Compare commits

..

7 Commits

1136 changed files with 20839 additions and 443832 deletions

View File

@ -1,28 +0,0 @@
ARG UBUNTU_VERSION=22.04
# This needs to generally match the container host's environment.
ARG CUDA_VERSION=11.7.1
# Target the CUDA build image
ARG BASE_CUDA_DEV_CONTAINER=nvidia/cuda:${CUDA_VERSION}-devel-ubuntu${UBUNTU_VERSION}
FROM ${BASE_CUDA_DEV_CONTAINER} as build
# Unless otherwise specified, we make a fat build.
ARG CUDA_DOCKER_ARCH=all
RUN apt-get update && \
apt-get install -y build-essential git cmake libsdl2-dev wget git
WORKDIR /app
COPY . .
# Set nvcc architecture
ENV CUDA_DOCKER_ARCH=${CUDA_DOCKER_ARCH}
# Enable cuBLAS
ENV GGML_CUDA=1
RUN make base.en
ENTRYPOINT ["/app/main"]

View File

@ -1,40 +0,0 @@
ARG UBUNTU_VERSION=22.04
# This needs to generally match the container host's environment.
ARG CUDA_VERSION=12.3.1
# Target the CUDA build image
ARG BASE_CUDA_DEV_CONTAINER=nvidia/cuda:${CUDA_VERSION}-devel-ubuntu${UBUNTU_VERSION}
# Target the CUDA runtime image
ARG BASE_CUDA_RUN_CONTAINER=nvidia/cuda:${CUDA_VERSION}-runtime-ubuntu${UBUNTU_VERSION}
FROM ${BASE_CUDA_DEV_CONTAINER} AS build
WORKDIR /app
# Unless otherwise specified, we make a fat build.
ARG CUDA_DOCKER_ARCH=all
# Set nvcc architecture
ENV CUDA_DOCKER_ARCH=${CUDA_DOCKER_ARCH}
RUN apt-get update && \
apt-get install -y build-essential libsdl2-dev wget cmake git \
&& rm -rf /var/lib/apt/lists/* /var/cache/apt/archives/*
# Ref: https://stackoverflow.com/a/53464012
ENV CUDA_MAIN_VERSION=12.3
ENV LD_LIBRARY_PATH /usr/local/cuda-${CUDA_MAIN_VERSION}/compat:$LD_LIBRARY_PATH
COPY .. .
# Enable cuBLAS
RUN make base.en CMAKE_ARGS="-DGGML_CUDA=1"
FROM ${BASE_CUDA_RUN_CONTAINER} AS runtime
ENV CUDA_MAIN_VERSION=12.3
ENV LD_LIBRARY_PATH /usr/local/cuda-${CUDA_MAIN_VERSION}/compat:$LD_LIBRARY_PATH
WORKDIR /app
RUN apt-get update && \
apt-get install -y curl ffmpeg wget cmake git \
&& rm -rf /var/lib/apt/lists/* /var/cache/apt/archives/*
COPY --from=build /app /app
ENV PATH=/app/build/bin:$PATH
ENTRYPOINT [ "bash", "-c" ]

View File

@ -1,28 +0,0 @@
ARG ONEAPI_VERSION=2025.1.1-0-devel-ubuntu24.04
FROM intel/oneapi-basekit:$ONEAPI_VERSION AS build
WORKDIR /app
RUN apt-get update && \
apt-get install -y build-essential libsdl2-dev wget cmake git \
&& rm -rf /var/lib/apt/lists/* /var/cache/apt/archives/*
COPY .. .
# Enable SYCL
ARG GGML_SYCL_F16=OFF
RUN if [ "${GGML_SYCL_F16}" = "ON" ]; then \
echo "GGML_SYCL_F16 is set" \
&& export OPT_SYCL_F16="-DGGML_SYCL_F16=ON"; \
fi && \
make base.en CMAKE_ARGS="-DGGML_SYCL=1 -DCMAKE_C_COMPILER=icx -DCMAKE_CXX_COMPILER=icpx ${OPT_SYCL_F16}"
FROM intel/oneapi-basekit:$ONEAPI_VERSION AS runtime
WORKDIR /app
RUN apt-get update && \
apt-get install -y curl ffmpeg libsdl2-dev wget cmake git \
&& rm -rf /var/lib/apt/lists/* /var/cache/apt/archives/*
COPY --from=build /app /app
ENV PATH=/app/build/bin:$PATH
ENTRYPOINT [ "bash", "-c" ]

View File

@ -1,39 +0,0 @@
ARG UBUNTU_VERSION=22.04
# This needs to generally match the container host's environment.
ARG MUSA_VERSION=rc4.0.1
# Target the MUSA build image
ARG BASE_MUSA_DEV_CONTAINER=mthreads/musa:${MUSA_VERSION}-mudnn-devel-ubuntu${UBUNTU_VERSION}
# Target the MUSA runtime image
ARG BASE_MUSA_RUN_CONTAINER=mthreads/musa:${MUSA_VERSION}-mudnn-runtime-ubuntu${UBUNTU_VERSION}
FROM ${BASE_MUSA_DEV_CONTAINER} AS build
WORKDIR /app
RUN apt-get update && \
apt-get install -y build-essential libsdl2-dev wget cmake git && \
apt-get clean && \
rm -rf /var/lib/apt/lists/* /var/cache/apt/archives/* /tmp/* /var/tmp/*
COPY .. .
# Enable muBLAS
RUN make base.en CMAKE_ARGS="-DGGML_MUSA=1"
RUN find /app/build -name "*.o" -delete && \
find /app/build -name "*.a" -delete && \
rm -rf /app/build/CMakeFiles && \
rm -rf /app/build/cmake_install.cmake && \
rm -rf /app/build/_deps
FROM ${BASE_MUSA_RUN_CONTAINER} AS runtime
WORKDIR /app
RUN apt-get update && \
apt-get install -y curl ffmpeg wget cmake git && \
apt-get clean && \
rm -rf /var/lib/apt/lists/* /var/cache/apt/archives/* /tmp/* /var/tmp/*
COPY --from=build /app /app
RUN du -sh /app/*
RUN find /app -type f -size +100M
ENV PATH=/app/build/bin:$PATH
ENTRYPOINT [ "bash", "-c" ]

View File

@ -1,20 +0,0 @@
FROM ubuntu:22.04 AS build
WORKDIR /app
RUN apt-get update && \
apt-get install -y build-essential wget cmake git \
&& rm -rf /var/lib/apt/lists/* /var/cache/apt/archives/*
COPY .. .
RUN make base.en
FROM ubuntu:22.04 AS runtime
WORKDIR /app
RUN apt-get update && \
apt-get install -y curl ffmpeg libsdl2-dev wget cmake git \
&& rm -rf /var/lib/apt/lists/* /var/cache/apt/archives/*
COPY --from=build /app /app
ENV PATH=/app/build/bin:$PATH
ENTRYPOINT [ "bash", "-c" ]

View File

@ -1,3 +0,0 @@
build*/
.github/
.devops/

View File

@ -1,22 +0,0 @@
name: Bindings Tests (Go)
on:
push:
paths:
- bindings/go/**
- whisper.h
pull_request:
paths:
- bindings/go/**
- whisper.h
jobs:
ubuntu-22:
runs-on: ubuntu-22.04
steps:
- uses: actions/setup-go@v5
with:
go-version: '^1.23'
- uses: actions/checkout@v4
- run: |
cd bindings/go
make test

View File

@ -1,21 +0,0 @@
name: Bindings Tests (Ruby)
on:
push:
branches:
- master
pull_request:
types: [opened, synchronize, reopened]
jobs:
ubuntu-22:
runs-on: ubuntu-22.04
defaults:
run:
working-directory: bindings/ruby
steps:
- uses: ruby/setup-ruby@v1
with:
ruby-version: '3.2'
- uses: actions/checkout@v4
- run: rake test

File diff suppressed because it is too large Load Diff

View File

@ -1,63 +0,0 @@
name: Publish Docker image
on:
pull_request:
push:
branches:
- master
jobs:
push_to_registry:
name: Push Docker image to Docker Hub
if: github.event.pull_request.draft == false
runs-on: ubuntu-22.04
env:
COMMIT_SHA: ${{ github.sha }}
strategy:
matrix:
config:
- { tag: "main", dockerfile: ".devops/main.Dockerfile", platform: "linux/amd64" }
- { tag: "main-musa", dockerfile: ".devops/main-musa.Dockerfile", platform: "linux/amd64" }
- { tag: "main-intel", dockerfile: ".devops/main-intel.Dockerfile", platform: "linux/amd64" }
#TODO: the cuda image keeps failing - disable for now
# https://github.com/ggerganov/whisper.cpp/actions/runs/11019444428/job/30602020339
#- { tag: "main-cuda", dockerfile: ".devops/main-cuda.Dockerfile", platform: "linux/amd64" }
steps:
- name: Check out the repo
uses: actions/checkout@v3
- name: Set up QEMU
uses: docker/setup-qemu-action@v3
with:
image: tonistiigi/binfmt:qemu-v7.0.0-28
- name: Set up Docker Buildx
uses: docker/setup-buildx-action@v3
- name: Log in to Docker Hub
uses: docker/login-action@v3
with:
registry: ghcr.io
username: ${{ github.repository_owner }}
password: ${{ secrets.GITHUB_TOKEN }}
- name: Build and push Docker image (versioned)
if: github.event_name == 'push'
uses: docker/build-push-action@v5
with:
context: .
push: true
platforms: ${{ matrix.config.platform }}
tags: "ghcr.io/${{ github.repository }}:${{ matrix.config.tag }}-${{ env.COMMIT_SHA }}"
file: ${{ matrix.config.dockerfile }}
- name: Build and push Docker image (tagged)
uses: docker/build-push-action@v4
with:
context: .
push: ${{ github.event_name == 'push' }}
platforms: ${{ matrix.config.platform }}
tags: "ghcr.io/${{ github.repository }}:${{ matrix.config.tag }}"
file: ${{ matrix.config.dockerfile }}

View File

@ -1,91 +0,0 @@
name: Examples WASM
on:
push:
branches: ["master"]
workflow_dispatch:
permissions:
contents: read
pages: write
id-token: write
concurrency:
group: "pages"
cancel-in-progress: false
jobs:
deploy-wasm-github-pages:
environment:
name: github-pages
url: ${{ steps.deployment.outputs.page_url }}
runs-on: ubuntu-latest
steps:
- name: Checkout
uses: actions/checkout@v4
- name: Setup Pages
uses: actions/configure-pages@v4
- name: Setup emsdk
uses: mymindstorm/setup-emsdk@v14
- name: Build WASM Examples
# Enable for real build later in whisper.cpp
run: |
mkdir -p build-em && cd build-em
emcmake cmake .. -DCMAKE_BUILD_TYPE=Release
make -j
- name: Create staging directory
run: mkdir -p staging
- name: Create .nojekyll file in staging directory
run: touch staging/.nojekyll
- name: Copy application files
run: |
build_dir=build-em/bin
ls ${build_dir}
# command.wasm
target_dir=staging/command.wasm
mkdir -p ${target_dir}
cp ${build_dir}/command.wasm/{index.html,command.js,helpers.js} ${target_dir}
cp ${build_dir}/libcommand.js ${target_dir}
# bench.wasm
target_dir=staging/bench.wasm
mkdir -p ${target_dir}
cp ${build_dir}/bench.wasm/{index.html,bench.js,helpers.js} ${target_dir}
cp ${build_dir}/libbench.js ${target_dir}
# stream.wasm
target_dir=staging/stream.wasm
mkdir -p ${target_dir}
cp ${build_dir}/stream.wasm/{index.html,stream.js,helpers.js} ${target_dir}
cp ${build_dir}/libstream.js ${target_dir}
# whisper.wasm (this will be the main example page)
target_dir=staging
mkdir -p ${target_dir}
cp ${build_dir}/whisper.wasm/{index.html,main.js,helpers.js} ${target_dir}
cp ${build_dir}/libmain.js ${target_dir}
# Copy Cross-Origin Isolation service worker
cp -v examples/coi-serviceworker.js staging/
- name: List files in staging directory (for debugging)
run: |
echo "Files in staging directory:"
find staging -type f | sort
- name: Upload artifact
uses: actions/upload-pages-artifact@v3
with:
path: ./staging
- name: Deploy to GitHub Pages
id: deployment
uses: actions/deploy-pages@v4

View File

@ -1,48 +0,0 @@
name: Examples Tests
on:
push:
paths:
- examples/addon.node/**
- whisper.h
pull_request:
paths:
- examples/addon.node/**
- whisper.h
jobs:
addon_node-ubuntu-22:
runs-on: ubuntu-22.04
strategy:
matrix:
node-version: [ 16.x, 18.x ]
steps:
- name: Clone
uses: actions/checkout@v1
- name: Dependencies
run: |
sudo apt-get update
sudo apt-get install build-essential git
sudo apt-get install cmake
sudo apt-get install libsdl2-dev
- name: Use Node.js ${{ matrix.node-version }}
uses: actions/setup-node@v1
with:
node-version: ${{ matrix.node-version }}
cache: 'npm'
- name: Install package.json dependencies
working-directory: ./examples/addon.node
run: npm install
- name: Compile addon.node
run: npx cmake-js compile -T addon.node -B Release
- name: Download test model
run: |
bash ./models/download-ggml-model.sh base.en
- name: Test
run: |
cd examples/addon.node
npm run test

59
.gitignore vendored
View File

@ -1,65 +1,22 @@
*.o
*.a
*.d
.cache/
.coreml/
.test/
.venv/
.vs/
.vscode/
.DS_Store
.vimspector.json
/CMakeSettings.json
/talk-llama.dSYM/
build/
build-*/
build_*/
build-em/
build-debug/
build-release/
build-sanitize-addr/
build-sanitize-thread/
# SPM
.build/
.swiftpm
*.metallib
ggml-metal-embed.metal
ggml-metal-embed.metal.tmp
/main
/stream
/command
/talk
/talk-llama
/bench
/quantize
/server
/lsp
arm_neon.h
main
stream
bench
sync.sh
libwhisper.a
libwhisper.so
compile_commands.json
examples/arm_neon.h
examples/whisper.objc/whisper.objc.xcodeproj/xcshareddata
examples/whisper.objc/whisper.objc.xcodeproj/xcuserdata/
examples/whisper.objc/whisper.objc.xcodeproj/project.xcworkspace/xcuserdata
extra/bench-gg.txt
models/*.mlmodel
models/*.mlmodelc
models/*.mlpackage
models/*-encoder-openvino.xml
models/*-encoder-openvino-cache/
bindings/java/.gradle/
bindings/java/.idea/
.idea/
benchmark_results.csv
cmake-build-debug/
.cxx/
.gradle/
local.properties
.log
.exe

510
AUTHORS
View File

@ -1,510 +0,0 @@
# date: Tue Feb 4 13:03:35 EET 2025
# this file is auto-generated by scripts/gen-authors.sh
0/0 <zero@imaskeleton.me>
0cc4m <picard12@live.de>
0xsourcecode <134374803+0xsourcecode@users.noreply.github.com>
65a <10104049+65a@users.noreply.github.com>
AIWintermuteAI <32562299+AIWintermuteAI@users.noreply.github.com>
AT <manyoso@users.noreply.github.com>
Aarni Koskela <akx@iki.fi>
Aaron Pham <29749331+aarnphm@users.noreply.github.com>
Aaron Taylor <aaron@exphat.com>
Abhilash Majumder <30946547+abhilash1910@users.noreply.github.com>
Abitofevrything <54505189+abitofevrything@users.noreply.github.com>
Adam Jones <domdomegg+git@gmail.com>
Adrien Gallouët <adrien@gallouet.fr>
Adrien Gallouët <angt@huggingface.co>
AfryMask <AfryMask@163.com>
Ahmad Bilal <ahmad.bilal@empglabs.com>
Ahmad Tameem <113388789+Tameem-10xE@users.noreply.github.com>
AidanBeltonS <87009434+AidanBeltonS@users.noreply.github.com>
AidanBeltonS <aidan.belton@codeplay.com>
Akarshan Biswas <akarshan.biswas@gmail.com>
Akarshan Biswas <akarshanbiswas@fedoraproject.org>
Akash Mahajan <akash7190@gmail.com>
Akash Mahajan <akashmjn@stanford.edu>
Al Hoang <3811822-hoanga@users.noreply.gitlab.com>
Alan <unknown>
Albert Jin <albert.jin@gmail.com>
Alberto Cabrera Pérez <alberto.cabrera@codeplay.com>
Alberto Cabrera Pérez <alberto.cabrera@intel.com>
Aleksander Andrzejewski <18704749+aleksanderandrzejewski@users.noreply.github.com>
Alex Azarov <alex@azarov.by>
Alex Bacart <13940752+alex-bacart@users.noreply.github.com>
Alex Evgrashin <aevgrashin@yandex.ru>
Alex O'Connell <35843486+acon96@users.noreply.github.com>
Alexandr Graschenkov <alexandr.graschenkov91@gmail.com>
Alexandru Mariuti <alex@mariuti.com>
Alexey Kharlamov <alexey@kharlamov.biz>
Alfredo Montesinos <alfredo.montesinos@g.austincc.edu>
Ali Alameh <ali.alameh@isae.edu.lb>
Alter <0x7c48@gmail.com>
Ananta Bastola <anantarajbastola@gmail.com>
Andreas Kieslinger <47689530+aendk@users.noreply.github.com>
Andreas Lubbe <git@lubbe.org>
Andreu Huguet <andreuhuguet@gmail.com>
Andrew Huynh <a5thuynh@gmail.com>
Andrew Minh Nguyen <40281306+amqdn@users.noreply.github.com>
Andrew S <andrews54757@gmail.com>
Andy Maloney <asmaloney@gmail.com>
Anton Kostin <masguit42@users.noreply.github.com>
Artyom Mezin <psycho.fading@gmail.com>
Asad Memon <asad.lionpk@gmail.com>
Ashraful Islam <ashraful.meche@gmail.com>
AsukaMinato <asukaminato@nyan.eu.org>
AustinMroz <austinmroz@utexas.edu>
Avik Sengupta <avik@sengupta.net>
Bader-eddine Ouaich <49657842+baderouaich@users.noreply.github.com>
Baffin Lee <baffinlee@gmail.com>
Ben Ashbaugh <ben.ashbaugh@intel.com>
Ben Nortier <bjnortier@gmail.com>
Benjamin Heiniger <benjamin.heiniger@bluewin.ch>
Bernhard M. Wiedemann <githubbmwprimary@lsmod.de>
Binozo <70137898+Binozo@users.noreply.github.com>
Bo-Yi Wu <appleboy.tw@gmail.com>
Boris Bliznioukov <blib@mail.com>
Borislav Stanimirov <b.stanimirov@abv.bg>
Brad Murray <59848399+bradmurray-dt@users.noreply.github.com>
Brian Murray <brian@bmurray.ca>
CRD716 <crd716@gmail.com>
Canis Lupus <Canis-UK@users.noreply.github.com>
Carlos Zoido <mrgalleta@gmail.com>
Carolinabanana <140120812+Carolinabanana@users.noreply.github.com>
CarterLi999 <664681047@qq.com>
ChangSeok Oh <shivamidow@users.noreply.github.com>
Changyeon Kim <cyzero.kim@samsung.com>
Chaoqun <27287694+OpenWaygate@users.noreply.github.com>
Charles Xu <63788048+chaxu01@users.noreply.github.com>
Charles Xu <charles.xu@arm.com>
Chen Xi <xi2.chen@intel.com>
Chen Xi <xixichen08@foxmail.com>
Chenguang Li <87689256+noemotiovon@users.noreply.github.com>
Chia-Hsiang Cheng <88014292+garychia@users.noreply.github.com>
Chidi Williams <williamschidi1@gmail.com>
Chris Elrod <elrodc@gmail.com>
Christian <12550267+iceychris@users.noreply.github.com>
Christian Kastner <ckk@kvr.at>
Clifford Heath <clifford.heath@gmail.com>
Clint Herron <hanclinto@gmail.com>
Colin <github@whoisc.cc>
Conrad Kramer <conrad@conradkramer.com>
Corey Earwood <iamcgn+github@gmail.com>
CrispStrobe <154636388+CrispStrobe@users.noreply.github.com>
DAN™ <dranger003@gmail.com>
DGdev91 <DGdev91@users.noreply.github.com>
Damian Czaja <trojan295@protonmail.com>
Dan Johansson <164997844+eddnjjn@users.noreply.github.com>
Dan Johansson <dan.johansson@arm.com>
Daniel Bevenius <daniel.bevenius@gmail.com>
Daniel Valdivia <18384552+dvaldivia@users.noreply.github.com>
Daniel Ziegenberg <daniel@ziegenberg.at>
Daniele <57776841+daniandtheweb@users.noreply.github.com>
Dave <dave-fl@users.noreply.github.com>
Dave Airlie <airlied@gmail.com>
Dave Airlie <airlied@redhat.com>
Daven Sanassy <daven@vochlea.co.uk>
David <dnhkng@gmail.com>
David Thorpe <djt@mutablelogic.com>
DavidKorczynski <david@adalogics.com>
Davidson Francis <davidsondfgl@gmail.com>
Dener Stassun <denerstassun@gmail.com>
Dibakar Gope <dibakar.gope@arm.com>
Didzis Gosko <didzis@users.noreply.github.com>
Diego Devesa <slarengh@gmail.com>
Digipom <admin@digipom.com>
Dimo <dimo@ieee.org>
Djip007 <3705339+Djip007@users.noreply.github.com>
Djip007 <djip.perois@free.fr>
Dody Suria Wijaya <dodysw@gmail.com>
Dou Xinpeng <15529241576@163.com>
Dou Xinpeng <81913537+Dou-Git@users.noreply.github.com>
Dr. Tom Murphy VII Ph.D <499244+tom7@users.noreply.github.com>
Duncan McConnell <ddmcconnell4@gmail.com>
Egor Egorov <me@egorfine.com>
Elkana Bardugo <ttv200@gmail.com>
Emmanuel Schmidbauer <eschmidbauer@gmail.com>
Engininja2 <139037756+Engininja2@users.noreply.github.com>
Eric Curtin <ericcurtin17@gmail.com>
Eric Swanson <eswanson@alloscomp.com>
Eric Tendian <erictendian@gmail.com>
Eric Zhang <34133756+EZForever@users.noreply.github.com>
Erik Scholz <Green-Sky@users.noreply.github.com>
Evan Jones <evan.q.jones@gmail.com>
Evan Martin <evan.martin@gmail.com>
Eve <139727413+netrunnereve@users.noreply.github.com>
Evgeny Kuznetsov <evgeny@kuznetsov.md>
F1L1P <78918286+F1L1Pv2@users.noreply.github.com>
Faisal Zaghloul <quic_fzaghlou@quicinc.com>
Fangjun Kuang <csukuangfj@gmail.com>
Felix <stenbackfelix@gmail.com>
Finn Voorhees <finnvoorhees@gmail.com>
FirstTimeEZ <179362031+FirstTimeEZ@users.noreply.github.com>
FlippFuzz <41221030+FlippFuzz@users.noreply.github.com>
Frankie Robertson <frankier@users.noreply.github.com>
Gang Chen <goncha@gmail.com>
Gavin Cai <gavin1818@hotmail.com>
George Hindle <george@georgehindle.com>
Georgi Gerganov <ggerganov@gmail.com>
Gilad S <7817232+giladgd@users.noreply.github.com>
Gilad S <giladgd@users.noreply.github.com>
Gilad S. <7817232+giladgd@users.noreply.github.com>
GitAritron <103900385+GitAritron@users.noreply.github.com>
GiviMAD <GiviMAD@users.noreply.github.com>
Gleicon Moraes <gleicon@gmail.com>
Gregor Jasny <gjasny@googlemail.com>
Guillaume Wenzek <gwenzek@users.noreply.github.com>
HY. Kelvin Lee <34256578+hykelvinlee42@users.noreply.github.com>
Halalaluyafail3 <55773281+Halalaluyafail3@users.noreply.github.com>
Hang <bebound@gmail.com>
Haus1 <haus.xda@gmail.com>
Herman Semenov <GermanAizek@yandex.ru>
HimariO <dsfhe49854@gmail.com>
Hong Bo PENG <penghb@cn.ibm.com>
Hrishikesh Barman <geekodour@users.noreply.github.com>
Hugo <hugo@whynothugo.nl>
Ian Bicking <ian@ianbicking.org>
Ian Bull <irbull@eclipsesource.com>
Ihar Hrachyshka <ihrachys@redhat.com>
Ikko Ashimine <eltociear@gmail.com>
Ikko Eltociear Ashimine <eltociear@gmail.com>
InconsolableCellist <23345188+InconsolableCellist@users.noreply.github.com>
Ismatulla Mansurov <47342870+sapoepsilon@users.noreply.github.com>
Ivan <nekotekina@gmail.com>
Ivan Filipov <159561759+vanaka11@users.noreply.github.com>
Ivan Gorin <ivangorin21@gmail.com>
Ivo von Putzer Reibegg <ivo.putzer@gmail.com>
JJ <103335846+computerscienceiscool@users.noreply.github.com>
Jack Mousseau <jmousseau@users.noreply.github.com>
JacobLinCool <jacoblincool@gmail.com>
Jakub Ráček <blizzcz@gmail.com>
Jared Van Bortel <jared@nomic.ai>
Jay Binks <jaybinks@gmail.com>
Jayant <jayantyadav202@gmail.com>
Jeff Bolz <jbolz@nvidia.com>
Jeroen Mostert <jeroen.mostert@cm.com>
Jhen-Jie Hong <developer@jhen.me>
Jhen-Jie Hong <iainst0409@gmail.com>
JidongZhang-THU <1119708529@qq.com>
Jo Liss <joliss42@gmail.com>
Joe Todd <joe.todd@codeplay.com>
Johan <jr.raffin@gmail.com>
Johannes Gäßler <johannesg@5d6.de>
John Balis <phobossystems@gmail.com>
JohnnyB <jboero@users.noreply.github.com>
Jonathan Soo <jcsoo@agora.com>
Jonno <1160532+razodactyl@users.noreply.github.com>
Joonas Pihlajamaa <joonas.pihlajamaa@iki.fi>
Jose <34888496+Jerry-Master@users.noreply.github.com>
Josh Bleecher Snyder <josharian@gmail.com>
Josscii <jossciiweiyi@gmail.com>
Judd <foldl@users.noreply.github.com>
Jumper775 <78500318+jumpers775@users.noreply.github.com>
Jun Hee Yoo <contact.jhyoo@gmail.com>
Junil Kim <logyourself@gmail.com>
Justina Cho <justcho5@gmail.com>
Justine Tunney <jtunney@gmail.com>
Justine Tunney <jtunney@mozilla.com>
KITAITI Makoto <KitaitiMakoto@gmail.com>
KP Kaiser <kirk@zothcorp.com>
Kamilake <exjang0@gmail.com>
Karol Kontny <82021046+kkontny@users.noreply.github.com>
Karthick <j.karthic2004@gmail.com>
Kartik Saranathan <278928+Kartiku@users.noreply.github.com>
Kasumi <90275229+kasumi-1@users.noreply.github.com>
Kawrakow <48489457+ikawrakow@users.noreply.github.com>
Kendrick Taylor <kendrick@circuitsix.com>
Kevin Brothaler <admin@digipom.com>
Kevin Gibbons <bakkot@gmail.com>
Konosuke Sakai <konosuke@konosuke.work>
Konstantin Zhuravlyov <konstantin.zhuravlyov@amd.com>
Kreijstal <rainb@tfwno.gf>
Kylin <56434533+KyL0N@users.noreply.github.com>
LBlue <153975653+lbluep@users.noreply.github.com>
Larry Battle <larry.battle.tech@gmail.com>
Laytan Laats <laytanlaats@hotmail.com>
Leo Moll <leo.moll@yeasoft.com>
Lexevolution <31176843+Lexevolution@users.noreply.github.com>
LittleLoli <26589867+WhichWho@users.noreply.github.com>
Lucas Zanek <57494138+LucasZNK@users.noreply.github.com>
Luis Herrera <herrera-luis@users.noreply.github.com>
Lukas Rist <glaslos@gmail.com>
M. A. Ali <73258591+MightyStud@users.noreply.github.com>
M. Eren Akbiyik <erenakbiyik@gmail.com>
Ma Mingfei <mingfei.ma@intel.com>
Maciek <maciek.mab122@gmail.com>
Mahesh Madhav <67384846+heshpdx@users.noreply.github.com>
Marcin Mielniczuk <marmistrz.dev@zoho.eu>
Mark Karpelès <MagicalTux@users.noreply.github.com>
Mark Zhuang <zhuangqiubin@gmail.com>
Markus Tavenrath <mtavenrath@users.noreply.github.com>
Martin Delille <martin@delille.org>
Martin Warnaar <martinwarnaar@gmail.com>
Masaya, Kato <62578291+msy-kato@users.noreply.github.com>
Matheus de Sousa <23645013+keyehzy@users.noreply.github.com>
Mathieu Baudier <mbaudier@argeo.org>
Mathijs de Bruin <mathijs@mathijsfietst.nl>
Matija Pevec <mightymatth@users.noreply.github.com>
Matt Stephenson <mstephenson6@users.noreply.github.com>
Max Krasnyansky <max.krasnyansky@gmail.com>
Max Krasnyansky <quic_maxk@quicinc.com>
Maximiliano Levi <8160966+maxilevi@users.noreply.github.com>
Meng, Hengyu <hengyu.meng@intel.com>
Mengqing Cao <cmq0113@163.com>
Michael Podvitskiy <podvitskiymichael@gmail.com>
Michael Rienstra <mrienstra@gmail.com>
Mikhail Grigorev <sleuthhound@gmail.com>
Mohammadreza Hendiani <hendiani.mohammadreza@gmail.com>
Mohit Agarwal <mohit@sdf.org>
Molly Sophia <mollysophia379@gmail.com>
Murilo Santana <mvrilo@gmail.com>
NETZkultur GmbH <mulholland@netzkultur.de>
Natsu <chino@hotococoa.moe>
Neil Chudleigh <nchudleigh@users.noreply.github.com>
Neo Zhang <14088817+arthw@users.noreply.github.com>
Neo Zhang Jianyu <jianyu.zhang@intel.com>
Neuman Vong <neuman.vong@gmail.com>
Nicholai Tukanov <nicholaitukanov@gmail.com>
Nicholas Albion <nalbion@yahoo.com>
Nico Bosshard <nico@bosshome.ch>
Nicolò Scipione <nicolo.scipione@codeplay.com>
Niels Mayer <Niels.Mayer@gmail.com>
Nikita Sarychev <42014488+sARY77@users.noreply.github.com>
Nikolaj Olsson <nikse.dk@gmail.com>
Okabintaro <103938900+Okabintaro@users.noreply.github.com>
Oleg Sidorov <me@whitebox.io>
Oleg Sidorov <oleg@sidorov.nl>
Olivier Chafik <ochafik@users.noreply.github.com>
Ondrej Kokes <ondrej.kokes@gmail.com>
Ouadie EL FAROUKI <ouadie.elfarouki@codeplay.com>
PAB <pierreantoine.bannier@gmail.com>
Paul Tsochantaris <ptsochantaris@icloud.com>
Pedro Probst <pprobst@insiberia.net>
Peng <hzp1024@qq.com>
Peter <peter277@users.noreply.github.com>
Philipp Zabel <philipp.zabel@gmail.com>
Philippe Normand <phil@base-art.net>
Philippe Normand <philn@igalia.com>
Plamen Minev <pacominev@gmail.com>
Prashant Vithule <119530321+Vithulep@users.noreply.github.com>
Przemysław Pawełczyk <przemoc@gmail.com>
Qianhe Chen <54462604+chenqianhe@users.noreply.github.com>
R0CKSTAR <xiaodong.ye@mthreads.com>
R0CKSTAR <yeahdongcn@gmail.com>
Radoslav Gerganov <rgerganov@gmail.com>
Radosław Gryta <radek.gryta@gmail.com>
Rahul Vadhyar <107788610+RahulVadhyar@users.noreply.github.com>
Raiya Araki <83504221+rai62@users.noreply.github.com>
Reinforce-II <fate@eastal.com>
Reinis Muiznieks <muiznieks.reinis@gmail.com>
RelatedTitle <r3latedtitle@gmail.com>
Rémy Oudompheng <oudomphe@phare.normalesup.org>
RhinoDevel <RhinoDevel@users.noreply.github.com>
Rich Jones <miserlou@gmail.com>
Robert Ormandi <52251610+ormandi@users.noreply.github.com>
Robin <robin.xw@hotmail.com>
Roddur Dasgupta <roddurd@gmail.com>
Roland Rabien <figbug@gmail.com>
Romain Biessy <romain.biessy@codeplay.com>
Ronsor <ronsor@ronsor.pw>
Rotem Dan <rotemdan@gmail.com>
Ryan Hitchman <hitchmanr@gmail.com>
Ryan Metcalfe <107415876+RyanMetcalfeInt8@users.noreply.github.com>
RyanChang <ftes90015@gmail.com>
SRHMorris <69468379+SRHMorris@users.noreply.github.com>
SXX <sxx1136965276@gmail.com>
Sacha Arbonel <sacha.arbonel@hotmail.fr>
Salman Faroz <stsfaroz@gmail.com>
Salvatore Mesoraca <s.mesoraca16@gmail.com>
Sam <49637763+Onlyartist9@users.noreply.github.com>
Sam Pullara <spullara@gmail.com>
Samuel Durante <44513615+samueldurantes@users.noreply.github.com>
Sanchit Gandhi <93869735+sanchit-gandhi@users.noreply.github.com>
Sandro Hanea <40202887+sandrohanea@users.noreply.github.com>
Sergio López <slp@redhat.com>
Sergio López <slp@sinrega.org>
Shanshan Shen <467638484@qq.com>
Shijie <821898965@qq.com>
Shupei Fan <dymarkfan@outlook.com>
Siddharth Ramakrishnan <srr2141@columbia.edu>
Sigbjørn Skjæret <sigbjorn.skjaeret@scala.com>
Simon Moisselin <simon.moisstoll@gmail.com>
Sindre Sorhus <sindresorhus@gmail.com>
Slava Primenko <primenko.s@gmail.com>
Srihari-mcw <96763064+Srihari-mcw@users.noreply.github.com>
Stavros Panakakis <53979866+Stavrospanakakis@users.noreply.github.com>
Stefan Sydow <s.sydow@heinlein-video.de>
Stefan Sydow <stefan@sydow.email>
Syahmi Azhar <prsyahmi@gmail.com>
Syed Jafri <syedjafri97@gmail.com>
Sơn Phan Trung <phantrungson17@gmail.com>
Taisei Mima <bhbstar.me@gmail.com>
Takeshi Inoue <inoue.takeshi@gmail.com>
Tamotsu Takahashi <ttakah+github@gmail.com>
Taras Glek <taras@thegp.com>
Tauseef Mohiuddin <35351464+tauseefmohammed2@users.noreply.github.com>
Thamster <Thamster@users.noreply.github.com>
Thijs Raymakers <thijs@raymakers.nl>
Thomas Fitzsimmons <fitzsim@fitzsim.org>
Tiago Fassoni <tiagofassoni@users.noreply.github.com>
Tienshiao Ma <tienshiao@tienshiao.org>
Tim Miller <drasticactions@users.noreply.github.com>
Timothy Cronin <40186632+4imothy@users.noreply.github.com>
Tobrun <tobrun.van.nuland@gmail.com>
Todd <taf2@users.noreply.github.com>
Toliver <teejae@gmail.com>
Tong Li <31761981+litongjava@users.noreply.github.com>
Tony Wasserka <4840017+neobrain@users.noreply.github.com>
Topping1 <78745143+Topping1@users.noreply.github.com>
Travis Cline <travis.cline@gmail.com>
UEXTM.com <84163508+uextm@users.noreply.github.com>
UsernamesLame <156965854+UsernamesLame@users.noreply.github.com>
Vadim Peretokin <vperetokin@hey.com>
Valentin Gosu <1454649+valenting@users.noreply.github.com>
Vin Misra <vinith@alum.mit.edu>
Vulcan <93451215+trholding@users.noreply.github.com>
WhiteOlivierus <36532695+WhiteOlivierus@users.noreply.github.com>
William Tambellini <william.tambellini@gmail.com>
William Tambellini <wtambellini@sdl.com>
Wilson Silva <wilson.dsigns@gmail.com>
Xiang (Kevin) Li <kevinli020508@gmail.com>
Xiao-Yong Jin <jinxiaoyong@gmail.com>
XiaotaoChen <chenxiaotao1234@gmail.com>
Xingchen Song(宋星辰) <xingchensong1996@163.com>
Xinpeng Dou <81913537+Dou-Git@users.noreply.github.com>
Xuan Son Nguyen <thichthat@gmail.com>
Yajing Tang <phillis@google.com>
Yang Shen <aplshenyang@gmail.com>
Yunès <jean.baptiste.yunes@free.fr>
Yuri Khrustalev <ykhrustalev@users.noreply.github.com>
Yusuf Redžić <48274562+redzic@users.noreply.github.com>
ZaBlazzingZephyrus <119159668+blazingzephyr@users.noreply.github.com>
Zhenwei Jin <109658203+kylo5aby@users.noreply.github.com>
Zhiyuan Li <lizhiyuan@uniartisan.com>
Zhiyuan Li <uniartisan2017@gmail.com>
Zigfrid Zvezdin <ziggerZZ@gmail.com>
Zollner <24618122+Zolliner@users.noreply.github.com>
a3sh <38979186+A3shTnT@users.noreply.github.com>
ag2s20150909 <19373730+ag2s20150909@users.noreply.github.com>
agray3 <agray3@users.noreply.github.com>
ai-at-home <149282006+ai-at-home@users.noreply.github.com>
aldorof <aldorof@users.noreply.github.com>
alonfaraj <alonfaraj@gmail.com>
amd-dwang <dong.wang@amd.com>
amritahs-ibm <amritahs@linux.vnet.ibm.com>
andypayne <apayne@gmail.com>
ardfork <134447697+ardfork@users.noreply.github.com>
arizhih <40765267+arizhih@users.noreply.github.com>
automaticcat <daogiatuank54@gmail.com>
bandoti <141645996+bandoti@users.noreply.github.com>
be-next <jerome.ramette@gmail.com>
bert hubert <bert@hubertnet.nl>
billyct <billy_allen@126.com>
bmwl <brian.marshall@tolko.com>
bobqianic <129547291+bobqianic@users.noreply.github.com>
bocytko <bocytko+github@gmail.com>
boolemancer <48014766+boolemancer@users.noreply.github.com>
boolemancer <boolemancer@gmail.com>
bradmit <151883577+bradmit@users.noreply.github.com>
brunofaustino <b.fa.amorim@gmail.com>
bssrdf <merlintiger@hotmail.com>
byte-6174 <88070277+byte-6174@users.noreply.github.com>
cdosoftei <ciprian.dosoftei@gmail.com>
clach04 <Chris.Clark@actian.com>
compilade <113953597+compilade@users.noreply.github.com>
compilade <git@compilade.net>
conradg <conradjgodfrey@gmail.com>
crummyh <elijah@crums.us>
ddpasa <112642920+ddpasa@users.noreply.github.com>
denersc <denerstassun@gmail.com>
dscripka <dscripka@users.noreply.github.com>
duthils <duthils@duthils.net>
ecneladis <ecneladis@users.noreply.github.com>
faker <nspyia2002@gmail.com>
fitzsim <fitzsim@fitzsim.org>
fj-y-saito <85871716+fj-y-saito@users.noreply.github.com>
fraxy-v <65565042+fraxy-v@users.noreply.github.com>
genevera (she/her) <genevera@users.noreply.github.com>
geniusnut <geniusnut@gmail.com>
gilbertgong <gilbert.gong@gmail.com>
gn64 <yukikaze.jp@gmail.com>
goldwaving <77494627+goldwaving@users.noreply.github.com>
greeshmay <greeshmay@gmail.com>
haopeng <657407891@qq.com>
hipudding <huafengchun@gmail.com>
hsinhoyeh <yhh92u@gmail.com>
hydai <z54981220@gmail.com>
iamthad <thadeus.j.fleming@gmail.com>
issixx <46835150+issixx@users.noreply.github.com>
james wolf <contractorwolf@hotmail.com>
jdomke <28772296+jdomke@users.noreply.github.com>
jettoblack <jettoblack@gmail.com>
jiez <373447296@qq.com>
joecryptotoo <80373433+joecryptotoo@users.noreply.github.com>
jorismertz <35079666+jorismertz@users.noreply.github.com>
junchao-loongson <68935141+junchao-loongson@users.noreply.github.com>
junkfood <69683722+JunkFood02@users.noreply.github.com>
jwijffels <jwijffels@bnosac.be>
k.h.lai <adrian.k.h.lai@outlook.com>
kamranjon <kamranjon@gmail.com>
katsu560 <katsu560oo-@docomo.ne.jp>
kennethge <57784063+kenneth-ge@users.noreply.github.com>
keyehzy <msamuel@aluno.puc-rio.br>
kunnis <kunnis@users.noreply.github.com>
l3utterfly <gc.pthzfoldr@gmail.com>
leejet <leejet714@gmail.com>
leo-pony <nengjunma@outlook.com>
lhez <quic_lih@quicinc.com>
litong <31761981+litongjava@users.noreply.github.com>
liuwei-git <14815172+liuwei-git@users.noreply.github.com>
lnyan <lkwq007@gmail.com>
luoyu-intel <yu.luo@intel.com>
m.bell <m.bell@techsmith.com>
mahorozte <41834471+mahorozte@users.noreply.github.com>
mashizora <30516315+mashizora@users.noreply.github.com>
matt23654 <matthew.webber@protonmail.com>
matteo <matteogeniaccio@yahoo.it>
mgrachten <maarten@grachten.eu>
mkiol <mkiol@users.noreply.github.com>
mky_coder <47767389+mkycoder@users.noreply.github.com>
novag <7754358+novag@users.noreply.github.com>
pajowu <pajowu@pajowu.de>
pengxin99 <pengxin.yuan@intel.com>
petterreinholdtsen <pere-github@hungry.com>
polarmoon <90010972+polarmoon@users.noreply.github.com>
rlapray <lapray.romain@gmail.com>
sandrohanea <40202887+sandrohanea@users.noreply.github.com>
semiformal-net <84111142+semiformal-net@users.noreply.github.com>
shibukazu <61775791+shibukazu@users.noreply.github.com>
shikokuchuo <53399081+shikokuchuo@users.noreply.github.com>
slaren <slarengh@gmail.com>
slashlib <slashlib@users.noreply.github.com>
snadampal <87143774+snadampal@users.noreply.github.com>
someone13574 <81528246+someone13574@users.noreply.github.com>
st-gr <38470677+st-gr@users.noreply.github.com>
stduhpf <stephduh@live.fr>
stormofice <58337328+stormofice@users.noreply.github.com>
texmex76 <40733439+texmex76@users.noreply.github.com>
thefinaldegree <thefinaldegree@gmail.com>
thewh1teagle <61390950+thewh1teagle@users.noreply.github.com>
toboil-features <160222185+toboil-features@users.noreply.github.com>
trixirt <trix@redhat.com>
ulatekh <ulatekh@yahoo.com>
undef <undefdev@gmail.com>
uvos <devnull@uvos.xyz>
uvos <philipp@uvos.xyz>
valVk <valVk@users.noreply.github.com>
venkr <venkateshrameshkumar+1@gmail.com>
vicalloy <zbirder@gmail.com>
wangshuai09 <391746016@qq.com>
woachk <24752637+woachk@users.noreply.github.com>
xctan <axunlei@gmail.com>
xdrudis <xavierdrudis@yahoo.es>
yuri@FreeBSD <yuri@FreeBSD>
zhangjixiong <code.zjx@gmail.com>
zhentaoyu <zhentao.yu@intel.com>
zhouwg <6889919+zhouwg@users.noreply.github.com>
zhouwg <zhouwg2000@gmail.com>
谢乃闻 <sienaiwun@users.noreply.github.com>
布客飞龙 <562826179@qq.com>
Артём Земляк <azemlyak@smart-consulting.ru>

