whisper.cpp/extra/bench-all.sh
Georgi Gerganov 93935980f8
whisper : Metal and ggml-alloc support (#1270)
* metal : init

* whisper : factor out graph builds

* whisper : allocate encoder and decoder using ggml-alloc

* whisper : ggml-alloc is now supported

* whisper : CoreML support ggml-alloc

* build : fix ggml-alloc

* ios : update submodule

* extra : update sync-ggml.sh script to also sync ggml-alloc

* ci : see if this is causing the crash

* whisper : refactor ggml-alloc init

* whisper.android : try to fix build

* whisper : initial Metal version

* ci : try to debug vmem issue

* metal : decoder works on GPU!

* metal : add multi-decoder support

* ggml : fix ggml_nbytes (probably temp solution)

* metal : run "cross" step on the GPU

* whisper : remove ggml_repeat in the encoder

* whisper : offload the Encoder to Metal

* ggml : use simpler ggml_bytes() implementation

* ggml-alloc : try to make CI happy by reducing vram to 128GB

* whisper : add whisper_allocr to wrap ggml_allocr

* whisper : factor out alloc init in a function

* cmake : update to support Metal build

* whisper : add <functional> header

* objc : fix build (no Metal yet)

* ios : add Metal support

* swiftui : fix build

* metal : speed-up KQ multiplication

* metal : sync latest llama.cpp kernels

* readme : add Metal info

* ios : update submodule

* coreml : add code to toggle Core ML config (CPU, ANE, GPU)

* bench : fix timings by running a pre-heat

* bench : start benching the decoder

* whisper : add ggml_mul_mat_pad

* bench : fix uninitialized vars

* whisper : add comment for disabling mul-mat padding

* whisper : add description of ggml_mul_mat_pad

* whisper : clean-up ggml_mul_mat_pad

* metal : remove the "concurrent" flag

* bench : variable n_past

* ios : update SPM package
2023-09-15 12:18:18 +03:00

92 lines
2.6 KiB
Bash
Executable File

#!/bin/bash
# Helper script to run the bench tool on all models and print the