View File

@ -1,32 +1,14 @@
cmake_minimum_required(VERSION 3.5) # for add_link_options and implicit target directories.
project("whisper.cpp" C CXX)
project("whisper.cpp" VERSION 1.7.5)
include(CheckIncludeFileCXX)
set(SOVERSION 1)
#set(CMAKE_WARN_DEPRECATED YES)
set(CMAKE_WARN_UNUSED_CLI YES)
set(CMAKE_EXPORT_COMPILE_COMMANDS ON)
if (NOT XCODE AND NOT MSVC AND NOT CMAKE_BUILD_TYPE)
set(CMAKE_BUILD_TYPE Release CACHE STRING "Build type" FORCE)
set_property(CACHE CMAKE_BUILD_TYPE PROPERTY STRINGS "Debug" "Release" "MinSizeRel" "RelWithDebInfo")
endif()
# Add path to modules
list(APPEND CMAKE_MODULE_PATH "${CMAKE_CURRENT_SOURCE_DIR}/cmake/")
cmake_minimum_required (VERSION 3.0)
project(whisper.cpp VERSION 1.0.0)
set(CMAKE_EXPORT_COMPILE_COMMANDS "on")
set(CMAKE_RUNTIME_OUTPUT_DIRECTORY ${CMAKE_BINARY_DIR}/bin)
set(CMAKE_INSTALL_RPATH "${CMAKE_INSTALL_PREFIX}/lib")
if (CMAKE_SOURCE_DIR STREQUAL CMAKE_CURRENT_SOURCE_DIR)
if(CMAKE_SOURCE_DIR STREQUAL CMAKE_CURRENT_SOURCE_DIR)
set(WHISPER_STANDALONE ON)
include(git-vars)
# configure project version
configure_file(${CMAKE_SOURCE_DIR}/bindings/javascript/package-tmpl.json ${CMAKE_SOURCE_DIR}/bindings/javascript/package.json @ONLY)
include(cmake/GitVars.cmake)
include(cmake/BuildTypes.cmake)
else()
set(WHISPER_STANDALONE OFF)
endif()
@ -35,16 +17,6 @@ if (EMSCRIPTEN)
set(BUILD_SHARED_LIBS_DEFAULT OFF)
option(WHISPER_WASM_SINGLE_FILE "whisper: embed WASM inside the generated whisper.js" ON)
# TODO: without these, we get the following error:
# wasm-ld: error: --shared-memory is disallowed by whisper.cpp.o because it was not compiled with 'atomics' or 'bulk-memory' features.
set(CMAKE_C_FLAGS "${CMAKE_C_FLAGS} -pthread")
set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} -pthread")
set(CMAKE_EXE_LINKER_FLAGS "${CMAKE_EXE_LINKER_FLAGS} -s TOTAL_STACK=5242880")
set(CMAKE_SHARED_LINKER_FLAGS "${CMAKE_SHARED_LINKER_FLAGS} -s TOTAL_STACK=5242880")
set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} -Wno-deprecated")
else()
if (MINGW)
set(BUILD_SHARED_LIBS_DEFAULT OFF)
@ -53,199 +25,195 @@ else()
endif()
endif()
option(BUILD_SHARED_LIBS "build shared libraries" ${BUILD_SHARED_LIBS_DEFAULT})
# options
#
# option list
#
option(BUILD_SHARED_LIBS "whisper: build shared libs" ${BUILD_SHARED_LIBS_DEFAULT})
# debug
option(WHISPER_ALL_WARNINGS "whisper: enable all compiler warnings" ON)
option(WHISPER_ALL_WARNINGS_3RD_PARTY "whisper: enable all compiler warnings in 3rd party libs" OFF)
option(WHISPER_ALL_WARNINGS "whisper: enable all compiler warnings" ON)
option(WHISPER_ALL_WARNINGS_3RD_PARTY "whisper: enable all compiler warnings in 3rd party libs" OFF)
# build
option(WHISPER_FATAL_WARNINGS "whisper: enable -Werror flag" OFF)
option(WHISPER_USE_SYSTEM_GGML "whisper: use system-installed GGML library" OFF)
option(WHISPER_SANITIZE_THREAD "whisper: enable thread sanitizer" OFF)
option(WHISPER_SANITIZE_ADDRESS "whisper: enable address sanitizer" OFF)
option(WHISPER_SANITIZE_UNDEFINED "whisper: enable undefined sanitizer" OFF)
option(WHISPER_BUILD_TESTS "whisper: build tests" ${WHISPER_STANDALONE})
option(WHISPER_BUILD_EXAMPLES "whisper: build examples" ${WHISPER_STANDALONE})
option(WHISPER_SUPPORT_SDL2 "whisper: support for libSDL2" OFF)
if (APPLE)
option(WHISPER_NO_ACCELERATE "whisper: disable Accelerate framework" OFF)
else()
option(WHISPER_SUPPORT_OPENBLAS "whisper: support for OpenBLAS" OFF)
endif()
option(WHISPER_PERF "whisper: enable perf timings" OFF)
# sanitizers
option(WHISPER_SANITIZE_THREAD "whisper: enable thread sanitizer" OFF)
option(WHISPER_SANITIZE_ADDRESS "whisper: enable address sanitizer" OFF)
option(WHISPER_SANITIZE_UNDEFINED "whisper: enable undefined sanitizer" OFF)
# extra artifacts
option(WHISPER_BUILD_TESTS "whisper: build tests" ${WHISPER_STANDALONE})
option(WHISPER_BUILD_EXAMPLES "whisper: build examples" ${WHISPER_STANDALONE})
option(WHISPER_BUILD_SERVER "whisper: build server example" ${WHISPER_STANDALONE})
# 3rd party libs
option(WHISPER_CURL "whisper: use libcurl to download model from an URL" OFF)
option(WHISPER_SDL2 "whisper: support for libSDL2" OFF)
if (CMAKE_SYSTEM_NAME MATCHES "Linux")
option(WHISPER_FFMPEG "whisper: support building and linking with ffmpeg libs (avcodec, swresample, ...)" OFF)
endif()
option(WHISPER_COREML "whisper: enable Core ML framework" OFF)
option(WHISPER_COREML_ALLOW_FALLBACK "whisper: allow non-CoreML fallback" OFF)
option(WHISPER_OPENVINO "whisper: support for OpenVINO" OFF)
# Required for relocatable CMake package
include(${CMAKE_CURRENT_SOURCE_DIR}/cmake/build-info.cmake)
# override ggml options
set(GGML_SANITIZE_THREAD ${WHISPER_SANITIZE_THREAD})
set(GGML_SANITIZE_ADDRESS ${WHISPER_SANITIZE_ADDRESS})
set(GGML_SANITIZE_UNDEFINED ${WHISPER_SANITIZE_UNDEFINED})
set(GGML_ALL_WARNINGS ${WHISPER_ALL_WARNINGS})
set(GGML_FATAL_WARNINGS ${WHISPER_FATAL_WARNINGS})
# transition helpers
function (whisper_option_depr TYPE OLD NEW)
if (${OLD})
message(${TYPE} "${OLD} is deprecated and will be removed in the future.\nUse ${NEW} instead\n")
set(${NEW} ON)
if (NOT MSVC)
if (WHISPER_SANITIZE_THREAD)
set(CMAKE_C_FLAGS "${CMAKE_C_FLAGS} -fsanitize=thread")
set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} -fsanitize=thread")
endif()
endfunction()
whisper_option_depr(FATAL_ERROR WHISPER_CUBLAS GGML_CUDA)
whisper_option_depr(WARNING WHISPER_CUDA GGML_CUDA)
whisper_option_depr(WARNING WHISPER_KOMPUTE GGML_KOMPUTE)
whisper_option_depr(WARNING WHISPER_METAL GGML_METAL)
whisper_option_depr(WARNING WHISPER_METAL_EMBED_LIBRARY GGML_METAL_EMBED_LIBRARY)
whisper_option_depr(WARNING WHISPER_NATIVE GGML_NATIVE)
whisper_option_depr(WARNING WHISPER_OPENMP GGML_OPENMP)
whisper_option_depr(WARNING WHISPER_RPC GGML_RPC)
whisper_option_depr(WARNING WHISPER_SYCL GGML_SYCL)
whisper_option_depr(WARNING WHISPER_SYCL_F16 GGML_SYCL_F16)
whisper_option_depr(WARNING WHISPER_CCACHE GGML_CCACHE)
if (WHISPER_SANITIZE_ADDRESS)
set(CMAKE_C_FLAGS "${CMAKE_C_FLAGS} -fsanitize=address -fno-omit-frame-pointer")
set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} -fsanitize=address -fno-omit-frame-pointer")
endif()
if (GGML_CUDA AND NOT MSVC)
#GGML_CUDA enabled, add the necessary compile options -Wno-deprecated-gpu-targets
add_compile_options(-Wno-deprecated-gpu-targets)
if (WHISPER_SANITIZE_UNDEFINED)
set(CMAKE_C_FLAGS "${CMAKE_C_FLAGS} -fsanitize=undefined")
set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} -fsanitize=undefined")
endif()
endif()
#
# build the library
#
#set(CMAKE_C_FLAGS "${CMAKE_C_FLAGS} -ffast-math")
#set(CMAKE_C_FLAGS "${CMAKE_C_FLAGS} -march=native")
if (NOT TARGET ggml)
if (WHISPER_USE_SYSTEM_GGML)
find_package(ggml REQUIRED)
if (NOT ggml_FOUND)
message(FATAL_ERROR "System-installed GGML library not found.")
endif()
add_library(ggml ALIAS ggml::ggml)
# dependencies
set(CMAKE_C_STANDARD 11)
set(CMAKE_CXX_STANDARD 20)
find_package(Threads REQUIRED)
# on APPLE - include Accelerate framework
if (APPLE AND NOT WHISPER_NO_ACCELERATE)
find_library(ACCELERATE_FRAMEWORK Accelerate)
if (ACCELERATE_FRAMEWORK)
message(STATUS "Accelerate framework found")
set(WHISPER_EXTRA_LIBS ${WHISPER_EXTRA_LIBS} ${ACCELERATE_FRAMEWORK})
set(WHISPER_EXTRA_FLAGS ${WHISPER_EXTRA_FLAGS} -DGGML_USE_ACCELERATE)
else()
add_subdirectory(ggml)
if(WIN32)
# The following adds a _DISABLE_CONSTEXPR_MUTEX_CONSTRUCTOR macro and is a workaround for
# the Windows C++ standard library which does not support constexpr mutexes.
# From the release notes://github.com/microsoft/STL/wiki/Changelog
# Disable constexpr mutex constructor on Windows
# Fixed mutex's constructor to be constexpr. #3824 #4000 #4339
# Note: Programs that aren't following the documented restrictions on binary compatibility may encounter
# null dereferences in mutex machinery. You must follow this rule:
# When you mix binaries built by different supported versions of the toolset, the Redistributable version
# must be at least as new as the latest toolset used by any app component.
# You can define _DISABLE_CONSTEXPR_MUTEX_CONSTRUCTOR as an escape hatch.
#
# Specifically to whisper.cpp this would cause a crash when using the Java bindings.
# resulting in a Invalid memory access error.
target_compile_definitions(ggml-base PRIVATE _DISABLE_CONSTEXPR_MUTEX_CONSTRUCTOR)
message(WARNING "Accelerate framework not found")
endif()
endif()
if (WHISPER_SUPPORT_OPENBLAS)
find_library(OPENBLAS_LIB openblas)
if (OPENBLAS_LIB)
message(STATUS "OpenBLAS found")
set(WHISPER_EXTRA_LIBS ${WHISPER_EXTRA_LIBS} ${OPENBLAS_LIB})
set(WHISPER_EXTRA_FLAGS ${WHISPER_EXTRA_FLAGS} -DGGML_USE_OPENBLAS)
else()
message(WARNING "OpenBLAS not found")
endif()
endif()
# compiler flags
if (NOT CMAKE_BUILD_TYPE AND NOT CMAKE_CONFIGURATION_TYPES)
set(CMAKE_BUILD_TYPE Release CACHE STRING "Build type" FORCE)
set_property(CACHE CMAKE_BUILD_TYPE PROPERTY STRINGS "Debug" "Release" "RelWithDebInfo")
endif ()
if (WHISPER_ALL_WARNINGS)
if (NOT MSVC)
set(CMAKE_C_FLAGS "${CMAKE_C_FLAGS} \
-Wall \
-Wextra \
-Wpedantic \
-Wshadow \
-Wcast-qual \
-Wstrict-prototypes \
-Wpointer-arith \
")
else()
# todo : msvc
endif()
endif()
if (NOT MSVC)
set(CMAKE_C_FLAGS "${CMAKE_C_FLAGS} -Werror=vla")
#set(CMAKE_C_FLAGS "${CMAKE_C_FLAGS} -fno-math-errno -ffinite-math-only -funsafe-math-optimizations")
endif()
message(STATUS "CMAKE_SYSTEM_PROCESSOR: ${CMAKE_SYSTEM_PROCESSOR}")
if (${CMAKE_SYSTEM_PROCESSOR} MATCHES "arm" OR ${CMAKE_SYSTEM_PROCESSOR} MATCHES "aarch64")
message(STATUS "ARM detected")
else()
message(STATUS "x86 detected")
if (MSVC)
set(CMAKE_C_FLAGS "${CMAKE_C_FLAGS} /arch:AVX2")
set(CMAKE_C_FLAGS_RELEASE "${CMAKE_C_FLAGS_RELEASE} /arch:AVX2")
else()
if (EMSCRIPTEN)
# we require support for WASM SIMD 128-bit
set(CMAKE_C_FLAGS "${CMAKE_C_FLAGS} -pthread -msimd128")
set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} -pthread")
else()
set(CMAKE_C_FLAGS "${CMAKE_C_FLAGS} -mavx -mavx2 -mfma -mf16c")
endif()
endif()
# ... otherwise assume ggml is added by a parent CMakeLists.txt
endif()
add_subdirectory(src)
if (WHISPER_PERF)
set(WHISPER_EXTRA_FLAGS ${WHISPER_EXTRA_FLAGS} -DGGML_PERF)
endif()
#
# install
# whisper - this is the main library of the project
#
include(GNUInstallDirs)
include(CMakePackageConfigHelpers)
set(TARGET whisper)
set(WHISPER_BUILD_NUMBER ${BUILD_NUMBER})
set(WHISPER_BUILD_COMMIT ${BUILD_COMMIT})
set(WHISPER_INSTALL_VERSION ${CMAKE_PROJECT_VERSION})
add_library(${TARGET}
ggml.c
whisper.cpp
)
set(WHISPER_INCLUDE_INSTALL_DIR ${CMAKE_INSTALL_INCLUDEDIR} CACHE PATH "Location of header files")
set(WHISPER_LIB_INSTALL_DIR ${CMAKE_INSTALL_LIBDIR} CACHE PATH "Location of library files")
set(WHISPER_BIN_INSTALL_DIR ${CMAKE_INSTALL_BINDIR} CACHE PATH "Location of binary files")
target_include_directories(${TARGET} PUBLIC
.
)
get_directory_property(WHISPER_TRANSIENT_DEFINES COMPILE_DEFINITIONS)
if (MSVC)
target_link_libraries(${TARGET} PRIVATE ${WHISPER_EXTRA_LIBS} ${CMAKE_THREAD_LIBS_INIT})
set_target_properties(whisper PROPERTIES PUBLIC_HEADER ${CMAKE_CURRENT_SOURCE_DIR}/include/whisper.h)
install(TARGETS whisper LIBRARY PUBLIC_HEADER)
set(WHISPER_EXTRA_FLAGS ${WHISPER_EXTRA_FLAGS} -D_CRT_SECURE_NO_WARNINGS)
else()
target_link_libraries(${TARGET} PRIVATE m ${WHISPER_EXTRA_LIBS} ${CMAKE_THREAD_LIBS_INIT})
endif()
configure_package_config_file(
${CMAKE_CURRENT_SOURCE_DIR}/cmake/whisper-config.cmake.in
${CMAKE_CURRENT_BINARY_DIR}/whisper-config.cmake
INSTALL_DESTINATION ${CMAKE_INSTALL_LIBDIR}/cmake/whisper
PATH_VARS
WHISPER_INCLUDE_INSTALL_DIR
WHISPER_LIB_INSTALL_DIR
WHISPER_BIN_INSTALL_DIR )
if (BUILD_SHARED_LIBS)
target_link_libraries(${TARGET} PUBLIC
${CMAKE_DL_LIBS}
)
write_basic_package_version_file(
${CMAKE_CURRENT_BINARY_DIR}/whisper-version.cmake
VERSION ${WHISPER_INSTALL_VERSION}
COMPATIBILITY SameMajorVersion)
target_compile_definitions(${TARGET} PUBLIC
WHISPER_SHARED
)
endif()
install(FILES ${CMAKE_CURRENT_BINARY_DIR}/whisper-config.cmake
${CMAKE_CURRENT_BINARY_DIR}/whisper-version.cmake
DESTINATION ${CMAKE_INSTALL_LIBDIR}/cmake/whisper)
target_compile_definitions(${TARGET} PUBLIC
${WHISPER_EXTRA_FLAGS}
)
configure_file(cmake/whisper.pc.in
"${CMAKE_CURRENT_BINARY_DIR}/whisper.pc"
@ONLY)
install(TARGETS ${TARGET}
LIBRARY DESTINATION lib
ARCHIVE DESTINATION lib/static
)
install(FILES "${CMAKE_CURRENT_BINARY_DIR}/whisper.pc"
DESTINATION lib/pkgconfig)
#
# bindings
#
add_subdirectory(bindings)
#
# programs, examples and tests
#
if (WHISPER_BUILD_TESTS AND NOT CMAKE_JS_VERSION)
include(CTest)
add_subdirectory(tests)
endif ()
if (WHISPER_BUILD_EXAMPLES)
add_subdirectory(examples)
endif()
if (MSVC)
set(MSVC_WARNING_FLAGS
/wd4101 # Unreferenced local variable
/wd4005 # Macro redefinition
/wd4065 # switch statement contains 'default' but no 'case' labels
/wd4267 # Conversion from 'size_t' to a smaller type, possible loss of data
/wd4244 # Conversion from one type to another type, possible loss of ata
/wd4805 # Unsafe mix of type
/wd4305 # Truncation from 'type1' to 'type2' (often double to float)
/wd4996 # Function or variable may be unsafe/deprecated
)
function(disable_msvc_warnings target_name)
if(TARGET ${target_name})
target_compile_options(${target_name} PRIVATE ${MSVC_WARNING_FLAGS})
endif()
endfunction()
if (WHISPER_STANDALONE)
if (WHISPER_BUILD_TESTS)
enable_testing()
add_subdirectory(tests)
endif ()
if (WHISPER_BUILD_EXAMPLES)
disable_msvc_warnings(whisper)
disable_msvc_warnings(common)
disable_msvc_warnings(common-sdl)
disable_msvc_warnings(lsp)
disable_msvc_warnings(wchess-core)
disable_msvc_warnings(whisper-command)
disable_msvc_warnings(whisper-cli)
disable_msvc_warnings(whisper-server)
disable_msvc_warnings(whisper-stream)
disable_msvc_warnings(whisper-talk-llama)
disable_msvc_warnings(whisper-bench)
disable_msvc_warnings(quantize)
disable_msvc_warnings(vad-speech-segments)
add_subdirectory(examples)
endif()
endif()
endif ()

View File

@ -1,6 +1,6 @@
MIT License
Copyright (c) 2023-2024 The ggml authors
Copyright (c) 2022 Georgi Gerganov
Permission is hereby granted, free of charge, to any person obtaining a copy
of this software and associated documentation files (the "Software"), to deal

145
Makefile
View File

@ -1,12 +1,122 @@
ifndef UNAME_S
UNAME_S := $(shell uname -s)
endif
ifndef UNAME_P
UNAME_P := $(shell uname -p)
endif
ifndef UNAME_M
UNAME_M := $(shell uname -m)
endif
# Mac OS + Arm can report x86_64
# ref: https://github.com/ggerganov/whisper.cpp/issues/66#issuecomment-1282546789
ifeq ($(UNAME_S),Darwin)
ifneq ($(UNAME_P),arm)
SYSCTL_M := $(shell sysctl -n hw.optional.arm64)
ifeq ($(SYSCTL_M),1)
# UNAME_P := arm
# UNAME_M := arm64
warn := $(warning Your arch is announced as x86_64, but it seems to actually be ARM64. Not fixing that can lead to bad performance. For more info see: https://github.com/ggerganov/whisper.cpp/issues/66\#issuecomment-1282546789)
endif
endif
endif
#
# Compile flags
#
CFLAGS = -I. -O3 -std=c11
CXXFLAGS = -I. -I./examples -O3 -std=c++11
LDFLAGS =
# OS specific
# TODO: support Windows
ifeq ($(UNAME_S),Linux)
CFLAGS += -pthread
CXXFLAGS += -pthread
endif
ifeq ($(UNAME_S),Darwin)
CFLAGS += -pthread
CXXFLAGS += -pthread
endif
ifeq ($(UNAME_S),FreeBSD)
CFLAGS += -pthread
CXXFLAGS += -pthread
endif
# Architecture specific
# TODO: probably these flags need to be tweaked on some architectures
# feel free to update the Makefile for your architecture and send a pull request or issue
ifeq ($(UNAME_M),x86_64)
# AVX 512
CFLAGS += -mavx512f -mfma -mf16c
# AVX 256
#CFLAGS += -mavx -mavx2 -mfma -mf16c
endif
ifeq ($(UNAME_M),amd64)
CFLAGS += -mavx -mavx2 -mfma -mf16c
endif
ifndef WHISPER_NO_ACCELERATE
# Mac M1 - include Accelerate framework
ifeq ($(UNAME_S),Darwin)
CFLAGS += -DGGML_USE_ACCELERATE
LDFLAGS += -framework Accelerate
endif
endif
ifneq ($(filter aarch64%,$(UNAME_M)),)
endif
ifneq ($(filter armv6%,$(UNAME_M)),)
# Raspberry Pi 1, 2, 3
CFLAGS += -mfpu=neon-fp-armv8 -mfp16-format=ieee -mno-unaligned-access
endif
ifneq ($(filter armv7%,$(UNAME_M)),)
# Raspberry Pi 4
CFLAGS += -mfpu=neon-fp-armv8 -mfp16-format=ieee -mno-unaligned-access -funsafe-math-optimizations
endif
ifneq ($(filter armv8%,$(UNAME_M)),)
# Raspberry Pi 4
CFLAGS += -mfp16-format=ieee -mno-unaligned-access
endif
#
# Build library + main
#
main: examples/main/main.cpp ggml.o whisper.o
$(CXX) $(CXXFLAGS) examples/main/main.cpp whisper.o ggml.o -o main $(LDFLAGS)
./main -h
ggml.o: ggml.c ggml.h
$(CC) $(CFLAGS) -c ggml.c -o ggml.o
whisper.o: whisper.cpp whisper.h
$(CXX) $(CXXFLAGS) -c whisper.cpp -o whisper.o
libwhisper.a: ggml.o whisper.o
$(AR) rcs libwhisper.a ggml.o whisper.o
clean:
rm -f *.o main stream bench libwhisper.a
#
# Examples
#
CC_SDL=`sdl2-config --cflags --libs`
stream: examples/stream/stream.cpp ggml.o whisper.o
$(CXX) $(CXXFLAGS) examples/stream/stream.cpp ggml.o whisper.o -o stream $(CC_SDL) $(LDFLAGS)
bench: examples/bench/bench.cpp ggml.o whisper.o
$(CXX) $(CXXFLAGS) examples/bench/bench.cpp ggml.o whisper.o -o bench $(LDFLAGS)
#
# Audio samples
#
.PHONY: build
build:
cmake -B build $(CMAKE_ARGS)
cmake --build build --config Release
# download a few audio samples into folder "./samples":
.PHONY: samples
samples:
@ -16,8 +126,12 @@ samples:
@wget --quiet --show-progress -O samples/gb1.ogg https://upload.wikimedia.org/wikipedia/commons/1/1f/George_W_Bush_Columbia_FINAL.ogg
@wget --quiet --show-progress -O samples/hp0.ogg https://upload.wikimedia.org/wikipedia/en/d/d4/En.henryfphillips.ogg
@wget --quiet --show-progress -O samples/mm1.wav https://cdn.openai.com/whisper/draft-20220913a/micro-machines.wav
@wget --quiet --show-progress -O samples/a13.mp3 https://upload.wikimedia.org/wikipedia/commons/transcoded/6/6f/Apollo13-wehaveaproblem.ogg/Apollo13-wehaveaproblem.ogg.mp3
@wget --quiet --show-progress -O samples/diffusion2023-07-03.flac https://archive.org/download/diffusion2023-07-03/diffusion2023-07-03.flac
@echo "Converting to 16-bit WAV ..."
@ffmpeg -loglevel -0 -y -i samples/gb0.ogg -ar 16000 -ac 1 -c:a pcm_s16le samples/gb0.wav
@ffmpeg -loglevel -0 -y -i samples/gb1.ogg -ar 16000 -ac 1 -c:a pcm_s16le samples/gb1.wav
@ffmpeg -loglevel -0 -y -i samples/hp0.ogg -ar 16000 -ac 1 -c:a pcm_s16le samples/hp0.wav
@ffmpeg -loglevel -0 -y -i samples/mm1.wav -ar 16000 -ac 1 -c:a pcm_s16le samples/mm0.wav
@rm samples/mm1.wav
#
# Models
@ -34,25 +148,20 @@ samples:
.PHONY: small
.PHONY: medium.en
.PHONY: medium
.PHONY: large-v1
.PHONY: large-v2
.PHONY: large-v3
.PHONY: large-v3-turbo
.PHONY: large
tiny.en tiny base.en base small.en small medium.en medium large-v1 large-v2 large-v3 large-v3-turbo:
tiny.en tiny base.en base small.en small medium.en medium large: main
bash ./models/download-ggml-model.sh $@
cmake -B build $(CMAKE_ARGS)
cmake --build build --config Release
@echo ""
@echo "==============================================="
@echo "Running $@ on all samples in ./samples ..."
@echo "==============================================="
@echo ""
@for f in samples/*.{flac,mp3,ogg,wav}; do \
echo "----------------------------------------------" ; \
echo "[+] Running $@ on $$f ... (run 'ffplay $$f' to listen)" ; \
@for f in samples/*.wav; do \
echo "----------------------------------------------" ; \
echo "[+] Running base.en on $$f ... (run 'ffplay $$f' to listen)" ; \
echo "----------------------------------------------" ; \
echo "" ; \
./build/bin/whisper-cli -m models/ggml-$@.bin -f $$f ; \
./main -m models/ggml-$@.bin -f $$f ; \
echo "" ; \
done

909
README.md

File diff suppressed because it is too large Load Diff

View File

@ -1,249 +0,0 @@
# whisper.cpp for SYCL
[Background](#background)
[OS](#os)
[Intel GPU](#intel-gpu)
[Linux](#linux)
[Environment Variable](#environment-variable)
[Known Issue](#known-issue)
[Todo](#todo)
## Background
SYCL is a higher-level programming model to improve programming productivity on various hardware accelerators—such as CPUs, GPUs, and FPGAs. It is a single-source embedded domain-specific language based on pure C++17.
oneAPI is a specification that is open and standards-based, supporting multiple architecture types including but not limited to GPU, CPU, and FPGA. The spec has both direct programming and API-based programming paradigms.
Intel uses the SYCL as direct programming language to support CPU, GPUs and FPGAs.
To avoid re-inventing the wheel, this code refers other code paths in llama.cpp (like OpenBLAS, cuBLAS, CLBlast). We use a open-source tool [SYCLomatic](https://github.com/oneapi-src/SYCLomatic) (Commercial release [Intel® DPC++ Compatibility Tool](https://www.intel.com/content/www/us/en/developer/tools/oneapi/dpc-compatibility-tool.html)) migrate to SYCL.
The whisper.cpp for SYCL is used to support Intel GPUs.
For Intel CPU, recommend to use whisper.cpp for X86 (Intel MKL build).
## OS
|OS|Status|Verified|
|-|-|-|
|Linux|Support|Ubuntu 22.04|
|Windows|Ongoing| |
## Intel GPU
|Intel GPU| Status | Verified Model|
|-|-|-|
|Intel Data Center Max Series| Support| Max 1550|
|Intel Data Center Flex Series| Support| Flex 170|
|Intel Arc Series| Support| Arc 770|
|Intel built-in Arc GPU| Support| built-in Arc GPU in Meteor Lake|
|Intel iGPU| Support| iGPU in i5-1250P, i7-1165G7|
## Linux
### Setup Environment
1. Install Intel GPU driver.
a. Please install Intel GPU driver by official guide: [Install GPU Drivers](https://dgpu-docs.intel.com/driver/installation.html).
Note: for iGPU, please install the client GPU driver.
b. Add user to group: video, render.
```
sudo usermod -aG render username
sudo usermod -aG video username
```
Note: re-login to enable it.
c. Check
```
sudo apt install clinfo
sudo clinfo -l
```
Output (example):
```
Platform #0: Intel(R) OpenCL Graphics
`-- Device #0: Intel(R) Arc(TM) A770 Graphics
Platform #0: Intel(R) OpenCL HD Graphics
`-- Device #0: Intel(R) Iris(R) Xe Graphics [0x9a49]
```
2. Install Intel® oneAPI Base toolkit.
a. Please follow the procedure in [Get the Intel® oneAPI Base Toolkit ](https://www.intel.com/content/www/us/en/developer/tools/oneapi/base-toolkit.html).
Recommend to install to default folder: **/opt/intel/oneapi**.
Following guide use the default folder as example. If you use other folder, please modify the following guide info with your folder.
b. Check
```
source /opt/intel/oneapi/setvars.sh
sycl-ls
```
There should be one or more level-zero devices. Like **[ext_oneapi_level_zero:gpu:0]**.
Output (example):
```
[opencl:acc:0] Intel(R) FPGA Emulation Platform for OpenCL(TM), Intel(R) FPGA Emulation Device OpenCL 1.2 [2023.16.10.0.17_160000]
[opencl:cpu:1] Intel(R) OpenCL, 13th Gen Intel(R) Core(TM) i7-13700K OpenCL 3.0 (Build 0) [2023.16.10.0.17_160000]
[opencl:gpu:2] Intel(R) OpenCL Graphics, Intel(R) Arc(TM) A770 Graphics OpenCL 3.0 NEO [23.30.26918.50]
[ext_oneapi_level_zero:gpu:0] Intel(R) Level-Zero, Intel(R) Arc(TM) A770 Graphics 1.3 [1.3.26918]
```
2. Build locally:
```
mkdir -p build
cd build
source /opt/intel/oneapi/setvars.sh
#for FP16
#cmake .. -DWHISPER_SYCL=ON -DCMAKE_C_COMPILER=icx -DCMAKE_CXX_COMPILER=icpx -DWHISPER_SYCL_F16=ON
#for FP32
cmake .. -DWHISPER_SYCL=ON -DCMAKE_C_COMPILER=icx -DCMAKE_CXX_COMPILER=icpx
#build example/main only
#cmake --build . --config Release --target main
#build all binary
cmake --build . --config Release -v
```
or
```
./examples/sycl/build.sh
```
Note:
- By default, it will build for all binary files. It will take more time. To reduce the time, we recommend to build for **example/main** only.
### Run
1. Put model file to folder **models**
2. Enable oneAPI running environment
```
source /opt/intel/oneapi/setvars.sh
```
3. List device ID
Run without parameter:
```
./build/bin/ls-sycl-device
or
./build/bin/main
```
Check the ID in startup log, like:
```
found 4 SYCL devices:
Device 0: Intel(R) Arc(TM) A770 Graphics, compute capability 1.3,
max compute_units 512, max work group size 1024, max sub group size 32, global mem size 16225243136
Device 1: Intel(R) FPGA Emulation Device, compute capability 1.2,
max compute_units 24, max work group size 67108864, max sub group size 64, global mem size 67065057280
Device 2: 13th Gen Intel(R) Core(TM) i7-13700K, compute capability 3.0,
max compute_units 24, max work group size 8192, max sub group size 64, global mem size 67065057280
Device 3: Intel(R) Arc(TM) A770 Graphics, compute capability 3.0,
max compute_units 512, max work group size 1024, max sub group size 32, global mem size 16225243136
```
|Attribute|Note|
|-|-|
|compute capability 1.3|Level-zero running time, recommended |
|compute capability 3.0|OpenCL running time, slower than level-zero in most cases|
4. Set device ID and execute whisper.cpp
Set device ID = 0 by **GGML_SYCL_DEVICE=0**
```
GGML_SYCL_DEVICE=0 ./build/bin/main -m models/ggml-base.en.bin -f samples/jfk.wav
```
or run by script:
```
./examples/sycl/run_whisper.sh
```
5. Check the device ID in output
Like:
```
Using device **0** (Intel(R) Arc(TM) A770 Graphics) as main device
```
## Environment Variable
#### Build
|Name|Value|Function|
|-|-|-|
|WHISPER_SYCL|ON (mandatory)|Enable build with SYCL code path. <br>For FP32/FP16, WHISPER_SYCL=ON is mandatory.|
|WHISPER_SYCL_F16|ON (optional)|Enable FP16 build with SYCL code path.For FP32, do not set it.|
|CMAKE_C_COMPILER|icx|Use icx compiler for SYCL code path|
|CMAKE_CXX_COMPILER|icpx|use icpx for SYCL code path|
#### Running
|Name|Value|Function|
|-|-|-|
|GGML_SYCL_DEVICE|0 (default) or 1|Set the device id used. Check the device ids by default running output|
|GGML_SYCL_DEBUG|0 (default) or 1|Enable log function by macro: GGML_SYCL_DEBUG|
## Known Issue
- Error: `error while loading shared libraries: libsycl.so.7: cannot open shared object file: No such file or directory`.
Miss to enable oneAPI running environment.
Install oneAPI base toolkit and enable it by: `source /opt/intel/oneapi/setvars.sh`.
- Hang during startup
llama.cpp use mmap as default way to read model file and copy to GPU. In some system, memcpy will be abnormal and block.
Solution: add **--no-mmap**.
## Todo
- Support to build in Windows.
- Support multiple cards.

View File

@ -1,19 +1,3 @@
if (EMSCRIPTEN)
add_subdirectory(javascript)
add_custom_command(
OUTPUT ${CMAKE_CURRENT_SOURCE_DIR}/javascript/publish.log
DEPENDS ${CMAKE_CURRENT_SOURCE_DIR}/javascript/whisper.js
DEPENDS ${CMAKE_CURRENT_SOURCE_DIR}/javascript/libwhisper.worker.js
DEPENDS ${CMAKE_CURRENT_SOURCE_DIR}/javascript/package.json
WORKING_DIRECTORY ${CMAKE_CURRENT_SOURCE_DIR}/javascript
COMMAND npm publish
COMMAND touch publish.log
COMMENT "Publishing npm module v${PROJECT_VERSION}"
VERBATIM
)
add_custom_target(publish-npm
DEPENDS javascript/publish.log
)
endif()

View File

@ -1,2 +0,0 @@
build
models

View File

@ -1,21 +0,0 @@
MIT License
Copyright (c) 2022 David Thorpe
Permission is hereby granted, free of charge, to any person obtaining a copy
of this software and associated documentation files (the "Software"), to deal
in the Software without restriction, including without limitation the rights
to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
copies of the Software, and to permit persons to whom the Software is
furnished to do so, subject to the following conditions:
The above copyright notice and this permission notice shall be included in all
copies or substantial portions of the Software.
THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
SOFTWARE.

View File

@ -1,71 +0,0 @@
ifndef UNAME_S
UNAME_S := $(shell uname -s)
endif
ifndef UNAME_P
UNAME_P := $(shell uname -p)
endif
ifndef UNAME_M
UNAME_M := $(shell uname -m)
endif
GGML_METAL_PATH_RESOURCES := $(abspath ../..)
BUILD_DIR := build_go
MODELS_DIR := models
EXAMPLES_DIR := $(wildcard examples/*)
INCLUDE_PATH := $(abspath ../../include):$(abspath ../../ggml/include)
LIBRARY_PATH := $(abspath ../../${BUILD_DIR}/src:$(abspath ../../${BUILD_DIR}/ggml/src))
ifeq ($(GGML_CUDA),1)
LIBRARY_PATH := $(LIBRARY_PATH):$(CUDA_PATH)/targets/$(UNAME_M)-linux/lib/
BUILD_FLAGS := -ldflags "-extldflags '-lcudart -lcuda -lcublas'"
endif
ifeq ($(UNAME_S),Darwin)
EXT_LDFLAGS := -framework Foundation -framework Metal -framework MetalKit
endif
all: clean whisper examples
whisper: mkdir
cmake -S ../.. -B ../../${BUILD_DIR} \
-DCMAKE_BUILD_TYPE=Release \
-DBUILD_SHARED_LIBS=OFF
cmake --build ../../${BUILD_DIR} --target whisper
test: model-small whisper modtidy
ifeq ($(UNAME_S),Darwin)
@C_INCLUDE_PATH=${INCLUDE_PATH} LIBRARY_PATH=${LIBRARY_PATH} GGML_METAL_PATH_RESOURCES=${GGML_METAL_PATH_RESOURCES} go test -ldflags "-extldflags '$(EXT_LDFLAGS)'" -v .
@C_INCLUDE_PATH=${INCLUDE_PATH} LIBRARY_PATH=${LIBRARY_PATH} GGML_METAL_PATH_RESOURCES=${GGML_METAL_PATH_RESOURCES} go test -ldflags "-extldflags '$(EXT_LDFLAGS)'" -v ./pkg/whisper/...
else
@C_INCLUDE_PATH=${INCLUDE_PATH} LIBRARY_PATH=${LIBRARY_PATH} go test -v .
@C_INCLUDE_PATH=${INCLUDE_PATH} LIBRARY_PATH=${LIBRARY_PATH} go test -v ./pkg/whisper/...
endif
examples: $(EXAMPLES_DIR)
model-small: mkdir examples/go-model-download
@${BUILD_DIR}/go-model-download -out models ggml-small.en.bin
$(EXAMPLES_DIR): mkdir whisper modtidy
@echo Build example $(notdir $@)
ifeq ($(UNAME_S),Darwin)
@C_INCLUDE_PATH=${INCLUDE_PATH} LIBRARY_PATH=${LIBRARY_PATH} GGML_METAL_PATH_RESOURCES=${GGML_METAL_PATH_RESOURCES} go build ${BUILD_FLAGS} -ldflags "-extldflags '$(EXT_LDFLAGS)'" -o ${BUILD_DIR}/$(notdir $@) ./$@
else
@C_INCLUDE_PATH=${INCLUDE_PATH} LIBRARY_PATH=${LIBRARY_PATH} go build ${BUILD_FLAGS} -o ${BUILD_DIR}/$(notdir $@) ./$@
endif
mkdir:
@echo Mkdir ${BUILD_DIR}
@install -d ${BUILD_DIR}
@echo Mkdir ${MODELS_DIR}
@install -d ${MODELS_DIR}
modtidy:
@go mod tidy
clean:
@echo Clean
@rm -fr $(BUILD_DIR)
@go clean

View File

@ -1,106 +0,0 @@
# Go bindings for Whisper
This package provides Go bindings for whisper.cpp. They have been tested on:
* Darwin (OS X) 12.6 on x64_64
* Debian Linux on arm64
* Fedora Linux on x86_64
The "low level" bindings are in the `bindings/go` directory and there is a more
Go-style package in the `bindings/go/pkg/whisper` directory. The most simple usage
is as follows:
```go
import (
"github.com/ggerganov/whisper.cpp/bindings/go/pkg/whisper"
)
func main() {
var modelpath string // Path to the model
var samples []float32 // Samples to process
// Load the model
model, err := whisper.New(modelpath)
if err != nil {
panic(err)
}
defer model.Close()
// Process samples
context, err := model.NewContext()
if err != nil {
panic(err)
}
if err := context.Process(samples, nil, nil, nil); err != nil {
return err
}
// Print out the results
for {
segment, err := context.NextSegment()
if err != nil {
break
}
fmt.Printf("[%6s->%6s] %s\n", segment.Start, segment.End, segment.Text)
}
}
```
## Building & Testing
In order to build, you need to have the Go compiler installed. You can get it from [here](https://golang.org/dl/). Run the tests with:
```bash
git clone https://github.com/ggml-org/whisper.cpp.git
cd whisper.cpp/bindings/go
make test
```
This will compile a static `libwhisper.a` in a `build` folder, download a model file, then run the tests. To build the examples:
```bash
make examples
```
To build using cuda support add `GGML_CUDA=1`:
```bash
GGML_CUDA=1 make examples
```
The examples are placed in the `build` directory. Once built, you can download all the models with the following command:
```bash
./build/go-model-download -out models
```
And you can then test a model against samples with the following command:
```bash
./build/go-whisper -model models/ggml-tiny.en.bin samples/jfk.wav
```
## Using the bindings
To use the bindings in your own software,
1. Import `github.com/ggerganov/whisper.cpp/bindings/go/pkg/whisper` (or `github.com/ggerganov/whisper.cpp/bindings/go` into your package;
2. Compile `libwhisper.a` (you can use `make whisper` in the `bindings/go` directory);
3. Link your go binary against whisper by setting the environment variables `C_INCLUDE_PATH` and `LIBRARY_PATH`
to point to the `whisper.h` file directory and `libwhisper.a` file directory respectively.
Look at the `Makefile` in the `bindings/go` directory for an example.
The API Documentation:
* https://pkg.go.dev/github.com/ggerganov/whisper.cpp/bindings/go
* https://pkg.go.dev/github.com/ggerganov/whisper.cpp/bindings/go/pkg/whisper
Getting help:
* Follow the discussion for the go bindings [here](https://github.com/ggml-org/whisper.cpp/discussions/312)
## License
The license for the Go bindings is the same as the license for the rest of the whisper.cpp project, which is the MIT License. See the `LICENSE` file for more details.

View File

@ -1,5 +0,0 @@
/*
github.com/ggml-org/whisper.cpp/bindings/go
provides a speech-to-text service bindings for the Go programming language.
*/
package whisper

View File

@ -1,31 +0,0 @@
package main
import (
"context"
"os"
"os/signal"
)
// ContextForSignal returns a context object which is cancelled when a signal
// is received. It returns nil if no signal parameter is provided
func ContextForSignal(signals ...os.Signal) context.Context {
if len(signals) == 0 {
return nil
}
ch := make(chan os.Signal, 1) // Buffered channel with space for 1 signal
ctx, cancel := context.WithCancel(context.Background())
// Send message on channel when signal received
signal.Notify(ch, signals...)
// When any signal is received, call cancel
go func() {
<-ch
cancel()
}()
// Return success
return ctx
}

View File

@ -1,304 +0,0 @@
package main
import (
"context"
"flag"
"fmt"
"io"
"net/http"
"net/url"
"os"
"path/filepath"
"strings"
"syscall"
"time"
)
///////////////////////////////////////////////////////////////////////////////
// CONSTANTS
const (
srcUrl = "https://huggingface.co/ggerganov/whisper.cpp/resolve/main/" // The location of the models
srcExt = ".bin" // Filename extension
bufSize = 1024 * 64 // Size of the buffer used for downloading the model
)
var (
// The models which will be downloaded, if no model is specified as an argument
modelNames = []string{
"tiny", "tiny-q5_1", "tiny-q8_0",
"tiny.en", "tiny.en-q5_1", "tiny.en-q8_0",
"base", "base-q5_1", "base-q8_0",
"base.en", "base.en-q5_1", "base.en-q8_0",
"small", "small-q5_1", "small-q8_0",
"small.en", "small.en-q5_1", "small.en-q8_0",
"medium", "medium-q5_0", "medium-q8_0",
"medium.en", "medium.en-q5_0", "medium.en-q8_0",
"large-v1",
"large-v2", "large-v2-q5_0", "large-v2-q8_0",
"large-v3", "large-v3-q5_0",
"large-v3-turbo", "large-v3-turbo-q5_0", "large-v3-turbo-q8_0",
}
)
var (
// The output folder. When not set, use current working directory.
flagOut = flag.String("out", "", "Output folder")
// HTTP timeout parameter - will timeout if takes longer than this to download a model
flagTimeout = flag.Duration("timeout", 30*time.Minute, "HTTP timeout")
// Quiet parameter - will not print progress if set
flagQuiet = flag.Bool("quiet", false, "Quiet mode")
)
///////////////////////////////////////////////////////////////////////////////
// MAIN
func main() {
flag.Usage = func() {
name := filepath.Base(flag.CommandLine.Name())
fmt.Fprintf(flag.CommandLine.Output(), `
Usage: %s [options] [<model>...]
Options:
-out string Specify the output folder where models will be saved.
Default: Current working directory.
-timeout duration Set the maximum duration for downloading a model.
Example: 10m, 1h (default: 30m0s).
-quiet Suppress all output except errors.
Examples:
1. Download a specific model:
%s -out ./models tiny-q8_0
2. Download all models:
%s -out ./models
`, name, name, name)
flag.PrintDefaults()
}
flag.Parse()
// Get output path
out, err := GetOut()
if err != nil {
fmt.Fprintln(os.Stderr, "Error:", err)
os.Exit(-1)
}
// Create context which quits on SIGINT or SIGQUIT
ctx := ContextForSignal(os.Interrupt, syscall.SIGQUIT)
// Progress filehandle
progress := os.Stdout
if *flagQuiet {
progress, err = os.Open(os.DevNull)
if err != nil {
fmt.Fprintln(os.Stderr, "Error:", err)
os.Exit(-1)
}
defer progress.Close()
}
// Download models - exit on error or interrupt
for _, model := range GetModels() {
url, err := URLForModel(model)
if err != nil {
fmt.Fprintln(os.Stderr, "Error:", err)
continue
} else if path, err := Download(ctx, progress, url, out); err == nil || err == io.EOF {
continue
} else if err == context.Canceled {
os.Remove(path)
fmt.Fprintln(progress, "\nInterrupted")
break
} else if err == context.DeadlineExceeded {
os.Remove(path)
fmt.Fprintln(progress, "Timeout downloading model")
continue
} else {
os.Remove(path)
fmt.Fprintln(os.Stderr, "Error:", err)
break
}
}
}
///////////////////////////////////////////////////////////////////////////////
// PUBLIC METHODS
// GetOut returns the path to the output directory
func GetOut() (string, error) {
if *flagOut == "" {
return os.Getwd()
}
if info, err := os.Stat(*flagOut); err != nil {
return "", err
} else if !info.IsDir() {
return "", fmt.Errorf("not a directory: %s", info.Name())
} else {
return *flagOut, nil
}
}
// GetModels returns the list of models to download
func GetModels() []string {
if flag.NArg() == 0 {
fmt.Println("No model specified.")
fmt.Println("Preparing to download all models...")
// Calculate total download size
fmt.Println("Calculating total download size...")
totalSize, err := CalculateTotalDownloadSize(modelNames)
if err != nil {
fmt.Println("Error calculating download sizes:", err)
os.Exit(1)
}
fmt.Println("View available models: https://huggingface.co/ggerganov/whisper.cpp/tree/main")
fmt.Printf("Total download size: %.2f GB\n", float64(totalSize)/(1024*1024*1024))
fmt.Println("Would you like to download all models? (y/N)")
// Prompt for user input
var response string
fmt.Scanln(&response)
if response != "y" && response != "Y" {
fmt.Println("Aborting. Specify a model to download.")
os.Exit(0)
}
return modelNames // Return all models if confirmed
}
return flag.Args() // Return specific models if arguments are provided
}
func CalculateTotalDownloadSize(models []string) (int64, error) {
var totalSize int64
client := http.Client{}
for _, model := range models {
modelURL, err := URLForModel(model)
if err != nil {
return 0, err
}
// Issue a HEAD request to get the file size
req, err := http.NewRequest("HEAD", modelURL, nil)
if err != nil {
return 0, err
}
resp, err := client.Do(req)
if err != nil {
return 0, err
}
resp.Body.Close()
if resp.StatusCode != http.StatusOK {
fmt.Printf("Warning: Unable to fetch size for %s (HTTP %d)\n", model, resp.StatusCode)
continue
}
size := resp.ContentLength
totalSize += size
}
return totalSize, nil
}
// URLForModel returns the URL for the given model on huggingface.co
func URLForModel(model string) (string, error) {
// Ensure "ggml-" prefix is added only once
if !strings.HasPrefix(model, "ggml-") {
model = "ggml-" + model
}
// Ensure ".bin" extension is added only once
if filepath.Ext(model) != srcExt {
model += srcExt
}
// Parse the base URL
url, err := url.Parse(srcUrl)
if err != nil {
return "", err
}
// Ensure no trailing slash in the base URL
url.Path = fmt.Sprintf("%s/%s", strings.TrimSuffix(url.Path, "/"), model)
return url.String(), nil
}
// Download downloads the model from the given URL to the given output directory
func Download(ctx context.Context, p io.Writer, model, out string) (string, error) {
// Create HTTP client
client := http.Client{
Timeout: *flagTimeout,
}
// Initiate the download
req, err := http.NewRequest("GET", model, nil)
if err != nil {
return "", err
}
resp, err := client.Do(req)
if err != nil {
return "", err
}
defer resp.Body.Close()
if resp.StatusCode != http.StatusOK {
return "", fmt.Errorf("%s: %s", model, resp.Status)
}
// If output file exists and is the same size as the model, skip
path := filepath.Join(out, filepath.Base(model))
if info, err := os.Stat(path); err == nil && info.Size() == resp.ContentLength {
fmt.Fprintln(p, "Skipping", model, "as it already exists")
return "", nil
}
// Create file
w, err := os.Create(path)
if err != nil {
return "", err
}
defer w.Close()
// Report
fmt.Fprintln(p, "Downloading", model, "to", out)
// Progressively download the model
data := make([]byte, bufSize)
count, pct := int64(0), int64(0)
ticker := time.NewTicker(5 * time.Second)
for {
select {
case <-ctx.Done():
// Cancelled, return error
return path, ctx.Err()
case <-ticker.C:
pct = DownloadReport(p, pct, count, resp.ContentLength)
default:
// Read body
n, err := resp.Body.Read(data)
if err != nil {
DownloadReport(p, pct, count, resp.ContentLength)
return path, err
} else if m, err := w.Write(data[:n]); err != nil {
return path, err
} else {
count += int64(m)
}
}
}
}
// Report periodically reports the download progress when percentage changes
func DownloadReport(w io.Writer, pct, count, total int64) int64 {
pct_ := count * 100 / total
if pct_ > pct {
fmt.Fprintf(w, " ...%d MB written (%d%%)\n", count/1e6, pct_)
}
return pct_
}

View File

@ -1,22 +0,0 @@
package main
import "fmt"
///////////////////////////////////////////////////////////////////////////////
// CONSTANTS
const (
Reset = "\033[0m"
RGBPrefix = "\033[38;5;" // followed by RGB values in decimal format separated by colons
RGBSuffix = "m"
)
///////////////////////////////////////////////////////////////////////////////
// PUBLIC METHODS
// Colorize text with RGB values, from 0 to 23
func Colorize(text string, v int) string {
// https://en.wikipedia.org/wiki/ANSI_escape_code#8-bit
// Grayscale colors are in the range 232-255
return RGBPrefix + fmt.Sprint(v%24+232) + RGBSuffix + text + Reset
}

View File

@ -1,147 +0,0 @@
package main
import (
"flag"
"fmt"
"strings"
"time"
// Packages
whisper "github.com/ggerganov/whisper.cpp/bindings/go/pkg/whisper"
)
///////////////////////////////////////////////////////////////////////////////
// TYPES
type Flags struct {
*flag.FlagSet
}
///////////////////////////////////////////////////////////////////////////////
// LIFECYCLE
func NewFlags(name string, args []string) (*Flags, error) {
flags := &Flags{
FlagSet: flag.NewFlagSet(name, flag.ContinueOnError),
}
// Register the command line arguments
registerFlags(flags)
// Parse command line
if err := flags.Parse(args); err != nil {
return nil, err
}
// Return success
return flags, nil
}
///////////////////////////////////////////////////////////////////////////////
// PUBLIC METHODS
func (flags *Flags) GetModel() string {
return flags.Lookup("model").Value.String()
}
func (flags *Flags) GetLanguage() string {
return flags.Lookup("language").Value.String()
}
func (flags *Flags) IsTranslate() bool {
return flags.Lookup("translate").Value.(flag.Getter).Get().(bool)
}
func (flags *Flags) GetOffset() time.Duration {
return flags.Lookup("offset").Value.(flag.Getter).Get().(time.Duration)
}
func (flags *Flags) GetDuration() time.Duration {
return flags.Lookup("duration").Value.(flag.Getter).Get().(time.Duration)
}
func (flags *Flags) GetThreads() uint {
return flags.Lookup("threads").Value.(flag.Getter).Get().(uint)
}
func (flags *Flags) GetOut() string {
return strings.ToLower(flags.Lookup("out").Value.String())
}
func (flags *Flags) IsTokens() bool {
return flags.Lookup("tokens").Value.String() == "true"
}
func (flags *Flags) IsColorize() bool {
return flags.Lookup("colorize").Value.String() == "true"
}
func (flags *Flags) GetMaxLen() uint {
return flags.Lookup("max-len").Value.(flag.Getter).Get().(uint)
}
func (flags *Flags) GetMaxTokens() uint {
return flags.Lookup("max-tokens").Value.(flag.Getter).Get().(uint)
}
func (flags *Flags) GetWordThreshold() float32 {
return float32(flags.Lookup("word-thold").Value.(flag.Getter).Get().(float64))
}
func (flags *Flags) SetParams(context whisper.Context) error {
if lang := flags.GetLanguage(); lang != "" && lang != "auto" {
fmt.Fprintf(flags.Output(), "Setting language to %q\n", lang)
if err := context.SetLanguage(lang); err != nil {
return err
}
}
if flags.IsTranslate() && context.IsMultilingual() {
fmt.Fprintf(flags.Output(), "Setting translate to true\n")
context.SetTranslate(true)
}
if offset := flags.GetOffset(); offset != 0 {
fmt.Fprintf(flags.Output(), "Setting offset to %v\n", offset)
context.SetOffset(offset)
}
if duration := flags.GetDuration(); duration != 0 {
fmt.Fprintf(flags.Output(), "Setting duration to %v\n", duration)
context.SetDuration(duration)
}
if threads := flags.GetThreads(); threads != 0 {
fmt.Fprintf(flags.Output(), "Setting threads to %d\n", threads)
context.SetThreads(threads)
}
if max_len := flags.GetMaxLen(); max_len != 0 {
fmt.Fprintf(flags.Output(), "Setting max_segment_length to %d\n", max_len)
context.SetMaxSegmentLength(max_len)
}
if max_tokens := flags.GetMaxTokens(); max_tokens != 0 {
fmt.Fprintf(flags.Output(), "Setting max_tokens to %d\n", max_tokens)
context.SetMaxTokensPerSegment(max_tokens)
}
if word_threshold := flags.GetWordThreshold(); word_threshold != 0 {
fmt.Fprintf(flags.Output(), "Setting word_threshold to %f\n", word_threshold)
context.SetTokenThreshold(word_threshold)
}
// Return success
return nil
}
///////////////////////////////////////////////////////////////////////////////
// PRIVATE METHODS
func registerFlags(flag *Flags) {
flag.String("model", "", "Path to the model file")
flag.String("language", "", "Spoken language")
flag.Bool("translate", false, "Translate from source language to english")
flag.Duration("offset", 0, "Time offset")
flag.Duration("duration", 0, "Duration of audio to process")
flag.Uint("threads", 0, "Number of threads to use")
flag.Uint("max-len", 0, "Maximum segment length in characters")
flag.Uint("max-tokens", 0, "Maximum tokens per segment")
flag.Float64("word-thold", 0, "Maximum segment score")
flag.Bool("tokens", false, "Display tokens")
flag.Bool("colorize", false, "Colorize tokens")
flag.String("out", "", "Output format (srt, none or leave as empty string)")
}

View File

@ -1,43 +0,0 @@
package main
import (
"flag"
"fmt"
"os"
"path/filepath"
// Packages
whisper "github.com/ggerganov/whisper.cpp/bindings/go/pkg/whisper"
)
func main() {
flags, err := NewFlags(filepath.Base(os.Args[0]), os.Args[1:])
if err == flag.ErrHelp {
os.Exit(0)
} else if err != nil {
fmt.Fprintln(os.Stderr, err)
os.Exit(1)
} else if flags.GetModel() == "" {
fmt.Fprintln(os.Stderr, "Use -model flag to specify which model file to use")
os.Exit(1)
} else if flags.NArg() == 0 {
fmt.Fprintln(os.Stderr, "No input files specified")
os.Exit(1)
}
// Load model
model, err := whisper.New(flags.GetModel())
if err != nil {
fmt.Fprintln(os.Stderr, err)
os.Exit(1)
}
defer model.Close()
// Process files
for _, filename := range flags.Args() {
if err := Process(model, filename, flags); err != nil {
fmt.Fprintln(os.Stderr, err)
continue
}
}
}

View File

@ -1,132 +0,0 @@
package main
import (
"fmt"
"io"
"os"
"time"
// Package imports
whisper "github.com/ggerganov/whisper.cpp/bindings/go/pkg/whisper"
wav "github.com/go-audio/wav"
)
func Process(model whisper.Model, path string, flags *Flags) error {
var data []float32
// Create processing context
context, err := model.NewContext()
if err != nil {
return err
}
// Set the parameters
if err := flags.SetParams(context); err != nil {
return err
}
fmt.Printf("\n%s\n", context.SystemInfo())
// Open the file
fmt.Fprintf(flags.Output(), "Loading %q\n", path)
fh, err := os.Open(path)
if err != nil {
return err
}
defer fh.Close()
// Decode the WAV file - load the full buffer
dec := wav.NewDecoder(fh)
if buf, err := dec.FullPCMBuffer(); err != nil {
return err
} else if dec.SampleRate != whisper.SampleRate {
return fmt.Errorf("unsupported sample rate: %d", dec.SampleRate)
} else if dec.NumChans != 1 {
return fmt.Errorf("unsupported number of channels: %d", dec.NumChans)
} else {
data = buf.AsFloat32Buffer().Data
}
// Segment callback when -tokens is specified
var cb whisper.SegmentCallback
if flags.IsTokens() {
cb = func(segment whisper.Segment) {
fmt.Fprintf(flags.Output(), "%02d [%6s->%6s] ", segment.Num, segment.Start.Truncate(time.Millisecond), segment.End.Truncate(time.Millisecond))
for _, token := range segment.Tokens {
if flags.IsColorize() && context.IsText(token) {
fmt.Fprint(flags.Output(), Colorize(token.Text, int(token.P*24.0)), " ")
} else {
fmt.Fprint(flags.Output(), token.Text, " ")
}
}
fmt.Fprintln(flags.Output(), "")
fmt.Fprintln(flags.Output(), "")
}
}
// Process the data
fmt.Fprintf(flags.Output(), " ...processing %q\n", path)
context.ResetTimings()
if err := context.Process(data, nil, cb, nil); err != nil {
return err
}
context.PrintTimings()
// Print out the results
switch {
case flags.GetOut() == "srt":
return OutputSRT(os.Stdout, context)
case flags.GetOut() == "none":
return nil
default:
return Output(os.Stdout, context, flags.IsColorize())
}
}
// Output text as SRT file
func OutputSRT(w io.Writer, context whisper.Context) error {
n := 1
for {
segment, err := context.NextSegment()
if err == io.EOF {
return nil
} else if err != nil {
return err
}
fmt.Fprintln(w, n)
fmt.Fprintln(w, srtTimestamp(segment.Start), " --> ", srtTimestamp(segment.End))
fmt.Fprintln(w, segment.Text)
fmt.Fprintln(w, "")
n++
}
}
// Output text to terminal
func Output(w io.Writer, context whisper.Context, colorize bool) error {
for {
segment, err := context.NextSegment()
if err == io.EOF {
return nil
} else if err != nil {
return err
}
fmt.Fprintf(w, "[%6s->%6s]", segment.Start.Truncate(time.Millisecond), segment.End.Truncate(time.Millisecond))
if colorize {
for _, token := range segment.Tokens {
if !context.IsText(token) {
continue
}
fmt.Fprint(w, " ", Colorize(token.Text, int(token.P*24.0)))
}
fmt.Fprint(w, "\n")
} else {
fmt.Fprintln(w, " ", segment.Text)
}
}
}
// Return srtTimestamp
func srtTimestamp(t time.Duration) string {
return fmt.Sprintf("%02d:%02d:%02d,%03d", t/time.Hour, (t%time.Hour)/time.Minute, (t%time.Minute)/time.Second, (t%time.Second)/time.Millisecond)
}

View File

@ -1,16 +0,0 @@
module github.com/ggerganov/whisper.cpp/bindings/go
go 1.23
require (
github.com/go-audio/wav v1.1.0
github.com/stretchr/testify v1.9.0
)
require (
github.com/davecgh/go-spew v1.1.1 // indirect
github.com/go-audio/audio v1.0.0 // indirect
github.com/go-audio/riff v1.0.0 // indirect
github.com/pmezard/go-difflib v1.0.0 // indirect
gopkg.in/yaml.v3 v3.0.1 // indirect
)

View File

@ -1,16 +0,0 @@
github.com/davecgh/go-spew v1.1.1 h1:vj9j/u1bqnvCEfJOwUhtlOARqs3+rkHYY13jYWTU97c=
github.com/davecgh/go-spew v1.1.1/go.mod h1:J7Y8YcW2NihsgmVo/mv3lAwl/skON4iLHjSsI+c5H38=
github.com/go-audio/audio v1.0.0 h1:zS9vebldgbQqktK4H0lUqWrG8P0NxCJVqcj7ZpNnwd4=
github.com/go-audio/audio v1.0.0/go.mod h1:6uAu0+H2lHkwdGsAY+j2wHPNPpPoeg5AaEFh9FlA+Zs=
github.com/go-audio/riff v1.0.0 h1:d8iCGbDvox9BfLagY94fBynxSPHO80LmZCaOsmKxokA=
github.com/go-audio/riff v1.0.0/go.mod h1:l3cQwc85y79NQFCRB7TiPoNiaijp6q8Z0Uv38rVG498=
github.com/go-audio/wav v1.1.0 h1:jQgLtbqBzY7G+BM8fXF7AHUk1uHUviWS4X39d5rsL2g=
github.com/go-audio/wav v1.1.0/go.mod h1:mpe9qfwbScEbkd8uybLuIpTgHyrISw/OTuvjUW2iGtE=
github.com/pmezard/go-difflib v1.0.0 h1:4DBwDE0NGyQoBHbLQYPwSUPoCMWR5BEzIk/f1lZbAQM=
github.com/pmezard/go-difflib v1.0.0/go.mod h1:iKH77koFhYxTK1pcRnkKkqfTogsbg7gZNVY4sRDYZ/4=
github.com/stretchr/testify v1.9.0 h1:HtqpIVDClZ4nwg75+f6Lvsy/wHu+3BoSGCbBAcpTsTg=
github.com/stretchr/testify v1.9.0/go.mod h1:r2ic/lqez/lEtzL7wO/rwa5dbSLXVDPFyf8C91i36aY=
gopkg.in/check.v1 v0.0.0-20161208181325-20d25e280405 h1:yhCVgyC4o1eVCa2tZl7eS0r+SDo693bJlVdllGtEeKM=
gopkg.in/check.v1 v0.0.0-20161208181325-20d25e280405/go.mod h1:Co6ibVJAznAaIkqp8huTwlJQCZ016jof/cbN4VW5Yz0=
gopkg.in/yaml.v3 v3.0.1 h1:fxVm/GzAzEWqLHuvctI91KS9hhNmmWOoWu0XTYJS7CA=
gopkg.in/yaml.v3 v3.0.1/go.mod h1:K4uyk7z7BCEPqu6E+C64Yfv1cQ7kz7rIZviUmN+EgEM=

View File

@ -1,204 +0,0 @@
package whisper
import (
"fmt"
)
///////////////////////////////////////////////////////////////////////////////
// CGO
/*
#include <whisper.h>
*/
import "C"
///////////////////////////////////////////////////////////////////////////////
// PUBLIC METHODS
func (p *Params) SetTranslate(v bool) {
p.translate = toBool(v)
}
func (p *Params) SetSplitOnWord(v bool) {
p.split_on_word = toBool(v)
}
func (p *Params) SetNoContext(v bool) {
p.no_context = toBool(v)
}
func (p *Params) SetSingleSegment(v bool) {
p.single_segment = toBool(v)
}
func (p *Params) SetPrintSpecial(v bool) {
p.print_special = toBool(v)
}
func (p *Params) SetPrintProgress(v bool) {
p.print_progress = toBool(v)
}
func (p *Params) SetPrintRealtime(v bool) {
p.print_realtime = toBool(v)
}
func (p *Params) SetPrintTimestamps(v bool) {
p.print_timestamps = toBool(v)
}
// Set language id
func (p *Params) SetLanguage(lang int) error {
if lang == -1 {
p.language = nil
return nil
}
str := C.whisper_lang_str(C.int(lang))
if str == nil {
return ErrInvalidLanguage
} else {
p.language = str
}
return nil
}
// Get language id
func (p *Params) Language() int {
if p.language == nil {
return -1
}
return int(C.whisper_lang_id(p.language))
}
// Threads available
func (p *Params) Threads() int {
return int(p.n_threads)
}
// Set number of threads to use
func (p *Params) SetThreads(threads int) {
p.n_threads = C.int(threads)
}
// Set start offset in ms
func (p *Params) SetOffset(offset_ms int) {
p.offset_ms = C.int(offset_ms)
}
// Set audio duration to process in ms
func (p *Params) SetDuration(duration_ms int) {
p.duration_ms = C.int(duration_ms)
}
// Set timestamp token probability threshold (~0.01)
func (p *Params) SetTokenThreshold(t float32) {
p.thold_pt = C.float(t)
}
// Set timestamp token sum probability threshold (~0.01)
func (p *Params) SetTokenSumThreshold(t float32) {
p.thold_ptsum = C.float(t)
}
// Set max segment length in characters
func (p *Params) SetMaxSegmentLength(n int) {
p.max_len = C.int(n)
}
func (p *Params) SetTokenTimestamps(b bool) {
p.token_timestamps = toBool(b)
}
// Set max tokens per segment (0 = no limit)
func (p *Params) SetMaxTokensPerSegment(n int) {
p.max_tokens = C.int(n)
}
// Set audio encoder context
func (p *Params) SetAudioCtx(n int) {
p.audio_ctx = C.int(n)
}
func (p *Params) SetMaxContext(n int) {
p.n_max_text_ctx = C.int(n)
}
func (p *Params) SetBeamSize(n int) {
p.beam_search.beam_size = C.int(n)
}
func (p *Params) SetEntropyThold(t float32) {
p.entropy_thold = C.float(t)
}
func (p *Params) SetTemperature(t float32) {
p.temperature = C.float(t)
}
// Sets the fallback temperature incrementation
// Pass -1.0 to disable this feature
func (p *Params) SetTemperatureFallback(t float32) {
p.temperature_inc = C.float(t)
}
// Set initial prompt
func (p *Params) SetInitialPrompt(prompt string) {
p.initial_prompt = C.CString(prompt)
}
///////////////////////////////////////////////////////////////////////////////
// PRIVATE METHODS
func toBool(v bool) C.bool {
if v {
return C.bool(true)
}
return C.bool(false)
}
///////////////////////////////////////////////////////////////////////////////
// STRINGIFY
func (p *Params) String() string {
str := "<whisper.params"
str += fmt.Sprintf(" strategy=%v", p.strategy)
str += fmt.Sprintf(" n_threads=%d", p.n_threads)
if p.language != nil {
str += fmt.Sprintf(" language=%s", C.GoString(p.language))
}
str += fmt.Sprintf(" n_max_text_ctx=%d", p.n_max_text_ctx)
str += fmt.Sprintf(" offset_ms=%d", p.offset_ms)
str += fmt.Sprintf(" duration_ms=%d", p.duration_ms)
str += fmt.Sprintf(" audio_ctx=%d", p.audio_ctx)
str += fmt.Sprintf(" initial_prompt=%s", C.GoString(p.initial_prompt))
str += fmt.Sprintf(" entropy_thold=%f", p.entropy_thold)
str += fmt.Sprintf(" temperature=%f", p.temperature)
str += fmt.Sprintf(" temperature_inc=%f", p.temperature_inc)
str += fmt.Sprintf(" beam_size=%d", p.beam_search.beam_size)
if p.translate {
str += " translate"
}
if p.no_context {
str += " no_context"
}
if p.single_segment {
str += " single_segment"
}
if p.print_special {
str += " print_special"
}
if p.print_progress {
str += " print_progress"
}
if p.print_realtime {
str += " print_realtime"
}
if p.print_timestamps {
str += " print_timestamps"
}
if p.token_timestamps {
str += " token_timestamps"
}
return str + ">"
}

View File

@ -1,28 +0,0 @@
package whisper
import (
"errors"
// Bindings
whisper "github.com/ggerganov/whisper.cpp/bindings/go"
)
///////////////////////////////////////////////////////////////////////////////
// ERRORS
var (
ErrUnableToLoadModel = errors.New("unable to load model")
ErrInternalAppError = errors.New("internal application error")
ErrProcessingFailed = errors.New("processing failed")
ErrUnsupportedLanguage = errors.New("unsupported language")
ErrModelNotMultilingual = errors.New("model is not multilingual")
)
///////////////////////////////////////////////////////////////////////////////
// CONSTANTS
// SampleRate is the sample rate of the audio data.
const SampleRate = whisper.SampleRate
// SampleBits is the number of bytes per sample.
const SampleBits = whisper.SampleBits

View File

@ -1,349 +0,0 @@
package whisper
import (
"fmt"
"io"
"runtime"
"strings"
"time"
// Bindings
whisper "github.com/ggerganov/whisper.cpp/bindings/go"
)
///////////////////////////////////////////////////////////////////////////////
// TYPES
type context struct {
n int
model *model
params whisper.Params
}
// Make sure context adheres to the interface
var _ Context = (*context)(nil)
///////////////////////////////////////////////////////////////////////////////
// LIFECYCLE
func newContext(model *model, params whisper.Params) (Context, error) {
context := new(context)
context.model = model
context.params = params
// Return success
return context, nil
}
///////////////////////////////////////////////////////////////////////////////
// PUBLIC METHODS
// Set the language to use for speech recognition.
func (context *context) SetLanguage(lang string) error {
if context.model.ctx == nil {
return ErrInternalAppError
}
if !context.model.IsMultilingual() {
return ErrModelNotMultilingual
}
if lang == "auto" {
context.params.SetLanguage(-1)
} else if id := context.model.ctx.Whisper_lang_id(lang); id < 0 {
return ErrUnsupportedLanguage
} else if err := context.params.SetLanguage(id); err != nil {
return err
}
// Return success
return nil
}
func (context *context) IsMultilingual() bool {
return context.model.IsMultilingual()
}
// Get language
func (context *context) Language() string {
id := context.params.Language()
if id == -1 {
return "auto"
}
return whisper.Whisper_lang_str(context.params.Language())
}
func (context *context) DetectedLanguage() string {
return whisper.Whisper_lang_str(context.model.ctx.Whisper_full_lang_id())
}
// Set translate flag
func (context *context) SetTranslate(v bool) {
context.params.SetTranslate(v)
}
func (context *context) SetSplitOnWord(v bool) {
context.params.SetSplitOnWord(v)
}
// Set number of threads to use
func (context *context) SetThreads(v uint) {
context.params.SetThreads(int(v))
}
// Set time offset
func (context *context) SetOffset(v time.Duration) {
context.params.SetOffset(int(v.Milliseconds()))
}
// Set duration of audio to process
func (context *context) SetDuration(v time.Duration) {
context.params.SetDuration(int(v.Milliseconds()))
}
// Set timestamp token probability threshold (~0.01)
func (context *context) SetTokenThreshold(t float32) {
context.params.SetTokenThreshold(t)
}
// Set timestamp token sum probability threshold (~0.01)
func (context *context) SetTokenSumThreshold(t float32) {
context.params.SetTokenSumThreshold(t)
}
// Set max segment length in characters
func (context *context) SetMaxSegmentLength(n uint) {
context.params.SetMaxSegmentLength(int(n))
}
// Set token timestamps flag
func (context *context) SetTokenTimestamps(b bool) {
context.params.SetTokenTimestamps(b)
}
// Set max tokens per segment (0 = no limit)
func (context *context) SetMaxTokensPerSegment(n uint) {
context.params.SetMaxTokensPerSegment(int(n))
}
// Set audio encoder context
func (context *context) SetAudioCtx(n uint) {
context.params.SetAudioCtx(int(n))
}
// Set maximum number of text context tokens to store
func (context *context) SetMaxContext(n int) {
context.params.SetMaxContext(n)
}
// Set Beam Size
func (context *context) SetBeamSize(n int) {
context.params.SetBeamSize(n)
}
// Set Entropy threshold
func (context *context) SetEntropyThold(t float32) {
context.params.SetEntropyThold(t)
}
// Set Temperature
func (context *context) SetTemperature(t float32) {
context.params.SetTemperature(t)
}
// Set the fallback temperature incrementation
// Pass -1.0 to disable this feature
func (context *context) SetTemperatureFallback(t float32) {
context.params.SetTemperatureFallback(t)
}
// Set initial prompt
func (context *context) SetInitialPrompt(prompt string) {
context.params.SetInitialPrompt(prompt)
}
// ResetTimings resets the mode timings. Should be called before processing
func (context *context) ResetTimings() {
context.model.ctx.Whisper_reset_timings()
}
// PrintTimings prints the model timings to stdout.
func (context *context) PrintTimings() {
context.model.ctx.Whisper_print_timings()
}
// SystemInfo returns the system information
func (context *context) SystemInfo() string {
return fmt.Sprintf("system_info: n_threads = %d / %d | %s\n",
context.params.Threads(),
runtime.NumCPU(),
whisper.Whisper_print_system_info(),
)
}
// Use mel data at offset_ms to try and auto-detect the spoken language
// Make sure to call whisper_pcm_to_mel() or whisper_set_mel() first.
// Returns the probabilities of all languages.
func (context *context) WhisperLangAutoDetect(offset_ms int, n_threads int) ([]float32, error) {
langProbs, err := context.model.ctx.Whisper_lang_auto_detect(offset_ms, n_threads)
if err != nil {
return nil, err
}
return langProbs, nil
}
// Process new sample data and return any errors
func (context *context) Process(
data []float32,
callEncoderBegin EncoderBeginCallback,
callNewSegment SegmentCallback,
callProgress ProgressCallback,
) error {
if context.model.ctx == nil {
return ErrInternalAppError
}
// If the callback is defined then we force on single_segment mode
if callNewSegment != nil {
context.params.SetSingleSegment(true)
}
// We don't do parallel processing at the moment
processors := 0
if processors > 1 {
if err := context.model.ctx.Whisper_full_parallel(context.params, data, processors, callEncoderBegin,
func(new int) {
if callNewSegment != nil {
num_segments := context.model.ctx.Whisper_full_n_segments()
s0 := num_segments - new
for i := s0; i < num_segments; i++ {
callNewSegment(toSegment(context.model.ctx, i))
}
}
}); err != nil {
return err
}
} else if err := context.model.ctx.Whisper_full(context.params, data, callEncoderBegin,
func(new int) {
if callNewSegment != nil {
num_segments := context.model.ctx.Whisper_full_n_segments()
s0 := num_segments - new
for i := s0; i < num_segments; i++ {
callNewSegment(toSegment(context.model.ctx, i))
}
}
}, func(progress int) {
if callProgress != nil {
callProgress(progress)
}
}); err != nil {
return err
}
// Return success
return nil
}
// Return the next segment of tokens
func (context *context) NextSegment() (Segment, error) {
if context.model.ctx == nil {
return Segment{}, ErrInternalAppError
}
if context.n >= context.model.ctx.Whisper_full_n_segments() {
return Segment{}, io.EOF
}
// Populate result
result := toSegment(context.model.ctx, context.n)
// Increment the cursor
context.n++
// Return success
return result, nil
}
// Test for text tokens
func (context *context) IsText(t Token) bool {
switch {
case context.IsBEG(t):
return false
case context.IsSOT(t):
return false
case whisper.Token(t.Id) >= context.model.ctx.Whisper_token_eot():
return false
case context.IsPREV(t):
return false
case context.IsSOLM(t):
return false
case context.IsNOT(t):
return false
default:
return true
}
}
// Test for "begin" token
func (context *context) IsBEG(t Token) bool {
return whisper.Token(t.Id) == context.model.ctx.Whisper_token_beg()
}
// Test for "start of transcription" token
func (context *context) IsSOT(t Token) bool {
return whisper.Token(t.Id) == context.model.ctx.Whisper_token_sot()
}
// Test for "end of transcription" token
func (context *context) IsEOT(t Token) bool {
return whisper.Token(t.Id) == context.model.ctx.Whisper_token_eot()
}
// Test for "start of prev" token
func (context *context) IsPREV(t Token) bool {
return whisper.Token(t.Id) == context.model.ctx.Whisper_token_prev()
}
// Test for "start of lm" token
func (context *context) IsSOLM(t Token) bool {
return whisper.Token(t.Id) == context.model.ctx.Whisper_token_solm()
}
// Test for "No timestamps" token
func (context *context) IsNOT(t Token) bool {
return whisper.Token(t.Id) == context.model.ctx.Whisper_token_not()
}
// Test for token associated with a specific language
func (context *context) IsLANG(t Token, lang string) bool {
if id := context.model.ctx.Whisper_lang_id(lang); id >= 0 {
return whisper.Token(t.Id) == context.model.ctx.Whisper_token_lang(id)
} else {
return false
}
}
///////////////////////////////////////////////////////////////////////////////
// PRIVATE METHODS
func toSegment(ctx *whisper.Context, n int) Segment {
return Segment{
Num: n,
Text: strings.TrimSpace(ctx.Whisper_full_get_segment_text(n)),
Start: time.Duration(ctx.Whisper_full_get_segment_t0(n)) * time.Millisecond * 10,
End: time.Duration(ctx.Whisper_full_get_segment_t1(n)) * time.Millisecond * 10,
Tokens: toTokens(ctx, n),
}
}
func toTokens(ctx *whisper.Context, n int) []Token {
result := make([]Token, ctx.Whisper_full_n_tokens(n))
for i := 0; i < len(result); i++ {
data := ctx.Whisper_full_get_token_data(n, i)
result[i] = Token{
Id: int(ctx.Whisper_full_get_token_id(n, i)),
Text: ctx.Whisper_full_get_token_text(n, i),
P: ctx.Whisper_full_get_token_p(n, i),
Start: time.Duration(data.T0()) * time.Millisecond * 10,
End: time.Duration(data.T1()) * time.Millisecond * 10,
}
}
return result
}

View File

@ -1,124 +0,0 @@
package whisper_test
import (
"os"
"testing"
"github.com/ggerganov/whisper.cpp/bindings/go/pkg/whisper"
"github.com/go-audio/wav"
assert "github.com/stretchr/testify/assert"
)
func TestSetLanguage(t *testing.T) {
assert := assert.New(t)
model, err := whisper.New(ModelPath)
assert.NoError(err)
assert.NotNil(model)
defer model.Close()
context, err := model.NewContext()
assert.NoError(err)
// This returns an error since
// the model 'models/ggml-small.en.bin'
// that is loaded is not multilingual
err = context.SetLanguage("en")
assert.Error(err)
}
func TestContextModelIsMultilingual(t *testing.T) {
assert := assert.New(t)
model, err := whisper.New(ModelPath)
assert.NoError(err)
assert.NotNil(model)
defer model.Close()
context, err := model.NewContext()
assert.NoError(err)
isMultilingual := context.IsMultilingual()
// This returns false since
// the model 'models/ggml-small.en.bin'
// that is loaded is not multilingual
assert.False(isMultilingual)
}
func TestLanguage(t *testing.T) {
assert := assert.New(t)
model, err := whisper.New(ModelPath)
assert.NoError(err)
assert.NotNil(model)
defer model.Close()
context, err := model.NewContext()
assert.NoError(err)
// This always returns en since
// the model 'models/ggml-small.en.bin'
// that is loaded is not multilingual
expectedLanguage := "en"
actualLanguage := context.Language()
assert.Equal(expectedLanguage, actualLanguage)
}
func TestProcess(t *testing.T) {
assert := assert.New(t)
fh, err := os.Open(SamplePath)
assert.NoError(err)
defer fh.Close()
// Decode the WAV file - load the full buffer
dec := wav.NewDecoder(fh)
buf, err := dec.FullPCMBuffer()
assert.NoError(err)
assert.Equal(uint16(1), dec.NumChans)
data := buf.AsFloat32Buffer().Data
model, err := whisper.New(ModelPath)
assert.NoError(err)
assert.NotNil(model)
defer model.Close()
context, err := model.NewContext()
assert.NoError(err)
err = context.Process(data, nil, nil, nil)
assert.NoError(err)
}
func TestDetectedLanguage(t *testing.T) {
assert := assert.New(t)
fh, err := os.Open(SamplePath)
assert.NoError(err)
defer fh.Close()
// Decode the WAV file - load the full buffer
dec := wav.NewDecoder(fh)
buf, err := dec.FullPCMBuffer()
assert.NoError(err)
assert.Equal(uint16(1), dec.NumChans)
data := buf.AsFloat32Buffer().Data
model, err := whisper.New(ModelPath)
assert.NoError(err)
assert.NotNil(model)
defer model.Close()
context, err := model.NewContext()
assert.NoError(err)
err = context.Process(data, nil, nil, nil)
assert.NoError(err)
expectedLanguage := "en"
actualLanguage := context.DetectedLanguage()
assert.Equal(expectedLanguage, actualLanguage)
}

View File

@ -1,4 +0,0 @@
/*
This is the higher-level speech-to-text whisper.cpp API for go
*/
package whisper

View File

@ -1,109 +0,0 @@
package whisper
import (
"io"
"time"
)
///////////////////////////////////////////////////////////////////////////////
// TYPES
// SegmentCallback is the callback function for processing segments in real
// time. It is called during the Process function
type SegmentCallback func(Segment)
// ProgressCallback is the callback function for reporting progress during
// processing. It is called during the Process function
type ProgressCallback func(int)
// EncoderBeginCallback is the callback function for checking if we want to
// continue processing. It is called during the Process function
type EncoderBeginCallback func() bool
// Model is the interface to a whisper model. Create a new model with the
// function whisper.New(string)
type Model interface {
io.Closer
// Return a new speech-to-text context.
NewContext() (Context, error)
// Return true if the model is multilingual.
IsMultilingual() bool
// Return all languages supported.
Languages() []string
}
// Context is the speech recognition context.
type Context interface {
SetLanguage(string) error // Set the language to use for speech recognition, use "auto" for auto detect language.
SetTranslate(bool) // Set translate flag
IsMultilingual() bool // Return true if the model is multilingual.
Language() string // Get language
DetectedLanguage() string // Get detected language
SetOffset(time.Duration) // Set offset
SetDuration(time.Duration) // Set duration
SetThreads(uint) // Set number of threads to use
SetSplitOnWord(bool) // Set split on word flag
SetTokenThreshold(float32) // Set timestamp token probability threshold
SetTokenSumThreshold(float32) // Set timestamp token sum probability threshold
SetMaxSegmentLength(uint) // Set max segment length in characters
SetTokenTimestamps(bool) // Set token timestamps flag
SetMaxTokensPerSegment(uint) // Set max tokens per segment (0 = no limit)
SetAudioCtx(uint) // Set audio encoder context
SetMaxContext(n int) // Set maximum number of text context tokens to store
SetBeamSize(n int) // Set Beam Size
SetEntropyThold(t float32) // Set Entropy threshold
SetInitialPrompt(prompt string) // Set initial prompt
SetTemperature(t float32) // Set temperature
SetTemperatureFallback(t float32) // Set temperature incrementation
// Process mono audio data and return any errors.
// If defined, newly generated segments are passed to the
// callback function during processing.
Process([]float32, EncoderBeginCallback, SegmentCallback, ProgressCallback) error
// After process is called, return segments until the end of the stream
// is reached, when io.EOF is returned.
NextSegment() (Segment, error)
IsBEG(Token) bool // Test for "begin" token
IsSOT(Token) bool // Test for "start of transcription" token
IsEOT(Token) bool // Test for "end of transcription" token
IsPREV(Token) bool // Test for "start of prev" token
IsSOLM(Token) bool // Test for "start of lm" token
IsNOT(Token) bool // Test for "No timestamps" token
IsLANG(Token, string) bool // Test for token associated with a specific language
IsText(Token) bool // Test for text token
// Timings
PrintTimings()
ResetTimings()
SystemInfo() string
}
// Segment is the text result of a speech recognition.
type Segment struct {
// Segment Number
Num int
// Time beginning and end timestamps for the segment.
Start, End time.Duration
// The text of the segment.
Text string
// The tokens of the segment.
Tokens []Token
}
// Token is a text or special token
type Token struct {
Id int
Text string
P float32
Start, End time.Duration
}

View File

@ -1,101 +0,0 @@
package whisper
import (
"fmt"
"os"
"runtime"
// Bindings
whisper "github.com/ggerganov/whisper.cpp/bindings/go"
)
///////////////////////////////////////////////////////////////////////////////
// TYPES
type model struct {
path string
ctx *whisper.Context
}
// Make sure model adheres to the interface
var _ Model = (*model)(nil)
///////////////////////////////////////////////////////////////////////////////
// LIFECYCLE
func New(path string) (Model, error) {
model := new(model)
if _, err := os.Stat(path); err != nil {
return nil, err
} else if ctx := whisper.Whisper_init(path); ctx == nil {
return nil, ErrUnableToLoadModel
} else {
model.ctx = ctx
model.path = path
}
// Return success
return model, nil
}
func (model *model) Close() error {
if model.ctx != nil {
model.ctx.Whisper_free()
}
// Release resources
model.ctx = nil
// Return success
return nil
}
///////////////////////////////////////////////////////////////////////////////
// STRINGIFY
func (model *model) String() string {
str := "<whisper.model"
if model.ctx != nil {
str += fmt.Sprintf(" model=%q", model.path)
}
return str + ">"
}
///////////////////////////////////////////////////////////////////////////////
// PUBLIC METHODS
// Return true if model is multilingual (language and translation options are supported)
func (model *model) IsMultilingual() bool {
return model.ctx.Whisper_is_multilingual() != 0
}
// Return all recognized languages. Initially it is set to auto-detect
func (model *model) Languages() []string {
result := make([]string, 0, whisper.Whisper_lang_max_id())
for i := 0; i < whisper.Whisper_lang_max_id(); i++ {
str := whisper.Whisper_lang_str(i)
if model.ctx.Whisper_lang_id(str) >= 0 {
result = append(result, str)
}
}
return result
}
func (model *model) NewContext() (Context, error) {
if model.ctx == nil {
return nil, ErrInternalAppError
}
// Create new context
params := model.ctx.Whisper_full_default_params(whisper.SAMPLING_GREEDY)
params.SetTranslate(false)
params.SetPrintSpecial(false)
params.SetPrintProgress(false)
params.SetPrintRealtime(false)
params.SetPrintTimestamps(false)
params.SetThreads(runtime.NumCPU())
params.SetNoContext(true)
// Return new context
return newContext(model, params)
}

View File

@ -1,91 +0,0 @@
package whisper_test
import (
"testing"
"github.com/ggerganov/whisper.cpp/bindings/go/pkg/whisper"
assert "github.com/stretchr/testify/assert"
)
func TestNew(t *testing.T) {
assert := assert.New(t)
t.Run("valid model path", func(t *testing.T) {
model, err := whisper.New(ModelPath)
assert.NoError(err)
assert.NotNil(model)
defer model.Close()
})
t.Run("invalid model path", func(t *testing.T) {
invalidModelPath := "invalid-model-path.bin"
model, err := whisper.New(invalidModelPath)
assert.Error(err)
assert.Nil(model)
})
}
func TestClose(t *testing.T) {
assert := assert.New(t)
model, err := whisper.New(ModelPath)
assert.NoError(err)
assert.NotNil(model)
err = model.Close()
assert.NoError(err)
}
func TestNewContext(t *testing.T) {
assert := assert.New(t)
model, err := whisper.New(ModelPath)
assert.NoError(err)
assert.NotNil(model)
defer model.Close()
context, err := model.NewContext()
assert.NoError(err)
assert.NotNil(context)
}
func TestIsMultilingual(t *testing.T) {
assert := assert.New(t)
model, err := whisper.New(ModelPath)
assert.NoError(err)
assert.NotNil(model)
defer model.Close()
isMultilingual := model.IsMultilingual()
// This returns false since
// the model 'models/ggml-small.en.bin'
// that is loaded is not multilingual
assert.False(isMultilingual)
}
func TestLanguages(t *testing.T) {
assert := assert.New(t)
model, err := whisper.New(ModelPath)
assert.NoError(err)
assert.NotNil(model)
defer model.Close()
expectedLanguages := []string{
"en", "zh", "de", "es", "ru", "ko", "fr", "ja", "pt", "tr", "pl",
"ca", "nl", "ar", "sv", "it", "id", "hi", "fi", "vi", "he", "uk",
"el", "ms", "cs", "ro", "da", "hu", "ta", "no", "th", "ur", "hr",
"bg", "lt", "la", "mi", "ml", "cy", "sk", "te", "fa", "lv", "bn",
"sr", "az", "sl", "kn", "et", "mk", "br", "eu", "is", "hy", "ne",
"mn", "bs", "kk", "sq", "sw", "gl", "mr", "pa", "si", "km", "sn",
"yo", "so", "af", "oc", "ka", "be", "tg", "sd", "gu", "am", "yi",
"lo", "uz", "fo", "ht", "ps", "tk", "nn", "mt", "sa", "lb", "my",
"bo", "tl", "mg", "as", "tt", "haw", "ln", "ha", "ba", "jw", "su",
}
actualLanguages := model.Languages()
assert.Equal(expectedLanguages, actualLanguages)
}

View File

@ -1,6 +0,0 @@
package whisper_test
const (
ModelPath = "../../models/ggml-small.en.bin"
SamplePath = "../../samples/jfk.wav"
)

Binary file not shown.

View File

@ -1,468 +0,0 @@
package whisper
import (
"errors"
"unsafe"
)
///////////////////////////////////////////////////////////////////////////////
// CGO
/*
#cgo LDFLAGS: -lwhisper -lggml -lggml-base -lggml-cpu -lm -lstdc++ -fopenmp
#cgo darwin LDFLAGS: -framework Accelerate -framework Metal -framework Foundation -framework CoreGraphics
#include <whisper.h>
#include <stdlib.h>
extern void callNewSegment(void* user_data, int new);
extern void callProgress(void* user_data, int progress);
extern bool callEncoderBegin(void* user_data);
// Text segment callback
// Called on every newly generated text segment
// Use the whisper_full_...() functions to obtain the text segments
static void whisper_new_segment_cb(struct whisper_context* ctx, struct whisper_state* state, int n_new, void* user_data) {
if(user_data != NULL && ctx != NULL) {
callNewSegment(user_data, n_new);
}
}
// Progress callback
// Called on every newly generated text segment
// Use the whisper_full_...() functions to obtain the text segments
static void whisper_progress_cb(struct whisper_context* ctx, struct whisper_state* state, int progress, void* user_data) {
if(user_data != NULL && ctx != NULL) {
callProgress(user_data, progress);
}
}
// Encoder begin callback
// If not NULL, called before the encoder starts
// If it returns false, the computation is aborted
static bool whisper_encoder_begin_cb(struct whisper_context* ctx, struct whisper_state* state, void* user_data) {
if(user_data != NULL && ctx != NULL) {
return callEncoderBegin(user_data);
}
return false;
}
// Get default parameters and set callbacks
static struct whisper_full_params whisper_full_default_params_cb(struct whisper_context* ctx, enum whisper_sampling_strategy strategy) {
struct whisper_full_params params = whisper_full_default_params(strategy);
params.new_segment_callback = whisper_new_segment_cb;
params.new_segment_callback_user_data = (void*)(ctx);
params.encoder_begin_callback = whisper_encoder_begin_cb;
params.encoder_begin_callback_user_data = (void*)(ctx);
params.progress_callback = whisper_progress_cb;
params.progress_callback_user_data = (void*)(ctx);
return params;
}
*/
import "C"
///////////////////////////////////////////////////////////////////////////////
// TYPES
type (
Context C.struct_whisper_context
Token C.whisper_token
TokenData C.struct_whisper_token_data
SamplingStrategy C.enum_whisper_sampling_strategy
Params C.struct_whisper_full_params
)
///////////////////////////////////////////////////////////////////////////////
// GLOBALS
const (
SAMPLING_GREEDY SamplingStrategy = C.WHISPER_SAMPLING_GREEDY
SAMPLING_BEAM_SEARCH SamplingStrategy = C.WHISPER_SAMPLING_BEAM_SEARCH
)
const (
SampleRate = C.WHISPER_SAMPLE_RATE // Expected sample rate, samples per second
SampleBits = uint16(unsafe.Sizeof(C.float(0))) * 8 // Sample size in bits
NumFFT = C.WHISPER_N_FFT
HopLength = C.WHISPER_HOP_LENGTH
ChunkSize = C.WHISPER_CHUNK_SIZE
)
var (
ErrTokenizerFailed = errors.New("whisper_tokenize failed")
ErrAutoDetectFailed = errors.New("whisper_lang_auto_detect failed")
ErrConversionFailed = errors.New("whisper_convert failed")
ErrInvalidLanguage = errors.New("invalid language")
)
///////////////////////////////////////////////////////////////////////////////
// PUBLIC METHODS
// Allocates all memory needed for the model and loads the model from the given file.
// Returns NULL on failure.
func Whisper_init(path string) *Context {
cPath := C.CString(path)
defer C.free(unsafe.Pointer(cPath))
if ctx := C.whisper_init_from_file_with_params(cPath, C.whisper_context_default_params()); ctx != nil {
return (*Context)(ctx)
} else {
return nil
}
}
// Frees all memory allocated by the model.
func (ctx *Context) Whisper_free() {
C.whisper_free((*C.struct_whisper_context)(ctx))
}
// Convert RAW PCM audio to log mel spectrogram.
// The resulting spectrogram is stored inside the provided whisper context.
func (ctx *Context) Whisper_pcm_to_mel(data []float32, threads int) error {
if C.whisper_pcm_to_mel((*C.struct_whisper_context)(ctx), (*C.float)(&data[0]), C.int(len(data)), C.int(threads)) == 0 {
return nil
} else {
return ErrConversionFailed
}
}
// This can be used to set a custom log mel spectrogram inside the provided whisper context.
// Use this instead of whisper_pcm_to_mel() if you want to provide your own log mel spectrogram.
// n_mel must be 80
func (ctx *Context) Whisper_set_mel(data []float32, n_mel int) error {
if C.whisper_set_mel((*C.struct_whisper_context)(ctx), (*C.float)(&data[0]), C.int(len(data)), C.int(n_mel)) == 0 {
return nil
} else {
return ErrConversionFailed
}
}
// Run the Whisper encoder on the log mel spectrogram stored inside the provided whisper context.
// Make sure to call whisper_pcm_to_mel() or whisper_set_mel() first.
// offset can be used to specify the offset of the first frame in the spectrogram.
func (ctx *Context) Whisper_encode(offset, threads int) error {
if C.whisper_encode((*C.struct_whisper_context)(ctx), C.int(offset), C.int(threads)) == 0 {
return nil
} else {
return ErrConversionFailed
}
}
// Run the Whisper decoder to obtain the logits and probabilities for the next token.
// Make sure to call whisper_encode() first.
// tokens + n_tokens is the provided context for the decoder.
// n_past is the number of tokens to use from previous decoder calls.
func (ctx *Context) Whisper_decode(tokens []Token, past, threads int) error {
if C.whisper_decode((*C.struct_whisper_context)(ctx), (*C.whisper_token)(&tokens[0]), C.int(len(tokens)), C.int(past), C.int(threads)) == 0 {
return nil
} else {
return ErrConversionFailed
}
}
// Convert the provided text into tokens. The tokens pointer must be large enough to hold the resulting tokens.
// Returns the number of tokens on success
func (ctx *Context) Whisper_tokenize(text string, tokens []Token) (int, error) {
cText := C.CString(text)
defer C.free(unsafe.Pointer(cText))
if n := C.whisper_tokenize((*C.struct_whisper_context)(ctx), cText, (*C.whisper_token)(&tokens[0]), C.int(len(tokens))); n >= 0 {
return int(n), nil
} else {
return 0, ErrTokenizerFailed
}
}
// Return the id of the specified language, returns -1 if not found
// Examples:
//
// "de" -> 2
// "german" -> 2
func (ctx *Context) Whisper_lang_id(lang string) int {
return int(C.whisper_lang_id(C.CString(lang)))
}
// Largest language id (i.e. number of available languages - 1)
func Whisper_lang_max_id() int {
return int(C.whisper_lang_max_id())
}
// Return the short string of the specified language id (e.g. 2 -> "de"),
// returns empty string if not found
func Whisper_lang_str(id int) string {
return C.GoString(C.whisper_lang_str(C.int(id)))
}
// Use mel data at offset_ms to try and auto-detect the spoken language
// Make sure to call whisper_pcm_to_mel() or whisper_set_mel() first.
// Returns the probabilities of all languages.
// ref: https://github.com/openai/whisper/blob/main/whisper/decoding.py#L18-L69
func (ctx *Context) Whisper_lang_auto_detect(offset_ms, n_threads int) ([]float32, error) {
probs := make([]float32, Whisper_lang_max_id()+1)
if n := int(C.whisper_lang_auto_detect((*C.struct_whisper_context)(ctx), C.int(offset_ms), C.int(n_threads), (*C.float)(&probs[0]))); n < 0 {
return nil, ErrAutoDetectFailed
} else {
return probs, nil
}
}
func (ctx *Context) Whisper_n_len() int {
return int(C.whisper_n_len((*C.struct_whisper_context)(ctx)))
}
func (ctx *Context) Whisper_n_vocab() int {
return int(C.whisper_n_vocab((*C.struct_whisper_context)(ctx)))
}
func (ctx *Context) Whisper_n_text_ctx() int {
return int(C.whisper_n_text_ctx((*C.struct_whisper_context)(ctx)))
}
func (ctx *Context) Whisper_n_audio_ctx() int {
return int(C.whisper_n_audio_ctx((*C.struct_whisper_context)(ctx)))
}
func (ctx *Context) Whisper_is_multilingual() int {
return int(C.whisper_is_multilingual((*C.struct_whisper_context)(ctx)))
}
// The probabilities for the next token
//func (ctx *Whisper_context) Whisper_get_probs() []float32 {
// return (*[1 << 30]float32)(unsafe.Pointer(C.whisper_get_probs((*C.struct_whisper_context)(ctx))))[:ctx.Whisper_n_vocab()]
//}
// Token Id -> String. Uses the vocabulary in the provided context
func (ctx *Context) Whisper_token_to_str(token Token) string {
return C.GoString(C.whisper_token_to_str((*C.struct_whisper_context)(ctx), C.whisper_token(token)))
}
// Special tokens
func (ctx *Context) Whisper_token_eot() Token {
return Token(C.whisper_token_eot((*C.struct_whisper_context)(ctx)))
}
// Special tokens
func (ctx *Context) Whisper_token_sot() Token {
return Token(C.whisper_token_sot((*C.struct_whisper_context)(ctx)))
}
// Special tokens
func (ctx *Context) Whisper_token_prev() Token {
return Token(C.whisper_token_prev((*C.struct_whisper_context)(ctx)))
}
// Special tokens
func (ctx *Context) Whisper_token_solm() Token {
return Token(C.whisper_token_solm((*C.struct_whisper_context)(ctx)))
}
// Special tokens
func (ctx *Context) Whisper_token_not() Token {
return Token(C.whisper_token_not((*C.struct_whisper_context)(ctx)))
}
// Special tokens
func (ctx *Context) Whisper_token_beg() Token {
return Token(C.whisper_token_beg((*C.struct_whisper_context)(ctx)))
}
// Special tokens
func (ctx *Context) Whisper_token_lang(lang_id int) Token {
return Token(C.whisper_token_lang((*C.struct_whisper_context)(ctx), C.int(lang_id)))
}
// Task tokens
func (ctx *Context) Whisper_token_translate() Token {
return Token(C.whisper_token_translate((*C.struct_whisper_context)(ctx)))
}
// Task tokens
func (ctx *Context) Whisper_token_transcribe() Token {
return Token(C.whisper_token_transcribe((*C.struct_whisper_context)(ctx)))
}
// Performance information
func (ctx *Context) Whisper_print_timings() {
C.whisper_print_timings((*C.struct_whisper_context)(ctx))
}
// Performance information
func (ctx *Context) Whisper_reset_timings() {
C.whisper_reset_timings((*C.struct_whisper_context)(ctx))
}
// Print system information
func Whisper_print_system_info() string {
return C.GoString(C.whisper_print_system_info())
}
// Return default parameters for a strategy
func (ctx *Context) Whisper_full_default_params(strategy SamplingStrategy) Params {
// Get default parameters
return Params(C.whisper_full_default_params_cb((*C.struct_whisper_context)(ctx), C.enum_whisper_sampling_strategy(strategy)))
}
// Run the entire model: PCM -> log mel spectrogram -> encoder -> decoder -> text
// Uses the specified decoding strategy to obtain the text.
func (ctx *Context) Whisper_full(
params Params,
samples []float32,
encoderBeginCallback func() bool,
newSegmentCallback func(int),
progressCallback func(int),
) error {
registerEncoderBeginCallback(ctx, encoderBeginCallback)
registerNewSegmentCallback(ctx, newSegmentCallback)
registerProgressCallback(ctx, progressCallback)
defer registerEncoderBeginCallback(ctx, nil)
defer registerNewSegmentCallback(ctx, nil)
defer registerProgressCallback(ctx, nil)
if C.whisper_full((*C.struct_whisper_context)(ctx), (C.struct_whisper_full_params)(params), (*C.float)(&samples[0]), C.int(len(samples))) == 0 {
return nil
} else {
return ErrConversionFailed
}
}
// Split the input audio in chunks and process each chunk separately using whisper_full()
// It seems this approach can offer some speedup in some cases.
// However, the transcription accuracy can be worse at the beginning and end of each chunk.
func (ctx *Context) Whisper_full_parallel(params Params, samples []float32, processors int, encoderBeginCallback func() bool, newSegmentCallback func(int)) error {
registerEncoderBeginCallback(ctx, encoderBeginCallback)
registerNewSegmentCallback(ctx, newSegmentCallback)
defer registerEncoderBeginCallback(ctx, nil)
defer registerNewSegmentCallback(ctx, nil)
if C.whisper_full_parallel((*C.struct_whisper_context)(ctx), (C.struct_whisper_full_params)(params), (*C.float)(&samples[0]), C.int(len(samples)), C.int(processors)) == 0 {
return nil
} else {
return ErrConversionFailed
}
}
// Return the id of the autodetected language, returns -1 if not found
// Added to whisper.cpp in
// https://github.com/ggerganov/whisper.cpp/commit/a1c1583cc7cd8b75222857afc936f0638c5683d6
//
// Examples:
//
// "de" -> 2
// "german" -> 2
func (ctx *Context) Whisper_full_lang_id() int {
return int(C.whisper_full_lang_id((*C.struct_whisper_context)(ctx)))
}
// Number of generated text segments.
// A segment can be a few words, a sentence, or even a paragraph.
func (ctx *Context) Whisper_full_n_segments() int {
return int(C.whisper_full_n_segments((*C.struct_whisper_context)(ctx)))
}
// Get the start and end time of the specified segment.
func (ctx *Context) Whisper_full_get_segment_t0(segment int) int64 {
return int64(C.whisper_full_get_segment_t0((*C.struct_whisper_context)(ctx), C.int(segment)))
}
// Get the start and end time of the specified segment.
func (ctx *Context) Whisper_full_get_segment_t1(segment int) int64 {
return int64(C.whisper_full_get_segment_t1((*C.struct_whisper_context)(ctx), C.int(segment)))
}
// Get the text of the specified segment.
func (ctx *Context) Whisper_full_get_segment_text(segment int) string {
return C.GoString(C.whisper_full_get_segment_text((*C.struct_whisper_context)(ctx), C.int(segment)))
}
// Get number of tokens in the specified segment.
func (ctx *Context) Whisper_full_n_tokens(segment int) int {
return int(C.whisper_full_n_tokens((*C.struct_whisper_context)(ctx), C.int(segment)))
}
// Get the token text of the specified token index in the specified segment.
func (ctx *Context) Whisper_full_get_token_text(segment int, token int) string {
return C.GoString(C.whisper_full_get_token_text((*C.struct_whisper_context)(ctx), C.int(segment), C.int(token)))
}
// Get the token of the specified token index in the specified segment.
func (ctx *Context) Whisper_full_get_token_id(segment int, token int) Token {
return Token(C.whisper_full_get_token_id((*C.struct_whisper_context)(ctx), C.int(segment), C.int(token)))
}
// Get token data for the specified token in the specified segment.
// This contains probabilities, timestamps, etc.
func (ctx *Context) Whisper_full_get_token_data(segment int, token int) TokenData {
return TokenData(C.whisper_full_get_token_data((*C.struct_whisper_context)(ctx), C.int(segment), C.int(token)))
}
// Get the probability of the specified token in the specified segment.
func (ctx *Context) Whisper_full_get_token_p(segment int, token int) float32 {
return float32(C.whisper_full_get_token_p((*C.struct_whisper_context)(ctx), C.int(segment), C.int(token)))
}
///////////////////////////////////////////////////////////////////////////////
// CALLBACKS
var (
cbNewSegment = make(map[unsafe.Pointer]func(int))
cbProgress = make(map[unsafe.Pointer]func(int))
cbEncoderBegin = make(map[unsafe.Pointer]func() bool)
)
func registerNewSegmentCallback(ctx *Context, fn func(int)) {
if fn == nil {
delete(cbNewSegment, unsafe.Pointer(ctx))
} else {
cbNewSegment[unsafe.Pointer(ctx)] = fn
}
}
func registerProgressCallback(ctx *Context, fn func(int)) {
if fn == nil {
delete(cbProgress, unsafe.Pointer(ctx))
} else {
cbProgress[unsafe.Pointer(ctx)] = fn
}
}
func registerEncoderBeginCallback(ctx *Context, fn func() bool) {
if fn == nil {
delete(cbEncoderBegin, unsafe.Pointer(ctx))
} else {
cbEncoderBegin[unsafe.Pointer(ctx)] = fn
}
}
//export callNewSegment
func callNewSegment(user_data unsafe.Pointer, new C.int) {
if fn, ok := cbNewSegment[user_data]; ok {
fn(int(new))
}
}
//export callProgress
func callProgress(user_data unsafe.Pointer, progress C.int) {
if fn, ok := cbProgress[user_data]; ok {
fn(int(progress))
}
}
//export callEncoderBegin
func callEncoderBegin(user_data unsafe.Pointer) C.bool {
if fn, ok := cbEncoderBegin[user_data]; ok {
if fn() {
return C.bool(true)
} else {
return C.bool(false)
}
}
return true
}
func (t TokenData) T0() int64 {
return int64(t.t0)
}
func (t TokenData) T1() int64 {
return int64(t.t1)
}
func (t TokenData) Id() Token {
return Token(t.id)
}

View File

@ -1,113 +0,0 @@
package whisper_test
import (
"os"
"runtime"
"testing"
"time"
// Packages
whisper "github.com/ggerganov/whisper.cpp/bindings/go"
wav "github.com/go-audio/wav"
assert "github.com/stretchr/testify/assert"
)
const (
ModelPath = "models/ggml-small.en.bin"
SamplePath = "samples/jfk.wav"
)
func Test_Whisper_000(t *testing.T) {
assert := assert.New(t)
if _, err := os.Stat(ModelPath); os.IsNotExist(err) {
t.Skip("Skipping test, model not found:", ModelPath)
}
ctx := whisper.Whisper_init(ModelPath)
assert.NotNil(ctx)
ctx.Whisper_free()
}
func Test_Whisper_001(t *testing.T) {
assert := assert.New(t)
if _, err := os.Stat(ModelPath); os.IsNotExist(err) {
t.Skip("Skipping test, model not found:", ModelPath)
}
if _, err := os.Stat(SamplePath); os.IsNotExist(err) {
t.Skip("Skipping test, sample not found:", SamplePath)
}
// Open samples
fh, err := os.Open(SamplePath)
assert.NoError(err)
defer fh.Close()
// Read samples
d := wav.NewDecoder(fh)
buf, err := d.FullPCMBuffer()
assert.NoError(err)
// Run whisper
ctx := whisper.Whisper_init(ModelPath)
assert.NotNil(ctx)
defer ctx.Whisper_free()
params := ctx.Whisper_full_default_params(whisper.SAMPLING_GREEDY)
data := buf.AsFloat32Buffer().Data
err = ctx.Whisper_full(params, data, nil, nil, nil)
assert.NoError(err)
// Print out tokens
num_segments := ctx.Whisper_full_n_segments()
assert.GreaterOrEqual(num_segments, 1)
for i := 0; i < num_segments; i++ {
str := ctx.Whisper_full_get_segment_text(i)
assert.NotEmpty(str)
t0 := time.Duration(ctx.Whisper_full_get_segment_t0(i)) * time.Millisecond
t1 := time.Duration(ctx.Whisper_full_get_segment_t1(i)) * time.Millisecond
t.Logf("[%6s->%-6s] %q", t0, t1, str)
}
}
func Test_Whisper_002(t *testing.T) {
assert := assert.New(t)
for i := 0; i < whisper.Whisper_lang_max_id(); i++ {
str := whisper.Whisper_lang_str(i)
assert.NotEmpty(str)
t.Log(str)
}
}
func Test_Whisper_003(t *testing.T) {
threads := runtime.NumCPU()
assert := assert.New(t)
if _, err := os.Stat(ModelPath); os.IsNotExist(err) {
t.Skip("Skipping test, model not found:", ModelPath)
}
if _, err := os.Stat(SamplePath); os.IsNotExist(err) {
t.Skip("Skipping test, sample not found:", SamplePath)
}
// Open samples
fh, err := os.Open(SamplePath)
assert.NoError(err)
defer fh.Close()
// Read samples
d := wav.NewDecoder(fh)
buf, err := d.FullPCMBuffer()
assert.NoError(err)
// Make the model
ctx := whisper.Whisper_init(ModelPath)
assert.NotNil(ctx)
defer ctx.Whisper_free()
// Get MEL
assert.NoError(ctx.Whisper_pcm_to_mel(buf.AsFloat32Buffer().Data, threads))
// Get Languages
languages, err := ctx.Whisper_lang_auto_detect(0, threads)
assert.NoError(err)
for i, p := range languages {
t.Logf("%s: %f", whisper.Whisper_lang_str(i), p)
}
}

View File

@ -1,124 +0,0 @@
<?xml version="1.0" encoding="UTF-8"?>
<project version="4">
<component name="Palette2">
<group name="Swing">
<item class="com.intellij.uiDesigner.HSpacer" tooltip-text="Horizontal Spacer" icon="/com/intellij/uiDesigner/icons/hspacer.svg" removable="false" auto-create-binding="false" can-attach-label="false">
<default-constraints vsize-policy="1" hsize-policy="6" anchor="0" fill="1" />
</item>
<item class="com.intellij.uiDesigner.VSpacer" tooltip-text="Vertical Spacer" icon="/com/intellij/uiDesigner/icons/vspacer.svg" removable="false" auto-create-binding="false" can-attach-label="false">
<default-constraints vsize-policy="6" hsize-policy="1" anchor="0" fill="2" />
</item>
<item class="javax.swing.JPanel" icon="/com/intellij/uiDesigner/icons/panel.svg" removable="false" auto-create-binding="false" can-attach-label="false">
<default-constraints vsize-policy="3" hsize-policy="3" anchor="0" fill="3" />
</item>
<item class="javax.swing.JScrollPane" icon="/com/intellij/uiDesigner/icons/scrollPane.svg" removable="false" auto-create-binding="false" can-attach-label="true">
<default-constraints vsize-policy="7" hsize-policy="7" anchor="0" fill="3" />
</item>
<item class="javax.swing.JButton" icon="/com/intellij/uiDesigner/icons/button.svg" removable="false" auto-create-binding="true" can-attach-label="false">
<default-constraints vsize-policy="0" hsize-policy="3" anchor="0" fill="1" />
<initial-values>
<property name="text" value="Button" />
</initial-values>
</item>
<item class="javax.swing.JRadioButton" icon="/com/intellij/uiDesigner/icons/radioButton.svg" removable="false" auto-create-binding="true" can-attach-label="false">
<default-constraints vsize-policy="0" hsize-policy="3" anchor="8" fill="0" />
<initial-values>
<property name="text" value="RadioButton" />
</initial-values>
</item>
<item class="javax.swing.JCheckBox" icon="/com/intellij/uiDesigner/icons/checkBox.svg" removable="false" auto-create-binding="true" can-attach-label="false">
<default-constraints vsize-policy="0" hsize-policy="3" anchor="8" fill="0" />
<initial-values>
<property name="text" value="CheckBox" />
</initial-values>
</item>
<item class="javax.swing.JLabel" icon="/com/intellij/uiDesigner/icons/label.svg" removable="false" auto-create-binding="false" can-attach-label="false">
<default-constraints vsize-policy="0" hsize-policy="0" anchor="8" fill="0" />
<initial-values>
<property name="text" value="Label" />
</initial-values>
</item>
<item class="javax.swing.JTextField" icon="/com/intellij/uiDesigner/icons/textField.svg" removable="false" auto-create-binding="true" can-attach-label="true">
<default-constraints vsize-policy="0" hsize-policy="6" anchor="8" fill="1">
<preferred-size width="150" height="-1" />
</default-constraints>
</item>
<item class="javax.swing.JPasswordField" icon="/com/intellij/uiDesigner/icons/passwordField.svg" removable="false" auto-create-binding="true" can-attach-label="true">
<default-constraints vsize-policy="0" hsize-policy="6" anchor="8" fill="1">
<preferred-size width="150" height="-1" />
</default-constraints>
</item>
<item class="javax.swing.JFormattedTextField" icon="/com/intellij/uiDesigner/icons/formattedTextField.svg" removable="false" auto-create-binding="true" can-attach-label="true">
<default-constraints vsize-policy="0" hsize-policy="6" anchor="8" fill="1">
<preferred-size width="150" height="-1" />
</default-constraints>
</item>
<item class="javax.swing.JTextArea" icon="/com/intellij/uiDesigner/icons/textArea.svg" removable="false" auto-create-binding="true" can-attach-label="true">
<default-constraints vsize-policy="6" hsize-policy="6" anchor="0" fill="3">
<preferred-size width="150" height="50" />
</default-constraints>
</item>
<item class="javax.swing.JTextPane" icon="/com/intellij/uiDesigner/icons/textPane.svg" removable="false" auto-create-binding="true" can-attach-label="true">
<default-constraints vsize-policy="6" hsize-policy="6" anchor="0" fill="3">
<preferred-size width="150" height="50" />
</default-constraints>
</item>
<item class="javax.swing.JEditorPane" icon="/com/intellij/uiDesigner/icons/editorPane.svg" removable="false" auto-create-binding="true" can-attach-label="true">
<default-constraints vsize-policy="6" hsize-policy="6" anchor="0" fill="3">
<preferred-size width="150" height="50" />
</default-constraints>
</item>
<item class="javax.swing.JComboBox" icon="/com/intellij/uiDesigner/icons/comboBox.svg" removable="false" auto-create-binding="true" can-attach-label="true">
<default-constraints vsize-policy="0" hsize-policy="2" anchor="8" fill="1" />
</item>
<item class="javax.swing.JTable" icon="/com/intellij/uiDesigner/icons/table.svg" removable="false" auto-create-binding="true" can-attach-label="false">
<default-constraints vsize-policy="6" hsize-policy="6" anchor="0" fill="3">
<preferred-size width="150" height="50" />
</default-constraints>
</item>
<item class="javax.swing.JList" icon="/com/intellij/uiDesigner/icons/list.svg" removable="false" auto-create-binding="true" can-attach-label="false">
<default-constraints vsize-policy="6" hsize-policy="2" anchor="0" fill="3">
<preferred-size width="150" height="50" />
</default-constraints>
</item>
<item class="javax.swing.JTree" icon="/com/intellij/uiDesigner/icons/tree.svg" removable="false" auto-create-binding="true" can-attach-label="false">
<default-constraints vsize-policy="6" hsize-policy="6" anchor="0" fill="3">
<preferred-size width="150" height="50" />
</default-constraints>
</item>
<item class="javax.swing.JTabbedPane" icon="/com/intellij/uiDesigner/icons/tabbedPane.svg" removable="false" auto-create-binding="true" can-attach-label="false">
<default-constraints vsize-policy="3" hsize-policy="3" anchor="0" fill="3">
<preferred-size width="200" height="200" />
</default-constraints>
</item>
<item class="javax.swing.JSplitPane" icon="/com/intellij/uiDesigner/icons/splitPane.svg" removable="false" auto-create-binding="false" can-attach-label="false">
<default-constraints vsize-policy="3" hsize-policy="3" anchor="0" fill="3">
<preferred-size width="200" height="200" />
</default-constraints>
</item>
<item class="javax.swing.JSpinner" icon="/com/intellij/uiDesigner/icons/spinner.svg" removable="false" auto-create-binding="true" can-attach-label="true">
<default-constraints vsize-policy="0" hsize-policy="6" anchor="8" fill="1" />
</item>
<item class="javax.swing.JSlider" icon="/com/intellij/uiDesigner/icons/slider.svg" removable="false" auto-create-binding="true" can-attach-label="false">
<default-constraints vsize-policy="0" hsize-policy="6" anchor="8" fill="1" />
</item>
<item class="javax.swing.JSeparator" icon="/com/intellij/uiDesigner/icons/separator.svg" removable="false" auto-create-binding="false" can-attach-label="false">
<default-constraints vsize-policy="6" hsize-policy="6" anchor="0" fill="3" />
</item>
<item class="javax.swing.JProgressBar" icon="/com/intellij/uiDesigner/icons/progressbar.svg" removable="false" auto-create-binding="true" can-attach-label="false">
<default-constraints vsize-policy="0" hsize-policy="6" anchor="0" fill="1" />
</item>
<item class="javax.swing.JToolBar" icon="/com/intellij/uiDesigner/icons/toolbar.svg" removable="false" auto-create-binding="false" can-attach-label="false">
<default-constraints vsize-policy="0" hsize-policy="6" anchor="0" fill="1">
<preferred-size width="-1" height="20" />
</default-constraints>
</item>
<item class="javax.swing.JToolBar$Separator" icon="/com/intellij/uiDesigner/icons/toolbarSeparator.svg" removable="false" auto-create-binding="false" can-attach-label="false">
<default-constraints vsize-policy="0" hsize-policy="0" anchor="0" fill="1" />
</item>
<item class="javax.swing.JScrollBar" icon="/com/intellij/uiDesigner/icons/scrollbar.svg" removable="false" auto-create-binding="true" can-attach-label="false">
<default-constraints vsize-policy="6" hsize-policy="0" anchor="0" fill="2" />
</item>
</group>
</component>
</project>

View File

@ -1,71 +0,0 @@
# Java JNI bindings for Whisper
This package provides Java JNI bindings for whisper.cpp. They have been tested on:
* <strike>Darwin (OS X) 12.6 on x64_64</strike>
* Ubuntu on x86_64
* Windows on x86_64
The "low level" bindings are in `WhisperCppJnaLibrary`. The most simple usage is as follows:
JNA will attempt to load the `whispercpp` shared library from:
- jna.library.path
- jna.platform.library
- ~/Library/Frameworks
- /Library/Frameworks
- /System/Library/Frameworks
- classpath
```java
import io.github.ggerganov.whispercpp.WhisperCpp;
public class Example {
public static void main(String[] args) {
WhisperCpp whisper = new WhisperCpp();
// By default, models are loaded from ~/.cache/whisper/ and are usually named "ggml-${name}.bin"
// or you can provide the absolute path to the model file.
long context = whisper.initContext("base.en");
try {
var whisperParams = whisper.getFullDefaultParams(WhisperSamplingStrategy.WHISPER_SAMPLING_GREEDY);
// custom configuration if required
whisperParams.temperature_inc = 0f;
var samples = readAudio(); // divide each value by 32767.0f
whisper.fullTranscribe(whisperParams, samples);
int segmentCount = whisper.getTextSegmentCount(context);
for (int i = 0; i < segmentCount; i++) {
String text = whisper.getTextSegment(context, i);
System.out.println(segment.getText());
}
} finally {
whisper.freeContext(context);
}
}
}
```
## Building & Testing
In order to build, you need to have the JDK 8 or higher installed. Run the tests with:
```bash
git clone https://github.com/ggml-org/whisper.cpp.git
cd whisper.cpp/bindings/java
./gradlew build
```
You need to have the `whisper` library in your [JNA library path](https://java-native-access.github.io/jna/4.2.1/com/sun/jna/NativeLibrary.html). On Windows the dll is included in the jar and you can update it:
```bash
copy /y ..\..\build\bin\Release\whisper.dll build\generated\resources\main\win32-x86-64\whisper.dll
```
## License
The license for the Java bindings is the same as the license for the rest of the whisper.cpp project, which is the MIT License. See the `LICENSE` file for more details.

View File

@ -1,159 +0,0 @@
plugins {
id 'java'
id 'java-library'
id 'maven-publish'
id 'signing'
}
archivesBaseName = 'whispercpp'
group = 'io.github.ggerganov'
version = '1.4.0'
sourceCompatibility = 1.8
targetCompatibility = 1.8
sourceSets {
main {
resources {
srcDirs = ['src/main/resources', 'build/generated/resources/main']
}
}
test {
runtimeClasspath += files('build/generated/resources/main')
}
}
tasks.register('copyLibwhisperDynlib', Copy) {
from '../../build/src'
include 'libwhisper.dylib'
into 'build/generated/resources/main'
}
tasks.register('copyLibwhisperSo', Copy) {
from '../../build/src'
include 'libwhisper.so'
into 'build/generated/resources/main'
}
tasks.register('copyWhisperDLL', Copy) {
from '../../build/bin/Release'
include 'whisper.dll'
into 'build/generated/resources/main'
}
tasks.register('copyGGML_BASE_DLL', Copy) {
from '../../build/bin/Release'
include 'ggml-base.dll'
into 'build/generated/resources/main'
}
tasks.register('copyGGML_DLL', Copy) {
from '../../build/bin/Release'
include 'ggml.dll'
into 'build/generated/resources/main'
}
tasks.register('copyGGML_CPU_DLL', Copy) {
from '../../build/bin/Release'
include 'ggml-cpu.dll'
into 'build/generated/resources/main'
}
tasks.register('copyLibs') {
dependsOn copyLibwhisperDynlib, copyLibwhisperSo, copyWhisperDLL, copyGGML_BASE_DLL, copyGGML_DLL, copyGGML_CPU_DLL
}
test {
systemProperty 'jna.library.path', project.file('build/generated/resources/main').absolutePath
}
java {
withSourcesJar()
withJavadocJar()
}
sourcesJar() {
dependsOn copyLibs
}
jar {
dependsOn copyLibs
exclude '**/whisper_java.exp', '**/whisper_java.lib'
}
javadoc {
options.addStringOption('Xdoclint:none', '-quiet')
}
tasks.withType(Test) {
useJUnitPlatform()
}
test.dependsOn copyLibs
processResources.dependsOn copyLibs
dependencies {
implementation "net.java.dev.jna:jna:5.13.0"
testImplementation "org.junit.jupiter:junit-jupiter:5.9.2"
testImplementation "org.assertj:assertj-core:3.24.2"
}
repositories {
mavenCentral()
}
publishing {
publications {
mavenJava(MavenPublication) {
artifactId = 'whispercpp'
from components.java
pom {
name = 'whispercpp'
description = "Java JNA bindings for OpenAI's Whisper model, implemented in C/C++"
url = 'https://github.com/ggerganov/whisper.cpp'
licenses {
license {
name = 'MIT licence'
url = 'https://raw.githubusercontent.com/ggerganov/whisper.cpp/master/LICENSE'
}
}
developers {
developer {
id = 'ggerganov'
name = 'Georgi Gerganov'
email = 'ggerganov@gmail.com'
}
developer {
id = 'nalbion'
name = 'Nicholas Albion'
email = 'nalbion@yahoo.com'
}
}
scm {
connection = 'scm:git:git://github.com/ggerganov/whisper.cpp.git'
url = 'https://github.com/ggerganov/whisper.cpp'
}
}
}
}
repositories {
maven {
def releasesRepoUrl = 'https://s01.oss.sonatype.org/service/local/staging/deploy/maven2/'
def snapshotsRepoUrl = 'https://s01.oss.sonatype.org/content/repositories/snapshots/'
url = version.endsWith('-SNAPSHOT') ? snapshotsRepoUrl : releasesRepoUrl
credentials {
username = System.getenv("MAVEN_USERNAME")
password = System.getenv("MAVEN_PASSWORD")
}
}
}
}
signing {
def signingKey = System.getenv("PGP_SECRET")
def signingPassword = System.getenv("PGP_PASSPHRASE")
useInMemoryPgpKeys(signingKey, signingPassword)
sign publishing.publications.mavenJava
}

View File

@ -1,6 +0,0 @@
org.gradle.jvmargs=-Xms256m -Xmx1024m
system.include.dir=/usr/include
#system.local.include.dir=../../include
system.local.include.dir=./build/generated/sources/headers/java/main
jni.include.dir=/usr/lib/jvm/java-8-openjdk-amd64/include/
jni.lib.dir=/usr/lib/jvm/java-8-openjdk-amd64/lib/

Binary file not shown.

View File

@ -1,6 +0,0 @@
distributionBase=GRADLE_USER_HOME
distributionPath=wrapper/dists
distributionUrl=https\://services.gradle.org/distributions/gradle-8.1-bin.zip
networkTimeout=10000
zipStoreBase=GRADLE_USER_HOME
zipStorePath=wrapper/dists

244
bindings/java/gradlew vendored
View File

@ -1,244 +0,0 @@
#!/bin/sh
#
# Copyright © 2015-2021 the original authors.
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# https://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
#
##############################################################################
#
# Gradle start up script for POSIX generated by Gradle.
#
# Important for running:
#
# (1) You need a POSIX-compliant shell to run this script. If your /bin/sh is
# noncompliant, but you have some other compliant shell such as ksh or
# bash, then to run this script, type that shell name before the whole
# command line, like:
#
# ksh Gradle
#
# Busybox and similar reduced shells will NOT work, because this script
# requires all of these POSIX shell features:
# * functions;
# * expansions «$var», «${var}», «${var:-default}», «${var+SET}»,
# «${var#prefix}», «${var%suffix}», and «$( cmd )»;
# * compound commands having a testable exit status, especially «case»;
# * various built-in commands including «command», «set», and «ulimit».
#
# Important for patching:
#
# (2) This script targets any POSIX shell, so it avoids extensions provided
# by Bash, Ksh, etc; in particular arrays are avoided.
#
# The "traditional" practice of packing multiple parameters into a
# space-separated string is a well documented source of bugs and security
# problems, so this is (mostly) avoided, by progressively accumulating
# options in "$@", and eventually passing that to Java.
#
# Where the inherited environment variables (DEFAULT_JVM_OPTS, JAVA_OPTS,
# and GRADLE_OPTS) rely on word-splitting, this is performed explicitly;
# see the in-line comments for details.
#
# There are tweaks for specific operating systems such as AIX, CygWin,
# Darwin, MinGW, and NonStop.
#
# (3) This script is generated from the Groovy template
# https://github.com/gradle/gradle/blob/HEAD/subprojects/plugins/src/main/resources/org/gradle/api/internal/plugins/unixStartScript.txt
# within the Gradle project.
#
# You can find Gradle at https://github.com/gradle/gradle/.
#
##############################################################################
# Attempt to set APP_HOME
# Resolve links: $0 may be a link
app_path=$0
# Need this for daisy-chained symlinks.
while
APP_HOME=${app_path%"${app_path##*/}"} # leaves a trailing /; empty if no leading path
[ -h "$app_path" ]
do
ls=$( ls -ld "$app_path" )
link=${ls#*' -> '}
case $link in #(
/*) app_path=$link ;; #(
*) app_path=$APP_HOME$link ;;
esac
done
# This is normally unused
# shellcheck disable=SC2034
APP_BASE_NAME=${0##*/}
APP_HOME=$( cd "${APP_HOME:-./}" && pwd -P ) || exit
# Add default JVM options here. You can also use JAVA_OPTS and GRADLE_OPTS to pass JVM options to this script.
DEFAULT_JVM_OPTS='"-Xmx64m" "-Xms64m"'
# Use the maximum available, or set MAX_FD != -1 to use that value.
MAX_FD=maximum
warn () {
echo "$*"
} >&2
die () {
echo
echo "$*"
echo
exit 1
} >&2
# OS specific support (must be 'true' or 'false').
cygwin=false
msys=false
darwin=false
nonstop=false
case "$( uname )" in #(
CYGWIN* ) cygwin=true ;; #(
Darwin* ) darwin=true ;; #(
MSYS* | MINGW* ) msys=true ;; #(
NONSTOP* ) nonstop=true ;;
esac
CLASSPATH=$APP_HOME/gradle/wrapper/gradle-wrapper.jar
# Determine the Java command to use to start the JVM.
if [ -n "$JAVA_HOME" ] ; then
if [ -x "$JAVA_HOME/jre/sh/java" ] ; then
# IBM's JDK on AIX uses strange locations for the executables
JAVACMD=$JAVA_HOME/jre/sh/java
else
JAVACMD=$JAVA_HOME/bin/java
fi
if [ ! -x "$JAVACMD" ] ; then
die "ERROR: JAVA_HOME is set to an invalid directory: $JAVA_HOME
Please set the JAVA_HOME variable in your environment to match the
location of your Java installation."
fi
else
JAVACMD=java
which java >/dev/null 2>&1 || die "ERROR: JAVA_HOME is not set and no 'java' command could be found in your PATH.
Please set the JAVA_HOME variable in your environment to match the
location of your Java installation."
fi
# Increase the maximum file descriptors if we can.
if ! "$cygwin" && ! "$darwin" && ! "$nonstop" ; then
case $MAX_FD in #(
max*)
# In POSIX sh, ulimit -H is undefined. That's why the result is checked to see if it worked.
# shellcheck disable=SC3045
MAX_FD=$( ulimit -H -n ) ||
warn "Could not query maximum file descriptor limit"
esac
case $MAX_FD in #(
'' | soft) :;; #(
*)
# In POSIX sh, ulimit -n is undefined. That's why the result is checked to see if it worked.
# shellcheck disable=SC3045
ulimit -n "$MAX_FD" ||
warn "Could not set maximum file descriptor limit to $MAX_FD"
esac
fi
# Collect all arguments for the java command, stacking in reverse order:
# * args from the command line
# * the main class name
# * -classpath
# * -D...appname settings
# * --module-path (only if needed)
# * DEFAULT_JVM_OPTS, JAVA_OPTS, and GRADLE_OPTS environment variables.
# For Cygwin or MSYS, switch paths to Windows format before running java
if "$cygwin" || "$msys" ; then
APP_HOME=$( cygpath --path --mixed "$APP_HOME" )
CLASSPATH=$( cygpath --path --mixed "$CLASSPATH" )
JAVACMD=$( cygpath --unix "$JAVACMD" )
# Now convert the arguments - kludge to limit ourselves to /bin/sh
for arg do
if
case $arg in #(
-*) false ;; # don't mess with options #(
/?*) t=${arg#/} t=/${t%%/*} # looks like a POSIX filepath
[ -e "$t" ] ;; #(
*) false ;;
esac
then
arg=$( cygpath --path --ignore --mixed "$arg" )
fi
# Roll the args list around exactly as many times as the number of
# args, so each arg winds up back in the position where it started, but
# possibly modified.
#
# NB: a `for` loop captures its iteration list before it begins, so
# changing the positional parameters here affects neither the number of
# iterations, nor the values presented in `arg`.
shift # remove old arg
set -- "$@" "$arg" # push replacement arg
done
fi
# Collect all arguments for the java command;
# * $DEFAULT_JVM_OPTS, $JAVA_OPTS, and $GRADLE_OPTS can contain fragments of
# shell script including quotes and variable substitutions, so put them in
# double quotes to make sure that they get re-expanded; and
# * put everything else in single quotes, so that it's not re-expanded.
set -- \
"-Dorg.gradle.appname=$APP_BASE_NAME" \
-classpath "$CLASSPATH" \
org.gradle.wrapper.GradleWrapperMain \
"$@"
# Stop when "xargs" is not available.
if ! command -v xargs >/dev/null 2>&1
then
die "xargs is not available"
fi
# Use "xargs" to parse quoted args.
#
# With -n1 it outputs one arg per line, with the quotes and backslashes removed.
#
# In Bash we could simply go:
#
# readarray ARGS < <( xargs -n1 <<<"$var" ) &&
# set -- "${ARGS[@]}" "$@"
#
# but POSIX shell has neither arrays nor command substitution, so instead we
# post-process each arg (as a line of input to sed) to backslash-escape any
# character that might be a shell metacharacter, then use eval to reverse
# that process (while maintaining the separation between arguments), and wrap
# the whole thing up as a single "set" statement.
#
# This will of course break if any of these variables contains a newline or
# an unmatched quote.
#
eval "set -- $(
printf '%s\n' "$DEFAULT_JVM_OPTS $JAVA_OPTS $GRADLE_OPTS" |
xargs -n1 |
sed ' s~[^-[:alnum:]+,./:=@_]~\\&~g; ' |
tr '\n' ' '
)" '"$@"'
exec "$JAVACMD" "$@"

View File

@ -1,92 +0,0 @@
@rem
@rem Copyright 2015 the original author or authors.
@rem
@rem Licensed under the Apache License, Version 2.0 (the "License");
@rem you may not use this file except in compliance with the License.
@rem You may obtain a copy of the License at
@rem
@rem https://www.apache.org/licenses/LICENSE-2.0
@rem
@rem Unless required by applicable law or agreed to in writing, software
@rem distributed under the License is distributed on an "AS IS" BASIS,
@rem WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
@rem See the License for the specific language governing permissions and
@rem limitations under the License.
@rem
@if "%DEBUG%"=="" @echo off
@rem ##########################################################################
@rem
@rem Gradle startup script for Windows
@rem
@rem ##########################################################################
@rem Set local scope for the variables with windows NT shell
if "%OS%"=="Windows_NT" setlocal
set DIRNAME=%~dp0
if "%DIRNAME%"=="" set DIRNAME=.
@rem This is normally unused
set APP_BASE_NAME=%~n0
set APP_HOME=%DIRNAME%
@rem Resolve any "." and ".." in APP_HOME to make it shorter.
for %%i in ("%APP_HOME%") do set APP_HOME=%%~fi
@rem Add default JVM options here. You can also use JAVA_OPTS and GRADLE_OPTS to pass JVM options to this script.
set DEFAULT_JVM_OPTS="-Xmx64m" "-Xms64m"
@rem Find java.exe
if defined JAVA_HOME goto findJavaFromJavaHome
set JAVA_EXE=java.exe
%JAVA_EXE% -version >NUL 2>&1
if %ERRORLEVEL% equ 0 goto execute
echo.
echo ERROR: JAVA_HOME is not set and no 'java' command could be found in your PATH.
echo.
echo Please set the JAVA_HOME variable in your environment to match the
echo location of your Java installation.
goto fail
:findJavaFromJavaHome
set JAVA_HOME=%JAVA_HOME:"=%
set JAVA_EXE=%JAVA_HOME%/bin/java.exe
if exist "%JAVA_EXE%" goto execute
echo.
echo ERROR: JAVA_HOME is set to an invalid directory: %JAVA_HOME%
echo.
echo Please set the JAVA_HOME variable in your environment to match the
echo location of your Java installation.
goto fail
:execute
@rem Setup the command line
set CLASSPATH=%APP_HOME%\gradle\wrapper\gradle-wrapper.jar
@rem Execute Gradle
"%JAVA_EXE%" %DEFAULT_JVM_OPTS% %JAVA_OPTS% %GRADLE_OPTS% "-Dorg.gradle.appname=%APP_BASE_NAME%" -classpath "%CLASSPATH%" org.gradle.wrapper.GradleWrapperMain %*
:end
@rem End local scope for the variables with windows NT shell
if %ERRORLEVEL% equ 0 goto mainEnd
:fail
rem Set variable GRADLE_EXIT_CONSOLE if you need the _script_ return code instead of
rem the _cmd.exe /c_ return code!
set EXIT_CODE=%ERRORLEVEL%
if %EXIT_CODE% equ 0 set EXIT_CODE=1
if not ""=="%GRADLE_EXIT_CONSOLE%" exit %EXIT_CODE%
exit /b %EXIT_CODE%
:mainEnd
if "%OS%"=="Windows_NT" endlocal
:omega

View File

@ -1 +0,0 @@
rootProject.name = "whispercpp"

View File

@ -1,24 +0,0 @@
package io.github.ggerganov.whispercpp;
/**
* Presets for alignment heads in DTW token timestamps
*/
public class WhisperConstants {
// Alignment heads presets
public static final int WHISPER_AHEADS_NONE = 0;
public static final int WHISPER_AHEADS_TINY_EN = 1;
public static final int WHISPER_AHEADS_TINY = 2;
public static final int WHISPER_AHEADS_BASE_EN = 3;
public static final int WHISPER_AHEADS_BASE = 4;
public static final int WHISPER_AHEADS_SMALL_EN = 5;
public static final int WHISPER_AHEADS_SMALL = 6;
public static final int WHISPER_AHEADS_MEDIUM_EN = 7;
public static final int WHISPER_AHEADS_MEDIUM = 8;
public static final int WHISPER_AHEADS_LARGE_V1 = 9;
public static final int WHISPER_AHEADS_LARGE_V2 = 10;
public static final int WHISPER_AHEADS_LARGE_V3 = 11;
public static final int WHISPER_AHEADS_LARGE_V3_TURBO = 12;
public static final int WHISPER_AHEADS_CUSTOM = 13;
public static final int WHISPER_AHEADS_N_TOP_MOST = 14;
public static final int WHISPER_AHEADS_COUNT = 15;
}

View File

@ -1,36 +0,0 @@
package io.github.ggerganov.whispercpp;
import com.sun.jna.NativeLong;
import com.sun.jna.Structure;
import com.sun.jna.ptr.PointerByReference;
import com.sun.jna.Pointer;
import io.github.ggerganov.whispercpp.ggml.GgmlType;
import io.github.ggerganov.whispercpp.WhisperModel;
import io.github.ggerganov.whispercpp.params.WhisperContextParams;
import java.util.List;
public class WhisperContext extends Structure {
public NativeLong t_load_us;
public NativeLong t_start_us;
/** weight type (FP32 / FP16 / QX) */
public GgmlType wtype = GgmlType.GGML_TYPE_F16;
/** intermediate type (FP32 or FP16) */
public GgmlType itype = GgmlType.GGML_TYPE_F16;
public WhisperContextParams.ByValue params;
public Pointer model;
public Pointer vocab;
public Pointer state;
/** populated by whisper_init_from_file_with_params() */
public Pointer path_model;
@Override
protected List<String> getFieldOrder() {
return List.of("t_load_us", "t_start_us", "wtype", "itype",
"params", "model", "vocab", "state", "path_model");
}
}

View File

@ -1,219 +0,0 @@
package io.github.ggerganov.whispercpp;
import com.sun.jna.Native;
import com.sun.jna.Pointer;
import io.github.ggerganov.whispercpp.bean.WhisperSegment;
import io.github.ggerganov.whispercpp.params.WhisperContextParams;
import io.github.ggerganov.whispercpp.params.WhisperFullParams;
import io.github.ggerganov.whispercpp.params.WhisperSamplingStrategy;
import java.io.File;
import java.io.FileNotFoundException;
import java.io.IOException;
import java.util.ArrayList;
import java.util.List;
/**
* Before calling most methods, you must call `initContext(modelPath)` to initialise the `ctx` Pointer.
*/
public class WhisperCpp implements AutoCloseable {
private WhisperCppJnaLibrary lib = WhisperCppJnaLibrary.instance;
private Pointer ctx = null;
private Pointer paramsPointer = null;
private Pointer greedyParamsPointer = null;
private Pointer beamParamsPointer = null;
public File modelDir() {
String modelDirPath = System.getenv("XDG_CACHE_HOME");
if (modelDirPath == null) {
modelDirPath = System.getProperty("user.home") + "/.cache";
}
return new File(modelDirPath, "whisper");
}
/**
* @param modelPath - absolute path, or just the name (eg: "base", "base-en" or "base.en")
*/
public void initContext(String modelPath) throws FileNotFoundException {
initContextImpl(modelPath, getContextDefaultParams());
}
/**
* @param modelPath - absolute path, or just the name (eg: "base", "base-en" or "base.en")
* @param params - params to use when initialising the context
*/
public void initContext(String modelPath, WhisperContextParams.ByValue params) throws FileNotFoundException {
initContextImpl(modelPath, params);
}
private void initContextImpl(String modelPath, WhisperContextParams.ByValue params) throws FileNotFoundException {
if (ctx != null) {
lib.whisper_free(ctx);
}
if (!modelPath.contains("/") && !modelPath.contains("\\")) {
if (!modelPath.endsWith(".bin")) {
modelPath = "ggml-" + modelPath.replace("-", ".") + ".bin";
}
modelPath = new File(modelDir(), modelPath).getAbsolutePath();
}
ctx = lib.whisper_init_from_file_with_params(modelPath, params);
if (ctx == null) {
throw new FileNotFoundException(modelPath);
}
}
/**
* Provides default params which can be used with `whisper_init_from_file_with_params()` etc.
* Returns a ByValue instance to ensure proper parameter passing to native code.
*/
public WhisperContextParams.ByValue getContextDefaultParams() {
WhisperContextParams.ByValue valueParams = new WhisperContextParams.ByValue(
lib.whisper_context_default_params_by_ref());
valueParams.read();
return valueParams;
}
/**
* Provides default params which can be used with `whisper_full()` etc.
* Because this function allocates memory for the params, the caller must call either:
* - call `whisper_free_params()`
* - `Native.free(Pointer.nativeValue(pointer));`
*
* @param strategy - GREEDY
*/
public WhisperFullParams.ByValue getFullDefaultParams(WhisperSamplingStrategy strategy) {
Pointer pointer;
// whisper_full_default_params_by_ref allocates memory which we need to delete, so only create max 1 pointer for each strategy.
if (strategy == WhisperSamplingStrategy.WHISPER_SAMPLING_GREEDY) {
if (greedyParamsPointer == null) {
greedyParamsPointer = lib.whisper_full_default_params_by_ref(strategy.ordinal());
}
pointer = greedyParamsPointer;
} else {
if (beamParamsPointer == null) {
beamParamsPointer = lib.whisper_full_default_params_by_ref(strategy.ordinal());
}
pointer = beamParamsPointer;
}
WhisperFullParams.ByValue params = new WhisperFullParams.ByValue(pointer);
params.read();
return params;
}
@Override
public void close() {
freeContext();
freeParams();
System.out.println("Whisper closed");
}
private void freeContext() {
if (ctx != null) {
lib.whisper_free(ctx);
}
}
private void freeParams() {
if (paramsPointer != null) {
Native.free(Pointer.nativeValue(paramsPointer));
paramsPointer = null;
}
if (greedyParamsPointer != null) {
Native.free(Pointer.nativeValue(greedyParamsPointer));
greedyParamsPointer = null;
}
if (beamParamsPointer != null) {
Native.free(Pointer.nativeValue(beamParamsPointer));
beamParamsPointer = null;
}
}
/**
* Run the entire model: PCM -&gt; log mel spectrogram -&gt; encoder -&gt; decoder -&gt; text.
* Not thread safe for same context
* Uses the specified decoding strategy to obtain the text.
*/
public String fullTranscribe(WhisperFullParams.ByValue whisperParams, float[] audioData) throws IOException {
if (ctx == null) {
throw new IllegalStateException("Model not initialised");
}
/*
WhisperFullParams.ByValue valueParams = new WhisperFullParams.ByValue(
lib.whisper_full_default_params_by_ref(WhisperSamplingStrategy.WHISPER_SAMPLING_BEAM_SEARCH.ordinal()));
valueParams.read();
*/
if (lib.whisper_full(ctx, whisperParams, audioData, audioData.length) != 0) {
throw new IOException("Failed to process audio");
}
int nSegments = lib.whisper_full_n_segments(ctx);
StringBuilder str = new StringBuilder();
for (int i = 0; i < nSegments; i++) {
String text = lib.whisper_full_get_segment_text(ctx, i);
System.out.println("Segment:" + text);
str.append(text);
}
return str.toString().trim();
}
/**
* Full transcribe with time list.
*
* @param whisperParams the whisper params
* @param audioData the audio data
* @return the list
* @throws IOException the io exception
*/
public List<WhisperSegment> fullTranscribeWithTime(WhisperFullParams.ByValue whisperParams, float[] audioData) throws IOException {
if (ctx == null) {
throw new IllegalStateException("Model not initialised");
}
if (lib.whisper_full(ctx, whisperParams, audioData, audioData.length) != 0) {
throw new IOException("Failed to process audio");
}
int nSegments = lib.whisper_full_n_segments(ctx);
List<WhisperSegment> segments= new ArrayList<>(nSegments);
for (int i = 0; i < nSegments; i++) {
long t0 = lib.whisper_full_get_segment_t0(ctx, i);
String text = lib.whisper_full_get_segment_text(ctx, i);
long t1 = lib.whisper_full_get_segment_t1(ctx, i);
segments.add(new WhisperSegment(t0,t1,text));
}
return segments;
}
// public int getTextSegmentCount(Pointer ctx) {
// return lib.whisper_full_n_segments(ctx);
// }
// public String getTextSegment(Pointer ctx, int index) {
// return lib.whisper_full_get_segment_text(ctx, index);
// }
public String getSystemInfo() {
return lib.whisper_print_system_info();
}
public int benchMemcpy(int nthread) {
return lib.whisper_bench_memcpy(nthread);
}
public int benchGgmlMulMat(int nthread) {
return lib.whisper_bench_ggml_mul_mat(nthread);
}
}

View File

@ -1,390 +0,0 @@
package io.github.ggerganov.whispercpp;
import com.sun.jna.Library;
import com.sun.jna.Native;
import com.sun.jna.Pointer;
import io.github.ggerganov.whispercpp.model.WhisperModelLoader;
import io.github.ggerganov.whispercpp.model.WhisperTokenData;
import io.github.ggerganov.whispercpp.params.WhisperContextParams;
import io.github.ggerganov.whispercpp.params.WhisperFullParams;
public interface WhisperCppJnaLibrary extends Library {
WhisperCppJnaLibrary instance = Native.load("whisper", WhisperCppJnaLibrary.class);
String whisper_print_system_info();
/**
* DEPRECATED. Allocate (almost) all memory needed for the model by loading from a file.
*
* @param path_model Path to the model file
* @return Whisper context on success, null on failure
*/
Pointer whisper_init_from_file(String path_model);
/**
* Provides default params which can be used with `whisper_init_from_file_with_params()` etc.
* Because this function allocates memory for the params, the caller must call either:
* - call `whisper_free_context_params()`
* - `Native.free(Pointer.nativeValue(pointer));`
*/
Pointer whisper_context_default_params_by_ref();
void whisper_free_context_params(Pointer params);
/**
* Allocate (almost) all memory needed for the model by loading from a file.
*
* @param path_model Path to the model file
* @param params Pointer to whisper_context_params
* @return Whisper context on success, null on failure
*/
Pointer whisper_init_from_file_with_params(String path_model, WhisperContextParams.ByValue params);
/**
* Allocate (almost) all memory needed for the model by loading from a buffer.
*
* @param buffer Model buffer
* @param buffer_size Size of the model buffer
* @return Whisper context on success, null on failure
*/
Pointer whisper_init_from_buffer(Pointer buffer, int buffer_size);
/**
* Allocate (almost) all memory needed for the model using a model loader.
*
* @param loader Model loader
* @return Whisper context on success, null on failure
*/
Pointer whisper_init(WhisperModelLoader loader);
/**
* Allocate (almost) all memory needed for the model by loading from a file without allocating the state.
*
* @param path_model Path to the model file
* @return Whisper context on success, null on failure
*/
Pointer whisper_init_from_file_no_state(String path_model);
/**
* Allocate (almost) all memory needed for the model by loading from a buffer without allocating the state.
*
* @param buffer Model buffer
* @param buffer_size Size of the model buffer
* @return Whisper context on success, null on failure
*/
Pointer whisper_init_from_buffer_no_state(Pointer buffer, int buffer_size);
// Pointer whisper_init_from_buffer_no_state(Pointer buffer, long buffer_size);
/**
* Allocate (almost) all memory needed for the model using a model loader without allocating the state.
*
* @param loader Model loader
* @return Whisper context on success, null on failure
*/
Pointer whisper_init_no_state(WhisperModelLoader loader);
/**
* Allocate memory for the Whisper state.
*
* @param ctx Whisper context
* @return Whisper state on success, null on failure
*/
Pointer whisper_init_state(Pointer ctx);
/**
* Free all allocated memory associated with the Whisper context.
*
* @param ctx Whisper context
*/
void whisper_free(Pointer ctx);
/**
* Free all allocated memory associated with the Whisper state.
*
* @param state Whisper state
*/
void whisper_free_state(Pointer state);
/**
* Convert RAW PCM audio to log mel spectrogram.
* The resulting spectrogram is stored inside the default state of the provided whisper context.
*
* @param ctx - Pointer to a WhisperContext
* @return 0 on success
*/
int whisper_pcm_to_mel(Pointer ctx, final float[] samples, int n_samples, int n_threads);
/**
* @param ctx Pointer to a WhisperContext
* @param state Pointer to WhisperState
* @param n_samples
* @param n_threads
* @return 0 on success
*/
int whisper_pcm_to_mel_with_state(Pointer ctx, Pointer state, final float[] samples, int n_samples, int n_threads);
/**
* This can be used to set a custom log mel spectrogram inside the default state of the provided whisper context.
* Use this instead of whisper_pcm_to_mel() if you want to provide your own log mel spectrogram.
* n_mel must be 80
* @return 0 on success
*/
int whisper_set_mel(Pointer ctx, final float[] data, int n_len, int n_mel);
int whisper_set_mel_with_state(Pointer ctx, Pointer state, final float[] data, int n_len, int n_mel);
/**
* Run the Whisper encoder on the log mel spectrogram stored inside the default state in the provided whisper context.
* Make sure to call whisper_pcm_to_mel() or whisper_set_mel() first.
* Offset can be used to specify the offset of the first frame in the spectrogram.
* @return 0 on success
*/
int whisper_encode(Pointer ctx, int offset, int n_threads);
int whisper_encode_with_state(Pointer ctx, Pointer state, int offset, int n_threads);
/**
* Run the Whisper decoder to obtain the logits and probabilities for the next token.
* Make sure to call whisper_encode() first.
* tokens + n_tokens is the provided context for the decoder.
* n_past is the number of tokens to use from previous decoder calls.
* Returns 0 on success
* TODO: add support for multiple decoders
*/
int whisper_decode(Pointer ctx, Pointer tokens, int n_tokens, int n_past, int n_threads);
/**
* @param ctx
* @param state
* @param tokens Pointer to int tokens
* @param n_tokens
* @param n_past
* @param n_threads
* @return
*/
int whisper_decode_with_state(Pointer ctx, Pointer state, Pointer tokens, int n_tokens, int n_past, int n_threads);
/**
* Convert the provided text into tokens.
* The tokens pointer must be large enough to hold the resulting tokens.
* Returns the number of tokens on success, no more than n_max_tokens
* Returns -1 on failure
* TODO: not sure if correct
*/
int whisper_tokenize(Pointer ctx, String text, Pointer tokens, int n_max_tokens);
/** Largest language id (i.e. number of available languages - 1) */
int whisper_lang_max_id();
/**
* @return the id of the specified language, returns -1 if not found.
* Examples:
* "de" -&gt; 2
* "german" -&gt; 2
*/
int whisper_lang_id(String lang);
/** @return the short string of the specified language id (e.g. 2 -&gt; "de"), returns nullptr if not found */
String whisper_lang_str(int id);
/**
* Use mel data at offset_ms to try and auto-detect the spoken language.
* Make sure to call whisper_pcm_to_mel() or whisper_set_mel() first
* Returns the top language id or negative on failure
* If not null, fills the lang_probs array with the probabilities of all languages
* The array must be whisper_lang_max_id() + 1 in size
*
* ref: https://github.com/openai/whisper/blob/main/whisper/decoding.py#L18-L69
*/
int whisper_lang_auto_detect(Pointer ctx, int offset_ms, int n_threads, float[] lang_probs);
int whisper_lang_auto_detect_with_state(Pointer ctx, Pointer state, int offset_ms, int n_threads, float[] lang_probs);
int whisper_n_len (Pointer ctx); // mel length
int whisper_n_len_from_state(Pointer state); // mel length
int whisper_n_vocab (Pointer ctx);
int whisper_n_text_ctx (Pointer ctx);
int whisper_n_audio_ctx (Pointer ctx);
int whisper_is_multilingual (Pointer ctx);
int whisper_model_n_vocab (Pointer ctx);
int whisper_model_n_audio_ctx (Pointer ctx);
int whisper_model_n_audio_state(Pointer ctx);
int whisper_model_n_audio_head (Pointer ctx);
int whisper_model_n_audio_layer(Pointer ctx);
int whisper_model_n_text_ctx (Pointer ctx);
int whisper_model_n_text_state (Pointer ctx);
int whisper_model_n_text_head (Pointer ctx);
int whisper_model_n_text_layer (Pointer ctx);
int whisper_model_n_mels (Pointer ctx);
int whisper_model_ftype (Pointer ctx);
int whisper_model_type (Pointer ctx);
/**
* Token logits obtained from the last call to whisper_decode().
* The logits for the last token are stored in the last row
* Rows: n_tokens
* Cols: n_vocab
*/
float[] whisper_get_logits (Pointer ctx);
float[] whisper_get_logits_from_state(Pointer state);
// Token Id -> String. Uses the vocabulary in the provided context
String whisper_token_to_str(Pointer ctx, int token);
String whisper_model_type_readable(Pointer ctx);
// Special tokens
int whisper_token_eot (Pointer ctx);
int whisper_token_sot (Pointer ctx);
int whisper_token_prev(Pointer ctx);
int whisper_token_solm(Pointer ctx);
int whisper_token_not (Pointer ctx);
int whisper_token_beg (Pointer ctx);
int whisper_token_lang(Pointer ctx, int lang_id);
// Task tokens
int whisper_token_translate (Pointer ctx);
int whisper_token_transcribe(Pointer ctx);
// Performance information from the default state.
void whisper_print_timings(Pointer ctx);
void whisper_reset_timings(Pointer ctx);
// Note: Even if `whisper_full_params is stripped back to just 4 ints, JNA throws "Invalid memory access"
// when `whisper_full_default_params()` tries to return a struct.
// WhisperFullParams whisper_full_default_params(int strategy);
/**
* Provides default params which can be used with `whisper_full()` etc.
* Because this function allocates memory for the params, the caller must call either:
* - call `whisper_free_params()`
* - `Native.free(Pointer.nativeValue(pointer));`
*
* @param strategy - WhisperSamplingStrategy.value
*/
Pointer whisper_full_default_params_by_ref(int strategy);
void whisper_free_params(Pointer params);
/**
* Run the entire model: PCM -&gt; log mel spectrogram -&gt; encoder -&gt; decoder -&gt; text
* Not thread safe for same context
* Uses the specified decoding strategy to obtain the text.
*/
int whisper_full(Pointer ctx, WhisperFullParams.ByValue params, final float[] samples, int n_samples);
public int whisper_full_with_state(Pointer ctx, Pointer state, WhisperFullParams.ByValue params, float[] samples, int n_samples);
//int whisper_full_with_state(Pointer ctx, Pointer state, WhisperFullParams params, final float[] samples, int n_samples);
// Split the input audio in chunks and process each chunk separately using whisper_full_with_state()
// Result is stored in the default state of the context
// Not thread safe if executed in parallel on the same context.
// It seems this approach can offer some speedup in some cases.
// However, the transcription accuracy can be worse at the beginning and end of each chunk.
int whisper_full_parallel(Pointer ctx, WhisperFullParams.ByValue params, final float[] samples, int n_samples, int n_processors);
/**
* Number of generated text segments.
* A segment can be a few words, a sentence, or even a paragraph.
* @param ctx Pointer to WhisperContext
*/
int whisper_full_n_segments (Pointer ctx);
/**
* @param state Pointer to WhisperState
*/
int whisper_full_n_segments_from_state(Pointer state);
/**
* Language id associated with the context's default state.
* @param ctx Pointer to WhisperContext
*/
int whisper_full_lang_id(Pointer ctx);
/** Language id associated with the provided state */
int whisper_full_lang_id_from_state(Pointer state);
/** Get the start time of the specified segment. */
long whisper_full_get_segment_t0(Pointer ctx, int i_segment);
/** Get the start time of the specified segment from the state. */
long whisper_full_get_segment_t0_from_state(Pointer state, int i_segment);
/** Get the end time of the specified segment. */
long whisper_full_get_segment_t1(Pointer ctx, int i_segment);
/** Get the end time of the specified segment from the state. */
long whisper_full_get_segment_t1_from_state(Pointer state, int i_segment);
/** Get the text of the specified segment. */
String whisper_full_get_segment_text(Pointer ctx, int i_segment);
/** Get the text of the specified segment from the state. */
String whisper_full_get_segment_text_from_state(Pointer state, int i_segment);
/** Get the number of tokens in the specified segment. */
int whisper_full_n_tokens(Pointer ctx, int i_segment);
/** Get the number of tokens in the specified segment from the state. */
int whisper_full_n_tokens_from_state(Pointer state, int i_segment);
/** Get the token text of the specified token in the specified segment. */
String whisper_full_get_token_text(Pointer ctx, int i_segment, int i_token);
/** Get the token text of the specified token in the specified segment from the state. */
String whisper_full_get_token_text_from_state(Pointer ctx, Pointer state, int i_segment, int i_token);
/** Get the token ID of the specified token in the specified segment. */
int whisper_full_get_token_id(Pointer ctx, int i_segment, int i_token);
/** Get the token ID of the specified token in the specified segment from the state. */
int whisper_full_get_token_id_from_state(Pointer state, int i_segment, int i_token);
/** Get token data for the specified token in the specified segment. */
WhisperTokenData whisper_full_get_token_data(Pointer ctx, int i_segment, int i_token);
/** Get token data for the specified token in the specified segment from the state. */
WhisperTokenData whisper_full_get_token_data_from_state(Pointer state, int i_segment, int i_token);
/** Get the probability of the specified token in the specified segment. */
float whisper_full_get_token_p(Pointer ctx, int i_segment, int i_token);
/** Get the probability of the specified token in the specified segment from the state. */
float whisper_full_get_token_p_from_state(Pointer state, int i_segment, int i_token);
/**
* Benchmark function for memcpy.
*
* @param nThreads Number of threads to use for the benchmark.
* @return The result of the benchmark.
*/
int whisper_bench_memcpy(int nThreads);
/**
* Benchmark function for memcpy as a string.
*
* @param nThreads Number of threads to use for the benchmark.
* @return The result of the benchmark as a string.
*/
String whisper_bench_memcpy_str(int nThreads);
/**
* Benchmark function for ggml_mul_mat.
*
* @param nThreads Number of threads to use for the benchmark.
* @return The result of the benchmark.
*/
int whisper_bench_ggml_mul_mat(int nThreads);
/**
* Benchmark function for ggml_mul_mat as a string.
*
* @param nThreads Number of threads to use for the benchmark.
* @return The result of the benchmark as a string.
*/
String whisper_bench_ggml_mul_mat_str(int nThreads);
}

View File

@ -1,47 +0,0 @@
package io.github.ggerganov.whispercpp.bean;
/**
* Created by litonglinux@qq.com on 10/21/2023_7:48 AM
*/
public class WhisperSegment {
private long start, end;
private String sentence;
public WhisperSegment() {
}
public WhisperSegment(long start, long end, String sentence) {
this.start = start;
this.end = end;
this.sentence = sentence;
}
public long getStart() {
return start;
}
public long getEnd() {
return end;
}
public String getSentence() {
return sentence;
}
public void setStart(long start) {
this.start = start;
}
public void setEnd(long end) {
this.end = end;
}
public void setSentence(String sentence) {
this.sentence = sentence;
}
@Override
public String toString() {
return "[" + start + " --> " + end + "]:" + sentence;
}
}

View File

@ -1,17 +0,0 @@
package io.github.ggerganov.whispercpp.callbacks;
import com.sun.jna.Callback;
/**
* Callback for aborting GGML computation
* Maps to the C typedef: bool (*ggml_abort_callback)(void * data)
*/
public interface GgmlAbortCallback extends Callback {
/**
* Return true to abort the computation, false to continue
*
* @param data User data passed to the callback
* @return true to abort, false to continue
*/
boolean invoke(com.sun.jna.Pointer data);
}

View File

@ -1,24 +0,0 @@
package io.github.ggerganov.whispercpp.callbacks;
import com.sun.jna.Callback;
import com.sun.jna.Pointer;
import io.github.ggerganov.whispercpp.WhisperContext;
import io.github.ggerganov.whispercpp.model.WhisperState;
/**
* Callback before the encoder starts.
* If not null, called before the encoder starts.
* If it returns false, the computation is aborted.
*/
public interface WhisperEncoderBeginCallback extends Callback {
/**
* Callback method before the encoder starts.
*
* @param ctx The whisper context.
* @param state The whisper state.
* @param user_data User data.
* @return True if the computation should proceed, false otherwise.
*/
boolean callback(Pointer ctx, Pointer state, Pointer user_data);
}

View File

@ -1,25 +0,0 @@
package io.github.ggerganov.whispercpp.callbacks;
import com.sun.jna.Callback;
import com.sun.jna.Pointer;
import io.github.ggerganov.whispercpp.model.WhisperTokenData;
/**
* Callback to filter logits.
* Can be used to modify the logits before sampling.
* If not null, called after applying temperature to logits.
*/
public interface WhisperLogitsFilterCallback extends Callback {
/**
* Callback method to filter logits.
*
* @param ctx The whisper context.
* @param state The whisper state.
* @param tokens The array of whisper_token_data.
* @param n_tokens The number of tokens.
* @param logits The array of logits.
* @param user_data User data.
*/
void callback(Pointer ctx, Pointer state, WhisperTokenData[] tokens, int n_tokens, float[] logits, Pointer user_data);
}

View File

@ -1,24 +0,0 @@
package io.github.ggerganov.whispercpp.callbacks;
import com.sun.jna.Callback;
import com.sun.jna.Pointer;
import io.github.ggerganov.whispercpp.WhisperContext;
import io.github.ggerganov.whispercpp.model.WhisperState;
/**
* Callback for the text segment.
* Called on every newly generated text segment.
* Use the whisper_full_...() functions to obtain the text segments.
*/
public interface WhisperNewSegmentCallback extends Callback {
/**
* Callback method for the text segment.
*
* @param ctx The whisper context.
* @param state The whisper state.
* @param n_new The number of newly generated text segments.
* @param user_data User data.
*/
void callback(Pointer ctx, Pointer state, int n_new, Pointer user_data);
}

View File

@ -1,22 +0,0 @@
package io.github.ggerganov.whispercpp.callbacks;
import com.sun.jna.Callback;
import com.sun.jna.Pointer;
import io.github.ggerganov.whispercpp.WhisperContext;
import io.github.ggerganov.whispercpp.model.WhisperState;
/**
* Callback for progress updates.
*/
public interface WhisperProgressCallback extends Callback {
/**
* Callback method for progress updates.
*
* @param ctx The whisper context.
* @param state The whisper state.
* @param progress The progress value.
* @param user_data User data.
*/
void callback(Pointer ctx, Pointer state, int progress, Pointer user_data);
}

View File

@ -1,4 +0,0 @@
package io.github.ggerganov.whispercpp.ggml;
public class GgmlTensor {
}

View File

@ -1,18 +0,0 @@
package io.github.ggerganov.whispercpp.ggml;
public enum GgmlType {
GGML_TYPE_F32,
GGML_TYPE_F16,
GGML_TYPE_Q4_0,
GGML_TYPE_Q4_1,
REMOVED_GGML_TYPE_Q4_2, // support has been removed
REMOVED_GGML_TYPE_Q4_3, // support has been removed
GGML_TYPE_Q5_0,
GGML_TYPE_Q5_1,
GGML_TYPE_Q8_0,
GGML_TYPE_Q8_1,
GGML_TYPE_I8,
GGML_TYPE_I16,
GGML_TYPE_I32,
GGML_TYPE_COUNT,
}

View File

@ -1,10 +0,0 @@
package io.github.ggerganov.whispercpp.model;
public enum EModel {
MODEL_UNKNOWN,
MODEL_TINY,
MODEL_BASE,
MODEL_SMALL,
MODEL_MEDIUM,
MODEL_LARGE,
}

View File

@ -1,49 +0,0 @@
package io.github.ggerganov.whispercpp;
import io.github.ggerganov.whispercpp.ggml.GgmlTensor;
import io.github.ggerganov.whispercpp.model.EModel;
public class WhisperModel {
// EModel type = EModel.MODEL_UNKNOWN;
//
// WhisperHParams hparams;
// WhisperFilters filters;
//
// // encoder.positional_embedding
// GgmlTensor e_pe;
//
// // encoder.conv1
// GgmlTensor e_conv_1_w;
// GgmlTensor e_conv_1_b;
//
// // encoder.conv2
// GgmlTensor e_conv_2_w;
// GgmlTensor e_conv_2_b;
//
// // encoder.ln_post
// GgmlTensor e_ln_w;
// GgmlTensor e_ln_b;
//
// // decoder.positional_embedding
// GgmlTensor d_pe;
//
// // decoder.token_embedding
// GgmlTensor d_te;
//
// // decoder.ln
// GgmlTensor d_ln_w;
// GgmlTensor d_ln_b;
//
// std::vector<whisper_layer_encoder> layers_encoder;
// std::vector<whisper_layer_decoder> layers_decoder;
//
// // context
// struct ggml_context * ctx;
//
// // the model memory buffer is read-only and can be shared between processors
// std::vector<uint8_t> * buf;
//
// // tensors
// int n_loaded;
// Map<String, GgmlTensor> tensors;
}

View File

@ -1,62 +0,0 @@
package io.github.ggerganov.whispercpp.model;
import com.sun.jna.Callback;
import com.sun.jna.Pointer;
import com.sun.jna.Structure;
public class WhisperModelLoader extends Structure {
public Pointer context;
public ReadFunction read;
public EOFFunction eof;
public CloseFunction close;
public static class ReadFunction implements Callback {
public Pointer invoke(Pointer ctx, Pointer output, int readSize) {
// TODO
return ctx;
}
}
public static class EOFFunction implements Callback {
public boolean invoke(Pointer ctx) {
// TODO
return false;
}
}
public static class CloseFunction implements Callback {
public void invoke(Pointer ctx) {
// TODO
}
}
// public WhisperModelLoader(Pointer p) {
// super(p);
// read = new ReadFunction();
// eof = new EOFFunction();
// close = new CloseFunction();
// read.setCallback(this);
// eof.setCallback(this);
// close.setCallback(this);
// read.write();
// eof.write();
// close.write();
// }
public WhisperModelLoader() {
super();
}
public interface ReadCallback extends Callback {
Pointer invoke(Pointer ctx, Pointer output, int readSize);
}
public interface EOFCallback extends Callback {
boolean invoke(Pointer ctx);
}
public interface CloseCallback extends Callback {
void invoke(Pointer ctx);
}
}

View File

@ -1,4 +0,0 @@
package io.github.ggerganov.whispercpp.model;
public class WhisperState {
}

View File

@ -1,50 +0,0 @@
package io.github.ggerganov.whispercpp.model;
import com.sun.jna.Structure;
import java.util.Arrays;
import java.util.List;
/**
* Structure representing token data.
*/
public class WhisperTokenData extends Structure {
/** Token ID. */
public int id;
/** Forced timestamp token ID. */
public int tid;
/** Probability of the token. */
public float p;
/** Log probability of the token. */
public float plog;
/** Probability of the timestamp token. */
public float pt;
/** Sum of probabilities of all timestamp tokens. */
public float ptsum;
/**
* Start time of the token (token-level timestamp data).
* Do not use if you haven't computed token-level timestamps.
*/
public long t0;
/**
* End time of the token (token-level timestamp data).
* Do not use if you haven't computed token-level timestamps.
*/
public long t1;
/** Voice length of the token. */
public float vlen;
@Override
protected List<String> getFieldOrder() {
return Arrays.asList("id", "tid", "p", "plog", "pt", "ptsum", "t0", "t1", "vlen");
}
}

View File

@ -1,19 +0,0 @@
package io.github.ggerganov.whispercpp.params;
import com.sun.jna.Structure;
import java.util.Arrays;
import java.util.List;
public class BeamSearchParams extends Structure {
/** ref: <a href="https://github.com/openai/whisper/blob/f82bc59f5ea234d4b97fb2860842ed38519f7e65/whisper/transcribe.py#L265">...</a> */
public int beam_size;
/** ref: <a href="https://arxiv.org/pdf/2204.05424.pdf">...</a> */
public float patience;
@Override
protected List<String> getFieldOrder() {
return Arrays.asList("beam_size", "patience");
}
}

View File

@ -1,30 +0,0 @@
package io.github.ggerganov.whispercpp.params;
import com.sun.jna.IntegerType;
import java.util.function.BooleanSupplier;
public class CBool extends IntegerType implements BooleanSupplier {
public static final int SIZE = 1;
public static final CBool FALSE = new CBool(0);
public static final CBool TRUE = new CBool(1);
public CBool() {
this(0);
}
public CBool(long value) {
super(SIZE, value, true);
}
@Override
public boolean getAsBoolean() {
return intValue() == 1;
}
@Override
public String toString() {
return intValue() == 1 ? "true" : "false";
}
}

View File

@ -1,16 +0,0 @@
package io.github.ggerganov.whispercpp.params;
import com.sun.jna.Structure;
import java.util.Collections;
import java.util.List;
public class GreedyParams extends Structure {
/** <a href="https://github.com/openai/whisper/blob/f82bc59f5ea234d4b97fb2860842ed38519f7e65/whisper/transcribe.py#L264">...</a> */
public int best_of;
@Override
protected List<String> getFieldOrder() {
return Collections.singletonList("best_of");
}
}

View File

@ -1,30 +0,0 @@
package io.github.ggerganov.whispercpp.params;
import com.sun.jna.*;
import java.util.Arrays;
import java.util.List;
public class WhisperAhead extends Structure {
public int n_text_layer;
public int n_head;
public WhisperAhead() {
super();
}
public WhisperAhead(int textLayer, int head) {
super();
this.n_text_layer = textLayer;
this.n_head = head;
}
@Override
protected List<String> getFieldOrder() {
return Arrays.asList("n_text_layer", "n_head");
}
public static class ByReference extends WhisperAhead implements Structure.ByReference {}
public static class ByValue extends WhisperAhead implements Structure.ByValue {}
}

View File

@ -1,41 +0,0 @@
package io.github.ggerganov.whispercpp.params;
import com.sun.jna.*;
import java.util.Arrays;
import java.util.List;
public class WhisperAheads extends Structure {
public NativeLong n_heads;
public Pointer heads;
public WhisperAheads() {
super();
}
/**
* Create alignment heads from an array of WhisperAhead objects
*/
public void setHeads(WhisperAhead[] aheadsArray) {
this.n_heads = new NativeLong(aheadsArray.length);
int structSize = aheadsArray[0].size();
Memory mem = new Memory(structSize * aheadsArray.length);
for (int i = 0; i < aheadsArray.length; i++) {
aheadsArray[i].write();
byte[] buffer = aheadsArray[i].getPointer().getByteArray(0, structSize);
mem.write(i * structSize, buffer, 0, buffer.length);
}
this.heads = mem;
}
@Override
protected List<String> getFieldOrder() {
return Arrays.asList("n_heads", "heads");
}
public static class ByReference extends WhisperAheads implements Structure.ByReference {}
public static class ByValue extends WhisperAheads implements Structure.ByValue {}
}

View File

@ -1,81 +0,0 @@
package io.github.ggerganov.whispercpp.params;
import com.sun.jna.*;
import java.util.Arrays;
import java.util.List;
/**
* Parameters for the whisper_init_from_file_with_params() function.
* If you change the order or add new parameters, make sure to update the default values in whisper.cpp:
* whisper_context_default_params()
*/
public class WhisperContextParams extends Structure {
public WhisperContextParams(Pointer p) {
super(p);
}
public WhisperContextParams() {
super();
}
/** Use GPU for inference (default = true) */
public CBool use_gpu;
/** Use flash attention (default = false) */
public CBool flash_attn;
/** CUDA device to use (default = 0) */
public int gpu_device;
/** [EXPERIMENTAL] Enable token-level timestamps with DTW (default = false) */
public CBool dtw_token_timestamps;
/** [EXPERIMENTAL] Alignment heads preset for DTW */
public int dtw_aheads_preset;
/** Number of top layers to use for DTW when using WHISPER_AHEADS_N_TOP_MOST preset */
public int dtw_n_top;
public WhisperAheads.ByValue dtw_aheads;
/** DTW memory size (internal use) */
public NativeLong dtw_mem_size;
/** Use GPU for inference */
public void useGpu(boolean enable) {
use_gpu = enable ? CBool.TRUE : CBool.FALSE;
}
/** Use flash attention */
public void useFlashAttn(boolean enable) {
flash_attn = enable ? CBool.TRUE : CBool.FALSE;
}
/** Enable DTW token-level timestamps */
public void enableDtwTokenTimestamps(boolean enable) {
dtw_token_timestamps = enable ? CBool.TRUE : CBool.FALSE;
}
/** Set DTW alignment heads preset */
public void setDtwAheadsPreset(int preset) {
dtw_aheads_preset = preset;
}
@Override
protected List<String> getFieldOrder() {
return Arrays.asList(
"use_gpu",
"flash_attn",
"gpu_device",
"dtw_token_timestamps",
"dtw_aheads_preset",
"dtw_n_top",
"dtw_aheads",
"dtw_mem_size"
);
}
public static class ByValue extends WhisperContextParams implements Structure.ByValue {
public ByValue() { super(); }
public ByValue(Pointer p) { super(p); }
}
}

View File

@ -1,10 +0,0 @@
package io.github.ggerganov.whispercpp.params;
import java.util.List;
public class WhisperFilters {
int n_mel;
int n_fft;
List<Float> data;
}

View File

@ -1,358 +0,0 @@
package io.github.ggerganov.whispercpp.params;
import com.sun.jna.*;
import io.github.ggerganov.whispercpp.callbacks.WhisperEncoderBeginCallback;
import io.github.ggerganov.whispercpp.callbacks.WhisperLogitsFilterCallback;
import io.github.ggerganov.whispercpp.callbacks.WhisperNewSegmentCallback;
import io.github.ggerganov.whispercpp.callbacks.WhisperProgressCallback;
import io.github.ggerganov.whispercpp.callbacks.GgmlAbortCallback;
import java.util.Arrays;
import java.util.List;
/**
* Parameters for the whisper_full() function.
* If you change the order or add new parameters, make sure to update the default values in whisper.cpp:
* whisper_full_default_params()
*/
public class WhisperFullParams extends Structure {
public WhisperFullParams() {
super();
}
public WhisperFullParams(Pointer p) {
super(p);
}
/** Sampling strategy for whisper_full() function. */
public int strategy;
/** Number of threads. (default = 4) */
public int n_threads;
/** Maximum tokens to use from past text as a prompt for the decoder. (default = 16384) */
public int n_max_text_ctx;
/** Start offset in milliseconds. (default = 0) */
public int offset_ms;
/** Audio duration to process in milliseconds. (default = 0) */
public int duration_ms;
/** Translate flag. (default = false) */
public CBool translate;
/** The compliment of translateMode() */
public void transcribeMode() {
translate = CBool.FALSE;
}
/** The compliment of transcribeMode() */
public void translateMode() {
translate = CBool.TRUE;
}
/** Flag to indicate whether to use past transcription (if any) as an initial prompt for the decoder. (default = true) */
public CBool no_context;
/** Flag to indicate whether to use past transcription (if any) as an initial prompt for the decoder. (default = true) */
public void enableContext(boolean enable) {
no_context = enable ? CBool.FALSE : CBool.TRUE;
}
/** Generate timestamps or not? */
public CBool no_timestamps;
/** Flag to force single segment output (useful for streaming). (default = false) */
public CBool single_segment;
/** Flag to force single segment output (useful for streaming). (default = false) */
public void singleSegment(boolean single) {
single_segment = single ? CBool.TRUE : CBool.FALSE;
}
/** Flag to print special tokens (e.g., &lt;SOT&gt;, &lt;EOT&gt;, &lt;BEG&gt;, etc.). (default = false) */
public CBool print_special;
/** Flag to print special tokens (e.g., &lt;SOT&gt;, &lt;EOT&gt;, &lt;BEG&gt;, etc.). (default = false) */
public void printSpecial(boolean enable) {
print_special = enable ? CBool.TRUE : CBool.FALSE;
}
/** Flag to print progress information. (default = true) */
public CBool print_progress;
/** Flag to print progress information. (default = true) */
public void printProgress(boolean enable) {
print_progress = enable ? CBool.TRUE : CBool.FALSE;
}
/** Flag to print results from within whisper.cpp (avoid it, use callback instead). (default = true) */
public CBool print_realtime;
/** Flag to print results from within whisper.cpp (avoid it, use callback instead). (default = true) */
public void printRealtime(boolean enable) {
print_realtime = enable ? CBool.TRUE : CBool.FALSE;
}
/** Flag to print timestamps for each text segment when printing realtime. (default = true) */
public CBool print_timestamps;
/** Flag to print timestamps for each text segment when printing realtime. (default = true) */
public void printTimestamps(boolean enable) {
print_timestamps = enable ? CBool.TRUE : CBool.FALSE;
}
/** [EXPERIMENTAL] Flag to enable token-level timestamps. (default = false) */
public CBool token_timestamps;
/** [EXPERIMENTAL] Flag to enable token-level timestamps. (default = false) */
public void tokenTimestamps(boolean enable) {
token_timestamps = enable ? CBool.TRUE : CBool.FALSE;
}
/** [EXPERIMENTAL] Timestamp token probability threshold (~0.01). (default = 0.01) */
public float thold_pt;
/** [EXPERIMENTAL] Timestamp token sum probability threshold (~0.01). */
public float thold_ptsum;
/** Maximum segment length in characters. (default = 0) */
public int max_len;
/** Flag to split on word rather than on token (when used with max_len). (default = false) */
public CBool split_on_word;
/** Flag to split on word rather than on token (when used with max_len). (default = false) */
public void splitOnWord(boolean enable) {
split_on_word = enable ? CBool.TRUE : CBool.FALSE;
}
/** Maximum tokens per segment (0, default = no limit) */
public int max_tokens;
/** [EXPERIMENTAL] Enable debug mode for extra info */
public CBool debug_mode;
/** Enable debug mode */
public void enableDebugMode(boolean enable) {
debug_mode = enable ? CBool.TRUE : CBool.FALSE;
}
/** Overwrite the audio context size (0 = use default). */
public int audio_ctx;
/** Enable tinydiarize (default = false) */
public CBool tdrz_enable;
/** Enable tinydiarize (default = false) */
public void tdrzEnable(boolean enable) {
tdrz_enable = enable ? CBool.TRUE : CBool.FALSE;
}
/** Regular expression matching tokens to suppress. */
public String suppress_regex;
/** Tokens to provide to the whisper decoder as an initial prompt.
* These are prepended to any existing text context from a previous call. */
public String initial_prompt;
/** Prompt tokens. (int*) */
public Pointer prompt_tokens;
public void setPromptTokens(int[] tokens) {
Memory mem = new Memory(tokens.length * 4L);
mem.write(0, tokens, 0, tokens.length);
prompt_tokens = mem;
}
/** Number of prompt tokens. */
public int prompt_n_tokens;
/** Language for auto-detection.
* For auto-detection, set to `null`, `""`, or "auto". */
public String language;
/** Flag to indicate whether to detect language automatically. */
public CBool detect_language;
/** Flag to indicate whether to detect language automatically. */
public void detectLanguage(boolean enable) {
detect_language = enable ? CBool.TRUE : CBool.FALSE;
}
// Common decoding parameters.
/** Flag to suppress blank tokens. */
public CBool suppress_blank;
public void suppressBlanks(boolean enable) {
suppress_blank = enable ? CBool.TRUE : CBool.FALSE;
}
/** Flag to suppress non-speech tokens. */
public CBool suppress_nst;
/** Flag to suppress non-speech tokens. */
public void suppressNonSpeechTokens(boolean enable) {
suppress_nst = enable ? CBool.TRUE : CBool.FALSE;
}
/** Initial decoding temperature. */
public float temperature;
/** Maximum initial timestamp. */
public float max_initial_ts;
/** Length penalty. */
public float length_penalty;
// Fallback parameters.
/** Temperature increment. */
public float temperature_inc;
/** Entropy threshold (similar to OpenAI's "compression_ratio_threshold"). */
public float entropy_thold;
/** Log probability threshold. */
public float logprob_thold;
/** No speech threshold. */
public float no_speech_thold;
/** Greedy decoding parameters. */
public GreedyParams greedy;
/**
* Beam search decoding parameters.
*/
public BeamSearchParams beam_search;
public void setBestOf(int bestOf) {
if (greedy == null) {
greedy = new GreedyParams();
}
greedy.best_of = bestOf;
}
public void setBeamSize(int beamSize) {
if (beam_search == null) {
beam_search = new BeamSearchParams();
}
beam_search.beam_size = beamSize;
}
public void setBeamSizeAndPatience(int beamSize, float patience) {
if (beam_search == null) {
beam_search = new BeamSearchParams();
}
beam_search.beam_size = beamSize;
beam_search.patience = patience;
}
/**
* Callback for every newly generated text segment.
* WhisperNewSegmentCallback
*/
public Pointer new_segment_callback;
/**
* User data for the new_segment_callback.
*/
public Pointer new_segment_callback_user_data;
/**
* Callback on each progress update.
* WhisperProgressCallback
*/
public Pointer progress_callback;
/**
* User data for the progress_callback.
*/
public Pointer progress_callback_user_data;
/**
* Callback each time before the encoder starts.
* WhisperEncoderBeginCallback
*/
public Pointer encoder_begin_callback;
/**
* User data for the encoder_begin_callback.
*/
public Pointer encoder_begin_callback_user_data;
/** Callback used to abort GGML computation */
public Pointer abort_callback;
/** User data for the abort_callback */
public Pointer abort_callback_user_data;
public void setAbortCallback(GgmlAbortCallback callback) {
abort_callback = CallbackReference.getFunctionPointer(callback);
}
/**
* Callback by each decoder to filter obtained logits.
* WhisperLogitsFilterCallback
*/
public Pointer logits_filter_callback;
/**
* User data for the logits_filter_callback.
*/
public Pointer logits_filter_callback_user_data;
public void setNewSegmentCallback(WhisperNewSegmentCallback callback) {
new_segment_callback = CallbackReference.getFunctionPointer(callback);
}
public void setProgressCallback(WhisperProgressCallback callback) {
progress_callback = CallbackReference.getFunctionPointer(callback);
}
public void setEncoderBeginCallbackeginCallbackCallback(WhisperEncoderBeginCallback callback) {
encoder_begin_callback = CallbackReference.getFunctionPointer(callback);
}
public void setLogitsFilterCallback(WhisperLogitsFilterCallback callback) {
logits_filter_callback = CallbackReference.getFunctionPointer(callback);
}
/** Grammar stuff */
public Pointer grammar_rules;
public long n_grammar_rules;
public long i_start_rule;
public float grammar_penalty;
@Override
protected List<String> getFieldOrder() {
return Arrays.asList("strategy", "n_threads", "n_max_text_ctx",
"offset_ms", "duration_ms", "translate", "no_context",
"no_timestamps", "single_segment", "print_special",
"print_progress", "print_realtime", "print_timestamps",
"token_timestamps", "thold_pt", "thold_ptsum", "max_len",
"split_on_word", "max_tokens", "debug_mode", "audio_ctx",
"tdrz_enable", "suppress_regex", "initial_prompt",
"prompt_tokens", "prompt_n_tokens", "language", "detect_language",
"suppress_blank", "suppress_nst", "temperature",
"max_initial_ts", "length_penalty", "temperature_inc",
"entropy_thold", "logprob_thold", "no_speech_thold", "greedy",
"beam_search", "new_segment_callback", "new_segment_callback_user_data",
"progress_callback", "progress_callback_user_data",
"encoder_begin_callback", "encoder_begin_callback_user_data",
"abort_callback", "abort_callback_user_data",
"logits_filter_callback", "logits_filter_callback_user_data",
"grammar_rules", "n_grammar_rules", "i_start_rule", "grammar_penalty");
}
public static class ByValue extends WhisperFullParams implements Structure.ByValue {
public ByValue() { super(); }
public ByValue(Pointer p) { super(p); }
}
}

View File

@ -1,15 +0,0 @@
package io.github.ggerganov.whispercpp.params;
public class WhisperHParams {
int n_vocab = 51864;
int n_audio_ctx = 1500;
int n_audio_state = 384;
int n_audio_head = 6;
int n_audio_layer = 4;
int n_text_ctx = 448;
int n_text_state = 384;
int n_text_head = 6;
int n_text_layer = 4;
int n_mels = 80;
int ftype = 1;
}

View File

@ -1,10 +0,0 @@
package io.github.ggerganov.whispercpp.params;
/** Available sampling strategies */
public enum WhisperSamplingStrategy {
/** similar to OpenAI's GreedyDecoder */
WHISPER_SAMPLING_GREEDY,
/** similar to OpenAI's BeamSearchDecoder */
WHISPER_SAMPLING_BEAM_SEARCH
}

View File

@ -1,144 +0,0 @@
package io.github.ggerganov.whispercpp;
import static org.junit.jupiter.api.Assertions.*;
import io.github.ggerganov.whispercpp.bean.WhisperSegment;
import io.github.ggerganov.whispercpp.params.CBool;
import io.github.ggerganov.whispercpp.params.WhisperFullParams;
import io.github.ggerganov.whispercpp.params.WhisperSamplingStrategy;
import org.junit.jupiter.api.BeforeAll;
import org.junit.jupiter.api.Test;
import javax.sound.sampled.AudioInputStream;
import javax.sound.sampled.AudioSystem;
import java.io.File;
import java.io.FileNotFoundException;
import java.util.List;
class WhisperCppTest {
private static WhisperCpp whisper = new WhisperCpp();
private static boolean modelInitialised = false;
@BeforeAll
static void init() throws FileNotFoundException {
// By default, models are loaded from ~/.cache/whisper/ and are usually named "ggml-${name}.bin"
// or you can provide the absolute path to the model file.
//String modelName = "../../models/ggml-tiny.bin";
String modelName = "../../models/ggml-tiny.en.bin";
try {
whisper.initContext(modelName);
//whisper.getFullDefaultParams(WhisperSamplingStrategy.WHISPER_SAMPLING_GREEDY);
//whisper.getJavaDefaultParams(WhisperSamplingStrategy.WHISPER_SAMPLING_BEAM_SEARCH);
modelInitialised = true;
} catch (FileNotFoundException ex) {
System.out.println("Model " + modelName + " not found");
}
}
@Test
void testGetDefaultFullParams_BeamSearch() {
// When
WhisperFullParams params = whisper.getFullDefaultParams(WhisperSamplingStrategy.WHISPER_SAMPLING_BEAM_SEARCH);
// Then
assertEquals(WhisperSamplingStrategy.WHISPER_SAMPLING_BEAM_SEARCH.ordinal(), params.strategy);
assertNotEquals(0, params.n_threads);
assertEquals(16384, params.n_max_text_ctx);
assertFalse(params.translate);
assertEquals(0.01f, params.thold_pt);
assertEquals(5, params.beam_search.beam_size);
assertEquals(-1.0f, params.beam_search.patience);
}
@Test
void testGetDefaultFullParams_Greedy() {
// When
WhisperFullParams params = whisper.getFullDefaultParams(WhisperSamplingStrategy.WHISPER_SAMPLING_GREEDY);
// Then
assertEquals(WhisperSamplingStrategy.WHISPER_SAMPLING_GREEDY.ordinal(), params.strategy);
assertNotEquals(0, params.n_threads);
assertEquals(16384, params.n_max_text_ctx);
assertEquals(5, params.greedy.best_of);
}
@Test
void testFullTranscribe() throws Exception {
if (!modelInitialised) {
System.out.println("Model not initialised, skipping test");
return;
}
// Given
File file = new File(System.getProperty("user.dir"), "../../samples/jfk.wav");
AudioInputStream audioInputStream = AudioSystem.getAudioInputStream(file);
byte[] b = new byte[audioInputStream.available()];
float[] floats = new float[b.length / 2];
//WhisperFullParams params = whisper.getFullDefaultParams(WhisperSamplingStrategy.WHISPER_SAMPLING_GREEDY);
WhisperFullParams.ByValue params = whisper.getFullDefaultParams(WhisperSamplingStrategy.WHISPER_SAMPLING_BEAM_SEARCH);
params.setProgressCallback((ctx, state, progress, user_data) -> System.out.println("progress: " + progress));
params.print_progress = CBool.FALSE;
//params.initial_prompt = "and so my fellow Americans um, like";
try {
audioInputStream.read(b);
for (int i = 0, j = 0; i < b.length; i += 2, j++) {
int intSample = (int) (b[i + 1]) << 8 | (int) (b[i]) & 0xFF;
floats[j] = intSample / 32767.0f;
}
// When
String result = whisper.fullTranscribe(params, floats);
// Then
System.err.println(result);
assertEquals("And so my fellow Americans ask not what your country can do for you " +
"ask what you can do for your country.",
result.replace(",", ""));
} finally {
audioInputStream.close();
}
}
@Test
void testFullTranscribeWithTime() throws Exception {
if (!modelInitialised) {
System.out.println("Model not initialised, skipping test");
return;
}
// Given
File file = new File(System.getProperty("user.dir"), "../../samples/jfk.wav");
AudioInputStream audioInputStream = AudioSystem.getAudioInputStream(file);
byte[] b = new byte[audioInputStream.available()];
float[] floats = new float[b.length / 2];
//WhisperFullParams params = whisper.getFullDefaultParams(WhisperSamplingStrategy.WHISPER_SAMPLING_GREEDY);
WhisperFullParams.ByValue params = whisper.getFullDefaultParams(WhisperSamplingStrategy.WHISPER_SAMPLING_BEAM_SEARCH);
params.setProgressCallback((ctx, state, progress, user_data) -> System.out.println("progress: " + progress));
params.print_progress = CBool.FALSE;
//params.initial_prompt = "and so my fellow Americans um, like";
try {
audioInputStream.read(b);
for (int i = 0, j = 0; i < b.length; i += 2, j++) {
int intSample = (int) (b[i + 1]) << 8 | (int) (b[i]) & 0xFF;
floats[j] = intSample / 32767.0f;
}
List<WhisperSegment> segments = whisper.fullTranscribeWithTime(params, floats);
assertTrue(segments.size() > 0, "The size of segments should be greater than 0");
for (WhisperSegment segment : segments) {
System.out.println(segment);
}
} finally {
audioInputStream.close();
}
}
}

View File

@ -1,17 +0,0 @@
package io.github.ggerganov.whispercpp;
import static org.junit.jupiter.api.Assertions.*;
import org.junit.jupiter.api.Test;
class WhisperJnaLibraryTest {
@Test
void testWhisperPrint_system_info() {
String systemInfo = WhisperCppJnaLibrary.instance.whisper_print_system_info();
// eg: "AVX = 1 | AVX2 = 1 | AVX512 = 0 | FMA = 1 | NEON = 0 | ARM_FMA = 0 | F16C = 1 | FP16_VA = 0
// | WASM_SIMD = 0 | BLAS = 0 | SSE3 = 1 | VSX = 0 | COREML = 0 | "
System.out.println("System info: " + systemInfo);
assertTrue(systemInfo.length() > 10);
}
}

View File

@ -9,33 +9,25 @@ target_link_libraries(${TARGET} PRIVATE
)
unset(EXTRA_FLAGS)
if (WHISPER_WASM_SINGLE_FILE)
set(EXTRA_FLAGS "-s SINGLE_FILE=1")
message(STATUS "Embedding WASM inside whisper.js")
add_custom_command(
TARGET ${TARGET} POST_BUILD
TARGET libwhisper POST_BUILD
COMMAND ${CMAKE_COMMAND} -E copy
${CMAKE_BINARY_DIR}/bin/libwhisper.js
${CMAKE_CURRENT_SOURCE_DIR}/whisper.js
)
add_custom_command(
TARGET ${TARGET} POST_BUILD
COMMAND ${CMAKE_COMMAND} -E copy
${CMAKE_BINARY_DIR}/bin/libwhisper.worker.js
${CMAKE_CURRENT_SOURCE_DIR}/libwhisper.worker.js
)
endif()
set_target_properties(${TARGET} PROPERTIES LINK_FLAGS " \
--bind \
-s MODULARIZE=1 \
-s EXPORT_NAME=\"'whisper_factory'\" \
-s FORCE_FILESYSTEM=1 \
-s USE_PTHREADS=1 \
-s PTHREAD_POOL_SIZE=8 \
-s ALLOW_MEMORY_GROWTH=1 \
-s INITIAL_MEMORY=1610612736 \
-s TOTAL_MEMORY=1610612736 \
-s FORCE_FILESYSTEM=1 \
-s EXPORTED_RUNTIME_METHODS=\"['print', 'printErr', 'ccall', 'cwrap']\" \
${EXTRA_FLAGS} \
")

View File

@ -1,81 +0,0 @@
# whisper.cpp
Node.js package for Whisper speech recognition
Package: https://www.npmjs.com/package/whisper.cpp
## Details
The performance is comparable to when running `whisper.cpp` in the browser via WASM.
The API is currently very rudimentary: [bindings/javascript/emscripten.cpp](/bindings/javascript/emscripten.cpp)
For sample usage check [tests/test-whisper.js](/tests/test-whisper.js)
## Package building + test
```bash
# load emscripten
source /path/to/emsdk/emsdk_env.sh
# clone repo
git clone https://github.com/ggerganov/whisper.cpp
cd whisper.cpp
# grab base.en model
./models/download-ggml-model.sh base.en
# prepare PCM sample for testing
ffmpeg -i samples/jfk.wav -f f32le -acodec pcm_f32le samples/jfk.pcmf32
# build
mkdir build-em && cd build-em
emcmake cmake .. && make -j
# run test
node ../tests/test-whisper.js
# For Node.js versions prior to v16.4.0, experimental features need to be enabled:
node --experimental-wasm-threads --experimental-wasm-simd ../tests/test-whisper.js
# publish npm package
make publish-npm
```
## Sample run
```text
$ node --experimental-wasm-threads --experimental-wasm-simd ../tests/test-whisper.js
whisper_model_load: loading model from 'whisper.bin'
whisper_model_load: n_vocab = 51864
whisper_model_load: n_audio_ctx = 1500
whisper_model_load: n_audio_state = 512
whisper_model_load: n_audio_head = 8
whisper_model_load: n_audio_layer = 6
whisper_model_load: n_text_ctx = 448
whisper_model_load: n_text_state = 512
whisper_model_load: n_text_head = 8
whisper_model_load: n_text_layer = 6
whisper_model_load: n_mels = 80
whisper_model_load: f16 = 1
whisper_model_load: type = 2
whisper_model_load: adding 1607 extra tokens
whisper_model_load: mem_required = 506.00 MB
whisper_model_load: ggml ctx size = 140.60 MB
whisper_model_load: memory size = 22.83 MB
whisper_model_load: model size = 140.54 MB
system_info: n_threads = 8 / 10 | AVX = 0 | AVX2 = 0 | AVX512 = 0 | NEON = 0 | F16C = 0 | FP16_VA = 0 | WASM_SIMD = 1 | BLAS = 0 |
operator(): processing 176000 samples, 11.0 sec, 8 threads, 1 processors, lang = en, task = transcribe ...
[00:00:00.000 --> 00:00:11.000] And so my fellow Americans, ask not what your country can do for you, ask what you can do for your country.
whisper_print_timings: load time = 162.37 ms
whisper_print_timings: mel time = 183.70 ms
whisper_print_timings: sample time = 4.27 ms
whisper_print_timings: encode time = 8582.63 ms / 1430.44 ms per layer
whisper_print_timings: decode time = 436.16 ms / 72.69 ms per layer
whisper_print_timings: total time = 9370.90 ms
```

View File

@ -1,58 +1,59 @@
//
// This is the Javascript API of whisper.cpp
//
// Very crude at the moment.
// Feel free to contribute and make this better!
//
// See the tests/test-whisper.js for sample usage
//
#include "whisper.h"
#include <emscripten.h>
#include <emscripten/bind.h>
#include <thread>
#include <vector>
#include <thread>
struct whisper_context * g_context;
std::vector<struct whisper_context *> g_contexts(4, nullptr);
EMSCRIPTEN_BINDINGS(whisper) {
emscripten::function("init", emscripten::optional_override([](const std::string & path_model) {
if (g_context == nullptr) {
g_context = whisper_init_from_file_with_params(path_model.c_str(), whisper_context_default_params());
if (g_context != nullptr) {
return true;
} else {
return false;
for (size_t i = 0; i < g_contexts.size(); ++i) {
if (g_contexts[i] == nullptr) {
g_contexts[i] = whisper_init(path_model.c_str());
if (g_contexts[i] != nullptr) {
return i + 1;
} else {
return (size_t) 0;
}
}
}
return false;
return (size_t) 0;
}));
emscripten::function("free", emscripten::optional_override([]() {
if (g_context) {
whisper_free(g_context);
g_context = nullptr;
emscripten::function("free", emscripten::optional_override([](size_t index) {
--index;
if (index < g_contexts.size()) {
whisper_free(g_contexts[index]);
g_contexts[index] = nullptr;
}
}));
emscripten::function("full_default", emscripten::optional_override([](const emscripten::val & audio, const std::string & lang, bool translate) {
if (g_context == nullptr) {
emscripten::function("full_default", emscripten::optional_override([](size_t index, const emscripten::val & audio, const std::string & lang, bool translate) {
--index;
if (index >= g_contexts.size()) {
return -1;
}
if (g_contexts[index] == nullptr) {
return -2;
}
struct whisper_full_params params = whisper_full_default_params(whisper_sampling_strategy::WHISPER_SAMPLING_GREEDY);
params.print_realtime = true;
params.print_progress = false;
params.print_timestamps = true;
params.print_special = false;
params.translate = translate;
params.language = whisper_is_multilingual(g_context) ? lang.c_str() : "en";
params.n_threads = std::min(8, (int) std::thread::hardware_concurrency());
params.offset_ms = 0;
params.print_realtime = true;
params.print_progress = false;
params.print_timestamps = true;
params.print_special_tokens = false;
params.translate = translate;
params.language = whisper_is_multilingual(g_contexts[index]) ? lang.c_str() : "en";
params.n_threads = std::min(8, (int) std::thread::hardware_concurrency());
params.offset_ms = 0;
std::vector<float> pcmf32;
const int n = audio["length"].as<int>();
@ -67,11 +68,9 @@ EMSCRIPTEN_BINDINGS(whisper) {
// print system information
{
printf("\n");
printf("system_info: n_threads = %d / %d | %s\n",
params.n_threads, std::thread::hardware_concurrency(), whisper_print_system_info());
printf("\n");
printf("%s: processing %d samples, %.1f sec, %d threads, %d processors, lang = %s, task = %s ...\n",
__func__, int(pcmf32.size()), float(pcmf32.size())/WHISPER_SAMPLE_RATE,
params.n_threads, 1,
@ -81,13 +80,10 @@ EMSCRIPTEN_BINDINGS(whisper) {
printf("\n");
}
// run whisper
{
whisper_reset_timings(g_context);
whisper_full(g_context, params, pcmf32.data(), pcmf32.size());
whisper_print_timings(g_context);
}
int ret = whisper_full(g_contexts[index], params, pcmf32.data(), pcmf32.size());
return 0;
whisper_print_timings(g_contexts[index]);
return ret;
}));
}

View File

@ -1 +0,0 @@
"use strict";var Module={};var ENVIRONMENT_IS_NODE=typeof process=="object"&&typeof process.versions=="object"&&typeof process.versions.node=="string";if(ENVIRONMENT_IS_NODE){var nodeWorkerThreads=require("worker_threads");var parentPort=nodeWorkerThreads.parentPort;parentPort.on("message",data=>onmessage({data:data}));var fs=require("fs");Object.assign(global,{self:global,require:require,Module:Module,location:{href:__filename},Worker:nodeWorkerThreads.Worker,importScripts:f=>(0,eval)(fs.readFileSync(f,"utf8")+"//# sourceURL="+f),postMessage:msg=>parentPort.postMessage(msg),performance:global.performance||{now:Date.now}})}var initializedJS=false;function threadPrintErr(){var text=Array.prototype.slice.call(arguments).join(" ");if(ENVIRONMENT_IS_NODE){fs.writeSync(2,text+"\n");return}console.error(text)}function threadAlert(){var text=Array.prototype.slice.call(arguments).join(" ");postMessage({cmd:"alert",text:text,threadId:Module["_pthread_self"]()})}var err=threadPrintErr;self.alert=threadAlert;Module["instantiateWasm"]=(info,receiveInstance)=>{var module=Module["wasmModule"];Module["wasmModule"]=null;var instance=new WebAssembly.Instance(module,info);return receiveInstance(instance)};self.onunhandledrejection=e=>{throw e.reason||e};function handleMessage(e){try{if(e.data.cmd==="load"){let messageQueue=[];self.onmessage=e=>messageQueue.push(e);self.startWorker=instance=>{Module=instance;postMessage({"cmd":"loaded"});for(let msg of messageQueue){handleMessage(msg)}self.onmessage=handleMessage};Module["wasmModule"]=e.data.wasmModule;for(const handler of e.data.handlers){Module[handler]=(...args)=>{postMessage({cmd:"callHandler",handler:handler,args:args})}}Module["wasmMemory"]=e.data.wasmMemory;Module["buffer"]=Module["wasmMemory"].buffer;Module["ENVIRONMENT_IS_PTHREAD"]=true;if(typeof e.data.urlOrBlob=="string"){importScripts(e.data.urlOrBlob)}else{var objectUrl=URL.createObjectURL(e.data.urlOrBlob);importScripts(objectUrl);URL.revokeObjectURL(objectUrl)}whisper_factory(Module)}else if(e.data.cmd==="run"){Module["__emscripten_thread_init"](e.data.pthread_ptr,0,0,1);Module["__emscripten_thread_mailbox_await"](e.data.pthread_ptr);Module["establishStackSpace"]();Module["PThread"].receiveObjectTransfer(e.data);Module["PThread"].threadInitTLS();if(!initializedJS){Module["__embind_initialize_bindings"]();initializedJS=true}try{Module["invokeEntryPoint"](e.data.start_routine,e.data.arg)}catch(ex){if(ex!="unwind"){throw ex}}}else if(e.data.cmd==="cancel"){if(Module["_pthread_self"]()){Module["__emscripten_thread_exit"](-1)}}else if(e.data.target==="setimmediate"){}else if(e.data.cmd==="checkMailbox"){if(initializedJS){Module["checkMailbox"]()}}else if(e.data.cmd){err(`worker.js received unknown command ${e.data.cmd}`);err(e.data)}}catch(ex){if(Module["__emscripten_thread_crashed"]){Module["__emscripten_thread_crashed"]()}throw ex}}self.onmessage=handleMessage;

View File

@ -1,26 +0,0 @@
{
"name": "whisper.cpp",
"version": "@PROJECT_VERSION@",
"description": "Whisper speech recognition",
"main": "whisper.js",
"scripts": {
"test": "echo \"todo: add tests\" && exit 0"
},
"repository": {
"type": "git",
"url": "git+https://github.com/ggerganov/whisper.cpp"
},
"keywords": [
"openai",
"whisper",
"speech-to-text",
"speech-recognition",
"transformer"
],
"author": "Georgi Gerganov",
"license": "MIT",
"bugs": {
"url": "https://github.com/ggerganov/whisper.cpp/issues"
},
"homepage": "https://github.com/ggerganov/whisper.cpp#readme"
}

View File

@ -1,26 +0,0 @@
{
"name": "whisper.cpp",
"version": "1.7.5",
"description": "Whisper speech recognition",
"main": "whisper.js",
"scripts": {
"test": "echo \"todo: add tests\" && exit 0"
},
"repository": {
"type": "git",
"url": "git+https://github.com/ggerganov/whisper.cpp"
},
"keywords": [
"openai",
"whisper",
"speech-to-text",
"speech-recognition",
"transformer"
],
"author": "Georgi Gerganov",
"license": "MIT",
"bugs": {
"url": "https://github.com/ggerganov/whisper.cpp/issues"
},
"homepage": "https://github.com/ggerganov/whisper.cpp#readme"
}

File diff suppressed because one or more lines are too long

View File

@ -1,9 +0,0 @@
LICENSE
pkg/
lib/whisper.*
ext/examples/
ext/ggml/
ext/include/
ext/scripts/
ext/src/
test/fixtures/

View File

@ -1,349 +0,0 @@
whispercpp
==========
![whisper.cpp](https://user-images.githubusercontent.com/1991296/235238348-05d0f6a4-da44-4900-a1de-d0707e75b763.jpeg)
Ruby bindings for [whisper.cpp][], an interface of automatic speech recognition model.
Installation
------------
Install the gem and add to the application's Gemfile by executing:
$ bundle add whispercpp
If bundler is not being used to manage dependencies, install the gem by executing:
$ gem install whispercpp
You can pass build options for whisper.cpp, for instance:
$ bundle config build.whispercpp --enable-ggml-cuda
or,
$ gem install whispercpp -- --enable-ggml-cuda
See whisper.cpp's [README](https://github.com/ggml-org/whisper.cpp/blob/master/README.md) for available options. You need convert options present the README to Ruby-style options, for example:
Boolean options:
* `-DGGML_BLAS=1` -> `--enable-ggml-blas`
* `-DWHISER_COREML=OFF` -> `--disable-whisper-coreml`
Argument options:
* `-DGGML_CUDA_COMPRESSION_MODE=size` -> `--ggml-cuda-compression-mode=size`
Combination:
* `-DGGML_CUDA=1 -DCMAKE_CUDA_ARCHITECTURES="86"` -> `--enable-ggml-cuda --cmake_cuda-architectures="86"`
For boolean options like `GGML_CUDA`, the README says `-DGGML_CUDA=1`. You need strip `-D`, prepend `--enable-` for `1` or `ON` (`--disable-` for `0` or `OFF`) and make it kebab-case: `--enable-ggml-cuda`.
For options which require arguments like `CMAKE_CUDA_ARCHITECTURES`, the README says `-DCMAKE_CUDA_ARCHITECTURES="86"`. You need strip `-D`, prepend `--`, make it kebab-case, append `=` and append argument: `--cmake-cuda-architectures="86"`.
Usage
-----
```ruby
require "whisper"
whisper = Whisper::Context.new("base")
params = Whisper::Params.new(
language: "en",
offset: 10_000,
duration: 60_000,
max_text_tokens: 300,
translate: true,
print_timestamps: false,
initial_prompt: "Initial prompt here."
)
whisper.transcribe("path/to/audio.wav", params) do |whole_text|
puts whole_text
end
```
### Preparing model ###
Some models are prepared up-front:
You also can use shorthand for pre-converted models:
```ruby
whisper = Whisper::Context.new("base.en")
```
You can see the list of prepared model names by `Whisper::Model.pre_converted_models.keys`:
```ruby
puts Whisper::Model.pre_converted_models.keys
# tiny
# tiny.en
# tiny-q5_1
# tiny.en-q5_1
# tiny-q8_0
# base
# base.en
# base-q5_1
# base.en-q5_1
# base-q8_0
# :
# :
```
You can also retrieve each model:
```ruby
base_en = Whisper::Model.pre_converted_models["base.en"]
whisper = Whisper::Context.new(base_en)
```
At first time you use a model, it is downloaded automatically. After that, downloaded cached file is used. To clear cache, call `#clear_cache`:
```ruby
Whisper::Model.pre_converted_models["base"].clear_cache
```
You can also use local model files you prepared:
```ruby
whisper = Whisper::Context.new("path/to/your/model.bin")
```
Or, you can download model files:
```ruby
whisper = Whisper::Context.new("https://example.net/uri/of/your/model.bin")
# Or
whisper = Whisper::Context.new(URI("https://example.net/uri/of/your/model.bin"))
```
See [models][] page for details.
### Preparing audio file ###
Currently, whisper.cpp accepts only 16-bit WAV files.
### Voice Activity Detection (VAD) ###
Support for Voice Activity Detection (VAD) can be enabled by setting `Whisper::Params`'s `vad` argument to `true` and specifying VAD model:
```ruby
Whisper::Params.new(
vad: true,
vad_model_path: "silero-v5.1.2",
# other arguments...
)
```
When you pass the model name (`"silero-v5.1.2"`) or URI (`https://huggingface.co/ggml-org/whisper-vad/resolve/main/ggml-silero-v5.1.2.bin`), it will be downloaded automatically.
Currently, "silero-v5.1.2" is registered as pre-converted model like ASR models. You also specify file path or URI of model.
If you need configure VAD behavior, pass params for that:
```ruby
Whisper::Params.new(
vad: true,
vad_model_path: "silero-v5.1.2",
vad_params: Whisper::VAD::Params.new(
threshold: 1.0, # defaults to 0.5
min_speech_duration_ms: 500, # defaults to 250
min_silence_duration_ms: 200, # defaults to 100
max_speech_duration_s: 30000, # default is FLT_MAX,
speech_pad_ms: 50, # defaults to 30
samples_overlap: 0.5 # defaults to 0.1
),
# other arguments...
)
```
For details on VAD, see [whisper.cpp's README](https://github.com/ggml-org/whisper.cpp?tab=readme-ov-file#voice-activity-detection-vad).
### Output ###
whispercpp supports SRT and WebVTT output:
```ruby
puts whisper.transcribe("path/to/audio.wav", Whisper::Params.new).to_webvtt
# =>
WEBVTT
1
00:00:00.000 --> 00:00:03.860
My thought I have nobody by a beauty and will as you poured.
2
00:00:03.860 --> 00:00:09.840
Mr. Rochester is sub in that so-don't find simplest, and devoted about, to let might in
3
00:00:09.840 --> 00:00:09.940
a
```
You may call `#to_srt`, too
API
---
### Transcription ###
By default, `Whisper::Context#transcribe` works in a single thread. You can make it work in parallel by passing `n_processors` option:
```ruby
whisper.transcribe("path/to/audio.wav", params, n_processors: Etc.nprocessors)
```
Note that transcription occasionally might be low accuracy when it works in parallel.
### Segments ###
Once `Whisper::Context#transcribe` called, you can retrieve segments by `#each_segment`:
```ruby
def format_time(time_ms)
sec, decimal_part = time_ms.divmod(1000)
min, sec = sec.divmod(60)
hour, min = min.divmod(60)
"%02d:%02d:%02d.%03d" % [hour, min, sec, decimal_part]
end
whisper
.transcribe("path/to/audio.wav", params)
.each_segment.with_index do |segment, index|
line = "[%{nth}: %{st} --> %{ed}] %{text}" % {
nth: index + 1,
st: format_time(segment.start_time),
ed: format_time(segment.end_time),
text: segment.text
}
line << " (speaker turned)" if segment.speaker_turn_next?
puts line
end
```
You can also add hook to params called on new segment:
```ruby
# Add hook before calling #transcribe
params.on_new_segment do |segment|
line = "[%{st} --> %{ed}] %{text}" % {
st: format_time(segment.start_time),
ed: format_time(segment.end_time),
text: segment.text
}
line << " (speaker turned)" if segment.speaker_turn_next?
puts line
end
whisper.transcribe("path/to/audio.wav", params)
```
### Models ###
You can see model information:
```ruby
whisper = Whisper::Context.new("base")
model = whisper.model
model.n_vocab # => 51864
model.n_audio_ctx # => 1500
model.n_audio_state # => 512
model.n_audio_head # => 8
model.n_audio_layer # => 6
model.n_text_ctx # => 448
model.n_text_state # => 512
model.n_text_head # => 8
model.n_text_layer # => 6
model.n_mels # => 80
model.ftype # => 1
model.type # => "base"
```
### Logging ###
You can set log callback:
```ruby
prefix = "[MyApp] "
log_callback = ->(level, buffer, user_data) {
case level
when Whisper::LOG_LEVEL_NONE
puts "#{user_data}none: #{buffer}"
when Whisper::LOG_LEVEL_INFO
puts "#{user_data}info: #{buffer}"
when Whisper::LOG_LEVEL_WARN
puts "#{user_data}warn: #{buffer}"
when Whisper::LOG_LEVEL_ERROR
puts "#{user_data}error: #{buffer}"
when Whisper::LOG_LEVEL_DEBUG
puts "#{user_data}debug: #{buffer}"
when Whisper::LOG_LEVEL_CONT
puts "#{user_data}same to previous: #{buffer}"
end
}
Whisper.log_set log_callback, prefix
```
Using this feature, you are also able to suppress log:
```ruby
Whisper.log_set ->(level, buffer, user_data) {
# do nothing
}, nil
Whisper::Context.new("base")
```
### Low-level API to transcribe ###
You can also call `Whisper::Context#full` and `#full_parallel` with a Ruby array as samples. Although `#transcribe` with audio file path is recommended because it extracts PCM samples in C++ and is fast, `#full` and `#full_parallel` give you flexibility.
```ruby
require "whisper"
require "wavefile"
reader = WaveFile::Reader.new("path/to/audio.wav", WaveFile::Format.new(:mono, :float, 16000))
samples = reader.enum_for(:each_buffer).map(&:samples).flatten
whisper = Whisper::Context.new("base")
whisper
.full(Whisper::Params.new, samples)
.each_segment do |segment|
puts segment.text
end
```
The second argument `samples` may be an array, an object with `length` and `each` method, or a MemoryView. If you can prepare audio data as C array and export it as a MemoryView, whispercpp accepts and works with it with zero copy.
Development
-----------
% git clone https://github.com/ggml-org/whisper.cpp.git
% cd whisper.cpp/bindings/ruby
% rake test
First call of `rake test` builds an extension and downloads a model for testing. After that, you add tests in `tests` directory and modify `ext/ruby_whisper.cpp`.
If something seems wrong on build, running `rake clean` solves some cases.
### Need help ###
* Windows support
* Refinement of C/C++ code, especially memory management
License
-------
The same to [whisper.cpp][].
[whisper.cpp]: https://github.com/ggml-org/whisper.cpp
[models]: https://github.com/ggml-org/whisper.cpp/tree/master/models

View File

@ -1,96 +0,0 @@
require 'rake/clean'
require "bundler/gem_tasks"
require "rake/testtask"
require_relative "extsources"
SOURCES_DIR = "ext/sources"
SOURCES = FileList[]
EXTSOURCES.each do |src|
basename = src.pathmap("%f")
dest = basename == "LICENSE" ? basename
: src.pathmap("%{\\.\\./\\.\\.,#{SOURCES_DIR}}p")
.pathmap("%{\\.\\./javascript,#{SOURCES_DIR}/bindings/javascript}p")
dir = dest.pathmap("%d")
file src
directory dir
file dest => [src, dir] do |t|
cp t.source, t.name
end
SOURCES.include dest
end
CLEAN.include SOURCES
SRC = FileList["ext/*.{c,cpp,h}"]
task build: SOURCES
directory "pkg"
CLOBBER.include "pkg"
LIB_NAME = "whisper".ext(RbConfig::CONFIG["DLEXT"])
SO_FILE = File.join("ext", LIB_NAME)
LIB_FILE = File.join("lib", LIB_NAME)
file "ext/Makefile" => SRC + ["ext/extconf.rb"] + SOURCES do |t|
chdir "ext" do
ruby "extconf.rb"
end
end
if File.exist? "ext/Makefile"
task :make_clean do
cd "ext" do
sh "make", "clean"
end
end
task clean: :make_clean
task :make_distclean do
cd "ext" do
sh "make", "distclean"
end
end
task clobber: :make_distclean
end
file SO_FILE => "ext/Makefile" do |t|
chdir "ext" do
sh "make"
end
end
CLEAN.include SO_FILE
directory "lib"
file LIB_FILE => [SO_FILE, "lib"] do |t|
copy t.source, t.name
end
CLEAN.include LIB_FILE
Rake::TestTask.new
TEST_FIXTURE_AUDIO = "test/fixtures/jfk.wav"
TEST_FIXTURE_AUDIO_SRC = File.expand_path(File.join(__dir__, "..", "..", "samples", "jfk.wav"))
TEST_FIXTURE_AUDIO_DIR = TEST_FIXTURE_AUDIO.pathmap("%d")
directory TEST_FIXTURE_AUDIO_DIR
if File.exist? TEST_FIXTURE_AUDIO_SRC
file TEST_FIXTURE_AUDIO => [TEST_FIXTURE_AUDIO_SRC, TEST_FIXTURE_AUDIO_DIR] do |t|
symlink t.source, t.name
end
else
require "open-uri"
file TEST_FIXTURE_AUDIO => TEST_FIXTURE_AUDIO_DIR do |t|
File.write t.name, URI("https://github.com/ggml-org/whisper.cpp/raw/refs/heads/master/samples/jfk.wav").read
end
end
TEST_MEMORY_VIEW = "test/jfk_reader/jfk_reader.#{RbConfig::CONFIG['DLEXT']}"
file TEST_MEMORY_VIEW => "test/jfk_reader/jfk_reader.c" do |t|
chdir "test/jfk_reader" do
ruby "extconf.rb"
sh "make"
end
end
CLEAN.include TEST_MEMORY_VIEW
task test: [LIB_FILE, TEST_MEMORY_VIEW, TEST_FIXTURE_AUDIO]

View File

@ -1,9 +0,0 @@
Makefile
whisper.so
whisper.bundle
whisper.dll
*.o
*.a
sources/*
!sources/CMakeGraphVizOptions.cmake
mkmf.log

View File

@ -1,73 +0,0 @@
require "tsort"
class Dependencies
include TSort
def initialize(cmake, options)
@cmake = cmake
@options = options
@static_lib_shape = nil
@nodes = {}
@graph = Hash.new {|h, k| h[k] = []}
generate_dot
parse_dot
end
def libs
tsort.filter_map {|node|
label, shape = @nodes[node]
if shape == @static_lib_shape
label.gsub(/\\n\([^)]+\)/, '')
else
nil
end
}.reverse.collect {|lib| "lib#{lib}.a"}
end
def to_s
libs.join(" ")
end
private
def dot_path
File.join(__dir__, "build", "whisper.cpp.dot")
end
def generate_dot
args = ["-S", "sources", "-B", "build", "--graphviz", dot_path, "-D", "BUILD_SHARED_LIBS=OFF"]
args << @options.to_s unless @options.to_s.empty?
system @cmake, *args, exception: true
end
def parse_dot
File.open(dot_path).each_line do |line|
case line
when /\[\s*label\s*=\s*"Static Library"\s*,\s*shape\s*=\s*(?<shape>\w+)\s*\]/
@static_lib_shape = $~[:shape]
when /\A\s*"(?<node>\w+)"\s*\[\s*label\s*=\s*"(?<label>\S+)"\s*,\s*shape\s*=\s*(?<shape>\w+)\s*\]\s*;\s*\z/
node = $~[:node]
label = $~[:label]
shape = $~[:shape]
@nodes[node] = [label, shape]
when /\A\s*"(?<depender>\w+)"\s*->\s*"(?<dependee>\w+)"/
depender = $~[:depender]
dependee = $~[:dependee]
@graph[depender] << dependee
end
end
end
def tsort_each_node
@nodes.each_key do |node|
yield node
end
end
def tsort_each_child(node)
@graph[node].each do |child|
yield child
end
end
end

View File

@ -1,22 +0,0 @@
require "mkmf"
require_relative "options"
require_relative "dependencies"
cmake = find_executable("cmake") || abort
options = Options.new(cmake)
have_library("gomp") rescue nil
libs = Dependencies.new(cmake, options)
$INCFLAGS << " -Isources/include -Isources/ggml/include -Isources/examples"
$LOCAL_LIBS << " #{libs}"
$cleanfiles << " build #{libs}"
create_makefile "whisper" do |conf|
conf << <<~EOF
$(TARGET_SO): #{libs}
#{libs}: cmake-targets
cmake-targets:
#{"\t"}#{cmake} -S sources -B build -D BUILD_SHARED_LIBS=OFF -D CMAKE_ARCHIVE_OUTPUT_DIRECTORY=#{__dir__} -D CMAKE_POSITION_INDEPENDENT_CODE=ON #{options}
#{"\t"}#{cmake} --build build --config Release --target common whisper
EOF
end

View File

@ -1,82 +0,0 @@
class Options
def initialize(cmake="cmake")
@cmake = cmake
@options = {}
configure
end
def to_s
@options
.reject {|name, (type, value)| value.nil?}
.collect {|name, (type, value)| "-D #{name}=#{value == true ? "ON" : value == false ? "OFF" : value.shellescape}"}
.join(" ")
end
def cmake_options
return @cmake_options if @cmake_options
output = nil
Dir.chdir __dir__ do
output = `#{@cmake.shellescape} -S sources -B build -L`
end
@cmake_options = output.lines.drop_while {|line| line.chomp != "-- Cache values"}.drop(1)
.filter_map {|line|
option, value = line.chomp.split("=", 2)
name, type = option.split(":", 2)
[
name,
[
type,
type == "BOOL" ? value == "ON" : value
]
]
}.to_h
end
private
def configure
cmake_options.each_pair do |name, (type, default_value)|
option = option_name(name)
value = type == "BOOL" ? enable_config(option) : arg_config("--#{option}")
@options[name] = [type, value]
end
configure_accelerate
configure_metal
configure_coreml
end
# See ggml/src/ggml-cpu/CMakeLists.txt
def configure_accelerate
if RUBY_PLATFORM.match?(/darwin/) && enabled?("GGML_ACCELERATE")
$LDFLAGS << " -framework Accelerate"
end
end
# See ggml/src/ggml-metal/CMakeLists.txt
def configure_metal
$LDFLAGS << " -framework Foundation -framework Metal -framework MetalKit" if enabled?("GGML_METAL")
end
# See src/CmakeLists.txt
def configure_coreml
if enabled?("WHISPER_COREML")
$LDFLAGS << " -framework Foundation -framework CoreML"
$CPPFLAGS << " -DRUBY_WHISPER_USE_COREML"
end
end
def option_name(name)
name.downcase.gsub("_", "-")
end
def enabled?(option)
if @options[option][1].nil?
cmake_options[option][1]
else
@options[option][1]
end
end
end

View File

@ -1,176 +0,0 @@
#include <ruby.h>
#include <ruby/memory_view.h>
#include "ruby_whisper.h"
VALUE mWhisper;
VALUE mVAD;
VALUE cContext;
VALUE cParams;
VALUE cVADParams;
VALUE eError;
VALUE cSegment;
VALUE cModel;
ID id_to_s;
ID id_call;
ID id___method__;
ID id_to_enum;
ID id_length;
ID id_next;
ID id_new;
ID id_to_path;
ID id_URI;
ID id_pre_converted_models;
ID id_coreml_compiled_models;
ID id_cache;
ID id_n_processors;
static bool is_log_callback_finalized = false;
// High level API
extern VALUE ruby_whisper_segment_allocate(VALUE klass);
extern void init_ruby_whisper_context(VALUE *mWhisper);
extern void init_ruby_whisper_params(VALUE *mWhisper);
extern void init_ruby_whisper_error(VALUE *mWhisper);
extern void init_ruby_whisper_segment(VALUE *mWhisper, VALUE *cSegment);
extern void init_ruby_whisper_model(VALUE *mWhisper);
extern void init_ruby_whisper_vad_params(VALUE *mVAD);
extern void register_callbacks(ruby_whisper_params *rwp, VALUE *context);
/*
* call-seq:
* lang_max_id -> Integer
*/
static VALUE ruby_whisper_s_lang_max_id(VALUE self) {
return INT2NUM(whisper_lang_max_id());
}
/*
* call-seq:
* lang_id(lang_name) -> Integer
*/
static VALUE ruby_whisper_s_lang_id(VALUE self, VALUE lang) {
const char * lang_str = StringValueCStr(lang);
const int id = whisper_lang_id(lang_str);
if (-1 == id) {
rb_raise(rb_eArgError, "language not found: %s", lang_str);
}
return INT2NUM(id);
}
/*
* call-seq:
* lang_str(lang_id) -> String
*/
static VALUE ruby_whisper_s_lang_str(VALUE self, VALUE id) {
const int lang_id = NUM2INT(id);
const char * str = whisper_lang_str(lang_id);
if (NULL == str) {
rb_raise(rb_eIndexError, "id %d outside of language id", lang_id);
}
return rb_str_new2(str);
}
/*
* call-seq:
* lang_str(lang_id) -> String
*/
static VALUE ruby_whisper_s_lang_str_full(VALUE self, VALUE id) {
const int lang_id = NUM2INT(id);
const char * str_full = whisper_lang_str_full(lang_id);
if (NULL == str_full) {
rb_raise(rb_eIndexError, "id %d outside of language id", lang_id);
}
return rb_str_new2(str_full);
}
/*
* call-seq:
* system_info_str -> String
*/
static VALUE ruby_whisper_s_system_info_str(VALUE self) {
return rb_str_new2(whisper_print_system_info());
}
static VALUE ruby_whisper_s_finalize_log_callback(VALUE self, VALUE id) {
is_log_callback_finalized = true;
return Qnil;
}
static void
ruby_whisper_log_callback(enum ggml_log_level level, const char * buffer, void * user_data) {
if (is_log_callback_finalized) {
return;
}
VALUE log_callback = rb_iv_get(mWhisper, "log_callback");
VALUE udata = rb_iv_get(mWhisper, "user_data");
rb_funcall(log_callback, id_call, 3, INT2NUM(level), rb_str_new2(buffer), udata);
}
/*
* call-seq:
* log_set ->(level, buffer, user_data) { ... }, user_data -> nil
*/
static VALUE ruby_whisper_s_log_set(VALUE self, VALUE log_callback, VALUE user_data) {
VALUE old_callback = rb_iv_get(self, "log_callback");
if (!NIL_P(old_callback)) {
rb_undefine_finalizer(old_callback);
}
rb_iv_set(self, "log_callback", log_callback);
rb_iv_set(self, "user_data", user_data);
VALUE finalize_log_callback = rb_funcall(mWhisper, rb_intern("method"), 1, rb_str_new2("finalize_log_callback"));
rb_define_finalizer(log_callback, finalize_log_callback);
whisper_log_set(ruby_whisper_log_callback, NULL);
return Qnil;
}
void Init_whisper() {
id_to_s = rb_intern("to_s");
id_call = rb_intern("call");
id___method__ = rb_intern("__method__");
id_to_enum = rb_intern("to_enum");
id_length = rb_intern("length");
id_next = rb_intern("next");
id_new = rb_intern("new");
id_to_path = rb_intern("to_path");
id_URI = rb_intern("URI");
id_pre_converted_models = rb_intern("pre_converted_models");
id_coreml_compiled_models = rb_intern("coreml_compiled_models");
id_cache = rb_intern("cache");
id_n_processors = rb_intern("n_processors");
mWhisper = rb_define_module("Whisper");
mVAD = rb_define_module_under(mWhisper, "VAD");
rb_define_const(mWhisper, "LOG_LEVEL_NONE", INT2NUM(GGML_LOG_LEVEL_NONE));
rb_define_const(mWhisper, "LOG_LEVEL_INFO", INT2NUM(GGML_LOG_LEVEL_INFO));
rb_define_const(mWhisper, "LOG_LEVEL_WARN", INT2NUM(GGML_LOG_LEVEL_WARN));
rb_define_const(mWhisper, "LOG_LEVEL_ERROR", INT2NUM(GGML_LOG_LEVEL_ERROR));
rb_define_const(mWhisper, "LOG_LEVEL_DEBUG", INT2NUM(GGML_LOG_LEVEL_DEBUG));
rb_define_const(mWhisper, "LOG_LEVEL_CONT", INT2NUM(GGML_LOG_LEVEL_CONT));
rb_define_singleton_method(mWhisper, "lang_max_id", ruby_whisper_s_lang_max_id, 0);
rb_define_singleton_method(mWhisper, "lang_id", ruby_whisper_s_lang_id, 1);
rb_define_singleton_method(mWhisper, "lang_str", ruby_whisper_s_lang_str, 1);
rb_define_singleton_method(mWhisper, "lang_str_full", ruby_whisper_s_lang_str_full, 1);
rb_define_singleton_method(mWhisper, "system_info_str", ruby_whisper_s_system_info_str, 0);
rb_define_singleton_method(mWhisper, "log_set", ruby_whisper_s_log_set, 2);
rb_define_private_method(rb_singleton_class(mWhisper), "finalize_log_callback", ruby_whisper_s_finalize_log_callback, 1);
init_ruby_whisper_context(&mWhisper);
init_ruby_whisper_params(&mWhisper);
init_ruby_whisper_error(&mWhisper);
init_ruby_whisper_segment(&mWhisper, &cContext);
init_ruby_whisper_model(&mWhisper);
init_ruby_whisper_vad_params(&mVAD);
rb_require("whisper/context");
rb_require("whisper/segment");
rb_require("whisper/model/uri");
}

View File

@ -1,40 +0,0 @@
#ifndef RUBY_WHISPER_H
#define RUBY_WHISPER_H
#include "whisper.h"
typedef struct {
VALUE *context;
VALUE user_data;
VALUE callback;
VALUE callbacks;
} ruby_whisper_callback_container;
typedef struct {
struct whisper_context *context;
} ruby_whisper;
typedef struct {
struct whisper_full_params params;
bool diarize;
ruby_whisper_callback_container *new_segment_callback_container;
ruby_whisper_callback_container *progress_callback_container;
ruby_whisper_callback_container *encoder_begin_callback_container;
ruby_whisper_callback_container *abort_callback_container;
VALUE vad_params;
} ruby_whisper_params;
typedef struct {
struct whisper_vad_params params;
} ruby_whisper_vad_params;
typedef struct {
VALUE context;
int index;
} ruby_whisper_segment;
typedef struct {
VALUE context;
} ruby_whisper_model;
#endif

View File

@ -1,672 +0,0 @@
#include <ruby.h>
#include <ruby/memory_view.h>
#include "ruby_whisper.h"
extern ID id_to_s;
extern ID id___method__;
extern ID id_to_enum;
extern ID id_length;
extern ID id_next;
extern ID id_new;
extern ID id_to_path;
extern ID id_URI;
extern ID id_pre_converted_models;
extern ID id_coreml_compiled_models;
extern ID id_cache;
extern ID id_n_processors;
extern VALUE cContext;
extern VALUE eError;
extern VALUE cModel;
extern const rb_data_type_t ruby_whisper_params_type;
extern VALUE ruby_whisper_transcribe(int argc, VALUE *argv, VALUE self);
extern VALUE rb_whisper_model_s_new(VALUE context);
extern VALUE rb_whisper_segment_s_new(VALUE context, int index);
extern void prepare_transcription(ruby_whisper_params *rwp, VALUE *context);
ID transcribe_option_names[1];
static void
ruby_whisper_free(ruby_whisper *rw)
{
if (rw->context) {
whisper_free(rw->context);
rw->context = NULL;
}
}
void
rb_whisper_mark(ruby_whisper *rw)
{
// call rb_gc_mark on any ruby references in rw
}
void
rb_whisper_free(void *p)
{
ruby_whisper *rw = (ruby_whisper *)p;
ruby_whisper_free(rw);
free(rw);
}
static size_t
ruby_whisper_memsize(const void *p)
{
const ruby_whisper *rw = (const ruby_whisper *)p;
size_t size = sizeof(rw);
if (!rw) {
return 0;
}
if (rw->context) {
size += sizeof(rw->context);
}
return size;
}
const rb_data_type_t ruby_whisper_type = {
"ruby_whisper",
{0, rb_whisper_free, ruby_whisper_memsize,},
0, 0,
0
};
static VALUE
ruby_whisper_allocate(VALUE klass)
{
ruby_whisper *rw;
VALUE obj = TypedData_Make_Struct(klass, ruby_whisper, &ruby_whisper_type, rw);
rw->context = NULL;
return obj;
}
VALUE
ruby_whisper_normalize_model_path(VALUE model_path)
{
VALUE pre_converted_models = rb_funcall(cModel, id_pre_converted_models, 0);
VALUE pre_converted_model = rb_hash_aref(pre_converted_models, model_path);
if (!NIL_P(pre_converted_model)) {
model_path = pre_converted_model;
#ifdef RUBY_WHISPER_USE_COREML
VALUE coreml_converted_models = rb_funcall(cModel, id_coreml_compiled_models, 0);
VALUE coreml_converted_model = rb_hash_aref(coreml_converted_models, pre_converted_model);
if (!NIL_P(coreml_converted_model)) {
rb_funcall(coreml_converted_model, id_cache, 0);
}
#endif
}
else if (TYPE(model_path) == T_STRING) {
const char * model_path_str = StringValueCStr(model_path);
if (strncmp("http://", model_path_str, 7) == 0 || strncmp("https://", model_path_str, 8) == 0) {
VALUE uri_class = rb_const_get(cModel, id_URI);
model_path = rb_class_new_instance(1, &model_path, uri_class);
}
}
else if (rb_obj_is_kind_of(model_path, rb_path2class("URI::HTTP"))) {
VALUE uri_class = rb_const_get(cModel, id_URI);
model_path = rb_class_new_instance(1, &model_path, uri_class);
}
if (rb_respond_to(model_path, id_to_path)) {
model_path = rb_funcall(model_path, id_to_path, 0);
}
return model_path;
}
/*
* call-seq:
* new("base.en") -> Whisper::Context
* new("path/to/model.bin") -> Whisper::Context
* new(Whisper::Model::URI.new("https://example.net/uri/of/model.bin")) -> Whisper::Context
*/
static VALUE
ruby_whisper_initialize(int argc, VALUE *argv, VALUE self)
{
ruby_whisper *rw;
VALUE whisper_model_file_path;
// TODO: we can support init from buffer here too maybe another ruby object to expose
rb_scan_args(argc, argv, "01", &whisper_model_file_path);
TypedData_Get_Struct(self, ruby_whisper, &ruby_whisper_type, rw);
whisper_model_file_path = ruby_whisper_normalize_model_path(whisper_model_file_path);
if (!rb_respond_to(whisper_model_file_path, id_to_s)) {
rb_raise(rb_eRuntimeError, "Expected file path to model to initialize Whisper::Context");
}
rw->context = whisper_init_from_file_with_params(StringValueCStr(whisper_model_file_path), whisper_context_default_params());
if (rw->context == NULL) {
rb_raise(rb_eRuntimeError, "error: failed to initialize whisper context");
}
return self;
}
/*
* call-seq:
* model_n_vocab -> Integer
*/
VALUE ruby_whisper_model_n_vocab(VALUE self)
{
ruby_whisper *rw;
TypedData_Get_Struct(self, ruby_whisper, &ruby_whisper_type, rw);
return INT2NUM(whisper_model_n_vocab(rw->context));
}
/*
* call-seq:
* model_n_audio_ctx -> Integer
*/
VALUE ruby_whisper_model_n_audio_ctx(VALUE self)
{
ruby_whisper *rw;
TypedData_Get_Struct(self, ruby_whisper, &ruby_whisper_type, rw);
return INT2NUM(whisper_model_n_audio_ctx(rw->context));
}
/*
* call-seq:
* model_n_audio_state -> Integer
*/
VALUE ruby_whisper_model_n_audio_state(VALUE self)
{
ruby_whisper *rw;
TypedData_Get_Struct(self, ruby_whisper, &ruby_whisper_type, rw);
return INT2NUM(whisper_model_n_audio_state(rw->context));
}
/*
* call-seq:
* model_n_audio_head -> Integer
*/
VALUE ruby_whisper_model_n_audio_head(VALUE self)
{
ruby_whisper *rw;
TypedData_Get_Struct(self, ruby_whisper, &ruby_whisper_type, rw);
return INT2NUM(whisper_model_n_audio_head(rw->context));
}
/*
* call-seq:
* model_n_audio_layer -> Integer
*/
VALUE ruby_whisper_model_n_audio_layer(VALUE self)
{
ruby_whisper *rw;
TypedData_Get_Struct(self, ruby_whisper, &ruby_whisper_type, rw);
return INT2NUM(whisper_model_n_audio_layer(rw->context));
}
/*
* call-seq:
* model_n_text_ctx -> Integer
*/
VALUE ruby_whisper_model_n_text_ctx(VALUE self)
{
ruby_whisper *rw;
TypedData_Get_Struct(self, ruby_whisper, &ruby_whisper_type, rw);
return INT2NUM(whisper_model_n_text_ctx(rw->context));
}
/*
* call-seq:
* model_n_text_state -> Integer
*/
VALUE ruby_whisper_model_n_text_state(VALUE self)
{
ruby_whisper *rw;
TypedData_Get_Struct(self, ruby_whisper, &ruby_whisper_type, rw);
return INT2NUM(whisper_model_n_text_state(rw->context));
}
/*
* call-seq:
* model_n_text_head -> Integer
*/
VALUE ruby_whisper_model_n_text_head(VALUE self)
{
ruby_whisper *rw;
TypedData_Get_Struct(self, ruby_whisper, &ruby_whisper_type, rw);
return INT2NUM(whisper_model_n_text_head(rw->context));
}
/*
* call-seq:
* model_n_text_layer -> Integer
*/
VALUE ruby_whisper_model_n_text_layer(VALUE self)
{
ruby_whisper *rw;
TypedData_Get_Struct(self, ruby_whisper, &ruby_whisper_type, rw);
return INT2NUM(whisper_model_n_text_layer(rw->context));
}
/*
* call-seq:
* model_n_mels -> Integer
*/
VALUE ruby_whisper_model_n_mels(VALUE self)
{
ruby_whisper *rw;
TypedData_Get_Struct(self, ruby_whisper, &ruby_whisper_type, rw);
return INT2NUM(whisper_model_n_mels(rw->context));
}
/*
* call-seq:
* model_ftype -> Integer
*/
VALUE ruby_whisper_model_ftype(VALUE self)
{
ruby_whisper *rw;
TypedData_Get_Struct(self, ruby_whisper, &ruby_whisper_type, rw);
return INT2NUM(whisper_model_ftype(rw->context));
}
/*
* call-seq:
* model_type -> String
*/
VALUE ruby_whisper_model_type(VALUE self)
{
ruby_whisper *rw;
TypedData_Get_Struct(self, ruby_whisper, &ruby_whisper_type, rw);
return rb_str_new2(whisper_model_type_readable(rw->context));
}
/*
* Run the entire model: PCM -> log mel spectrogram -> encoder -> decoder -> text
* Not thread safe for same context
* Uses the specified decoding strategy to obtain the text.
*
* call-seq:
* full(params, samples, n_samples) -> nil
* full(params, samples) -> nil
*
* The second argument +samples+ must be an array of samples, respond to :length, or be a MemoryView of an array of float. It must be 32 bit float PCM audio data.
*/
VALUE ruby_whisper_full(int argc, VALUE *argv, VALUE self)
{
if (argc < 2 || argc > 3) {
rb_raise(rb_eArgError, "wrong number of arguments (given %d, expected 2..3)", argc);
}
ruby_whisper *rw;
ruby_whisper_params *rwp;
TypedData_Get_Struct(self, ruby_whisper, &ruby_whisper_type, rw);
VALUE params = argv[0];
TypedData_Get_Struct(params, ruby_whisper_params, &ruby_whisper_params_type, rwp);
VALUE samples = argv[1];
int n_samples;
rb_memory_view_t view;
const bool memory_view_available_p = rb_memory_view_available_p(samples);
if (argc == 3) {
n_samples = NUM2INT(argv[2]);
if (TYPE(samples) == T_ARRAY) {
if (RARRAY_LEN(samples) < n_samples) {
rb_raise(rb_eArgError, "samples length %ld is less than n_samples %d", RARRAY_LEN(samples), n_samples);
}
}
// Should check when samples.respond_to?(:length)?
} else {
if (TYPE(samples) == T_ARRAY) {
if (RARRAY_LEN(samples) > INT_MAX) {
rb_raise(rb_eArgError, "samples are too long");
}
n_samples = (int)RARRAY_LEN(samples);
} else if (memory_view_available_p) {
if (!rb_memory_view_get(samples, &view, RUBY_MEMORY_VIEW_SIMPLE)) {
view.obj = Qnil;
rb_raise(rb_eArgError, "unable to get a memory view");
}
ssize_t n_samples_size = view.byte_size / view.item_size;
if (n_samples_size > INT_MAX) {
rb_raise(rb_eArgError, "samples are too long");
}
n_samples = (int)n_samples_size;
} else if (rb_respond_to(samples, id_length)) {
n_samples = NUM2INT(rb_funcall(samples, id_length, 0));
} else {
rb_raise(rb_eArgError, "samples must respond to :length or be a MemoryView of an array of flaot when n_samples is not given");
}
}
float * c_samples = (float *)malloc(n_samples * sizeof(float));
if (memory_view_available_p) {
c_samples = (float *)view.data;
} else {
if (TYPE(samples) == T_ARRAY) {
for (int i = 0; i < n_samples; i++) {
c_samples[i] = RFLOAT_VALUE(rb_ary_entry(samples, i));
}
} else {
// TODO: use rb_block_call
VALUE iter = rb_funcall(samples, id_to_enum, 1, rb_str_new2("each"));
for (int i = 0; i < n_samples; i++) {
// TODO: check if iter is exhausted and raise ArgumentError appropriately
VALUE sample = rb_funcall(iter, id_next, 0);
c_samples[i] = RFLOAT_VALUE(sample);
}
}
}
prepare_transcription(rwp, &self);
const int result = whisper_full(rw->context, rwp->params, c_samples, n_samples);
if (0 == result) {
return self;
} else {
rb_exc_raise(rb_funcall(eError, id_new, 1, result));
}
}
/*
* Split the input audio in chunks and process each chunk separately using whisper_full_with_state()
* Result is stored in the default state of the context
* Not thread safe if executed in parallel on the same context.
* It seems this approach can offer some speedup in some cases.
* However, the transcription accuracy can be worse at the beginning and end of each chunk.
*
* call-seq:
* full_parallel(params, samples) -> nil
* full_parallel(params, samples, n_samples) -> nil
* full_parallel(params, samples, n_samples, n_processors) -> nil
* full_parallel(params, samples, nil, n_processors) -> nil
*/
static VALUE
ruby_whisper_full_parallel(int argc, VALUE *argv,VALUE self)
{
if (argc < 2 || argc > 4) {
rb_raise(rb_eArgError, "wrong number of arguments (given %d, expected 2..3)", argc);
}
ruby_whisper *rw;
ruby_whisper_params *rwp;
TypedData_Get_Struct(self, ruby_whisper, &ruby_whisper_type, rw);
VALUE params = argv[0];
TypedData_Get_Struct(params, ruby_whisper_params, &ruby_whisper_params_type, rwp);
VALUE samples = argv[1];
int n_samples;
int n_processors;
rb_memory_view_t view;
const bool memory_view_available_p = rb_memory_view_available_p(samples);
switch (argc) {
case 2:
n_processors = 1;
break;
case 3:
n_processors = 1;
break;
case 4:
n_processors = NUM2INT(argv[3]);
break;
}
if (argc >= 3 && !NIL_P(argv[2])) {
n_samples = NUM2INT(argv[2]);
if (TYPE(samples) == T_ARRAY) {
if (RARRAY_LEN(samples) < n_samples) {
rb_raise(rb_eArgError, "samples length %ld is less than n_samples %d", RARRAY_LEN(samples), n_samples);
}
}
// Should check when samples.respond_to?(:length)?
} else if (memory_view_available_p) {
if (!rb_memory_view_get(samples, &view, RUBY_MEMORY_VIEW_SIMPLE)) {
view.obj = Qnil;
rb_raise(rb_eArgError, "unable to get a memory view");
}
ssize_t n_samples_size = view.byte_size / view.item_size;
if (n_samples_size > INT_MAX) {
rb_raise(rb_eArgError, "samples are too long");
}
n_samples = (int)n_samples_size;
} else {
if (TYPE(samples) == T_ARRAY) {
if (RARRAY_LEN(samples) > INT_MAX) {
rb_raise(rb_eArgError, "samples are too long");
}
n_samples = (int)RARRAY_LEN(samples);
} else if (rb_respond_to(samples, id_length)) {
n_samples = NUM2INT(rb_funcall(samples, id_length, 0));
} else {
rb_raise(rb_eArgError, "samples must respond to :length or be a MemoryView of an array of flaot when n_samples is not given");
}
}
float * c_samples = (float *)malloc(n_samples * sizeof(float));
if (memory_view_available_p) {
c_samples = (float *)view.data;
} else {
if (TYPE(samples) == T_ARRAY) {
for (int i = 0; i < n_samples; i++) {
c_samples[i] = RFLOAT_VALUE(rb_ary_entry(samples, i));
}
} else {
// FIXME: use rb_block_call
VALUE iter = rb_funcall(samples, id_to_enum, 1, rb_str_new2("each"));
for (int i = 0; i < n_samples; i++) {
// TODO: check if iter is exhausted and raise ArgumentError
VALUE sample = rb_funcall(iter, id_next, 0);
c_samples[i] = RFLOAT_VALUE(sample);
}
}
}
prepare_transcription(rwp, &self);
const int result = whisper_full_parallel(rw->context, rwp->params, c_samples, n_samples, n_processors);
if (0 == result) {
return self;
} else {
rb_exc_raise(rb_funcall(eError, id_new, 1, result));
}
}
/*
* Number of segments.
*
* call-seq:
* full_n_segments -> Integer
*/
static VALUE
ruby_whisper_full_n_segments(VALUE self)
{
ruby_whisper *rw;
TypedData_Get_Struct(self, ruby_whisper, &ruby_whisper_type, rw);
return INT2NUM(whisper_full_n_segments(rw->context));
}
/*
* Language ID, which can be converted to string by Whisper.lang_str and Whisper.lang_str_full.
*
* call-seq:
* full_lang_id -> Integer
*/
static VALUE
ruby_whisper_full_lang_id(VALUE self)
{
ruby_whisper *rw;
TypedData_Get_Struct(self, ruby_whisper, &ruby_whisper_type, rw);
return INT2NUM(whisper_full_lang_id(rw->context));
}
static int ruby_whisper_full_check_segment_index(const ruby_whisper * rw, const VALUE i_segment)
{
const int c_i_segment = NUM2INT(i_segment);
if (c_i_segment < 0 || c_i_segment >= whisper_full_n_segments(rw->context)) {
rb_raise(rb_eIndexError, "segment index %d out of range", c_i_segment);
}
return c_i_segment;
}
/*
* Start time of a segment indexed by +segment_index+ in centiseconds (10 times milliseconds).
*
* full_get_segment_t0(3) # => 1668 (16680 ms)
*
* call-seq:
* full_get_segment_t0(segment_index) -> Integer
*/
static VALUE
ruby_whisper_full_get_segment_t0(VALUE self, VALUE i_segment)
{
ruby_whisper *rw;
TypedData_Get_Struct(self, ruby_whisper, &ruby_whisper_type, rw);
const int c_i_segment = ruby_whisper_full_check_segment_index(rw, i_segment);
const int64_t t0 = whisper_full_get_segment_t0(rw->context, c_i_segment);
return LONG2NUM(t0);
}
/*
* End time of a segment indexed by +segment_index+ in centiseconds (10 times milliseconds).
*
* full_get_segment_t1(3) # => 1668 (16680 ms)
*
* call-seq:
* full_get_segment_t1(segment_index) -> Integer
*/
static VALUE
ruby_whisper_full_get_segment_t1(VALUE self, VALUE i_segment)
{
ruby_whisper *rw;
TypedData_Get_Struct(self, ruby_whisper, &ruby_whisper_type, rw);
const int c_i_segment = ruby_whisper_full_check_segment_index(rw, i_segment);
const int64_t t1 = whisper_full_get_segment_t1(rw->context, c_i_segment);
return LONG2NUM(t1);
}
/*
* Whether the next segment indexed by +segment_index+ is predicated as a speaker turn.
*
* full_get_segment_speacker_turn_next(3) # => true
*
* call-seq:
* full_get_segment_speacker_turn_next(segment_index) -> bool
*/
static VALUE
ruby_whisper_full_get_segment_speaker_turn_next(VALUE self, VALUE i_segment)
{
ruby_whisper *rw;
TypedData_Get_Struct(self, ruby_whisper, &ruby_whisper_type, rw);
const int c_i_segment = ruby_whisper_full_check_segment_index(rw, i_segment);
const bool speaker_turn_next = whisper_full_get_segment_speaker_turn_next(rw->context, c_i_segment);
return speaker_turn_next ? Qtrue : Qfalse;
}
/*
* Text of a segment indexed by +segment_index+.
*
* full_get_segment_text(3) # => "ask not what your country can do for you, ..."
*
* call-seq:
* full_get_segment_text(segment_index) -> String
*/
static VALUE
ruby_whisper_full_get_segment_text(VALUE self, VALUE i_segment)
{
ruby_whisper *rw;
TypedData_Get_Struct(self, ruby_whisper, &ruby_whisper_type, rw);
const int c_i_segment = ruby_whisper_full_check_segment_index(rw, i_segment);
const char * text = whisper_full_get_segment_text(rw->context, c_i_segment);
return rb_str_new2(text);
}
/*
* call-seq:
* full_get_segment_no_speech_prob(segment_index) -> Float
*/
static VALUE
ruby_whisper_full_get_segment_no_speech_prob(VALUE self, VALUE i_segment)
{
ruby_whisper *rw;
TypedData_Get_Struct(self, ruby_whisper, &ruby_whisper_type, rw);
const int c_i_segment = ruby_whisper_full_check_segment_index(rw, i_segment);
const float no_speech_prob = whisper_full_get_segment_no_speech_prob(rw->context, c_i_segment);
return DBL2NUM(no_speech_prob);
}
// High level API
static VALUE
ruby_whisper_full_get_segment(VALUE self, VALUE i_segment)
{
return rb_whisper_segment_s_new(self, NUM2INT(i_segment));
}
/*
* Yields each Whisper::Segment:
*
* whisper.transcribe("path/to/audio.wav", params)
* whisper.each_segment do |segment|
* puts segment.text
* end
*
* Returns an Enumerator if no block given:
*
* whisper.transcribe("path/to/audio.wav", params)
* enum = whisper.each_segment
* enum.to_a # => [#<Whisper::Segment>, ...]
*
* call-seq:
* each_segment {|segment| ... }
* each_segment -> Enumerator
*/
static VALUE
ruby_whisper_each_segment(VALUE self)
{
if (!rb_block_given_p()) {
const VALUE method_name = rb_funcall(self, id___method__, 0);
return rb_funcall(self, id_to_enum, 1, method_name);
}
ruby_whisper *rw;
TypedData_Get_Struct(self, ruby_whisper, &ruby_whisper_type, rw);
const int n_segments = whisper_full_n_segments(rw->context);
for (int i = 0; i < n_segments; ++i) {
rb_yield(rb_whisper_segment_s_new(self, i));
}
return self;
}
/*
* call-seq:
* model -> Whisper::Model
*/
static VALUE
ruby_whisper_get_model(VALUE self)
{
return rb_whisper_model_s_new(self);
}
void
init_ruby_whisper_context(VALUE *mWhisper)
{
cContext = rb_define_class_under(*mWhisper, "Context", rb_cObject);
transcribe_option_names[0] = id_n_processors;
rb_define_alloc_func(cContext, ruby_whisper_allocate);
rb_define_method(cContext, "initialize", ruby_whisper_initialize, -1);
rb_define_method(cContext, "transcribe", ruby_whisper_transcribe, -1);
rb_define_method(cContext, "model_n_vocab", ruby_whisper_model_n_vocab, 0);
rb_define_method(cContext, "model_n_audio_ctx", ruby_whisper_model_n_audio_ctx, 0);
rb_define_method(cContext, "model_n_audio_state", ruby_whisper_model_n_audio_state, 0);
rb_define_method(cContext, "model_n_audio_head", ruby_whisper_model_n_audio_head, 0);
rb_define_method(cContext, "model_n_audio_layer", ruby_whisper_model_n_audio_layer, 0);
rb_define_method(cContext, "model_n_text_ctx", ruby_whisper_model_n_text_ctx, 0);
rb_define_method(cContext, "model_n_text_state", ruby_whisper_model_n_text_state, 0);
rb_define_method(cContext, "model_n_text_head", ruby_whisper_model_n_text_head, 0);
rb_define_method(cContext, "model_n_text_layer", ruby_whisper_model_n_text_layer, 0);
rb_define_method(cContext, "model_n_mels", ruby_whisper_model_n_mels, 0);
rb_define_method(cContext, "model_ftype", ruby_whisper_model_ftype, 0);
rb_define_method(cContext, "model_type", ruby_whisper_model_type, 0);
rb_define_method(cContext, "full_n_segments", ruby_whisper_full_n_segments, 0);
rb_define_method(cContext, "full_lang_id", ruby_whisper_full_lang_id, 0);
rb_define_method(cContext, "full_get_segment_t0", ruby_whisper_full_get_segment_t0, 1);
rb_define_method(cContext, "full_get_segment_t1", ruby_whisper_full_get_segment_t1, 1);
rb_define_method(cContext, "full_get_segment_speaker_turn_next", ruby_whisper_full_get_segment_speaker_turn_next, 1);
rb_define_method(cContext, "full_get_segment_text", ruby_whisper_full_get_segment_text, 1);
rb_define_method(cContext, "full_get_segment_no_speech_prob", ruby_whisper_full_get_segment_no_speech_prob, 1);
rb_define_method(cContext, "full", ruby_whisper_full, -1);
rb_define_method(cContext, "full_parallel", ruby_whisper_full_parallel, -1);
// High level
rb_define_method(cContext, "full_get_segment", ruby_whisper_full_get_segment, 1);
rb_define_method(cContext, "each_segment", ruby_whisper_each_segment, 0);
rb_define_method(cContext, "model", ruby_whisper_get_model, 0);
}

Some files were not shown because too many files have changed in this diff Show More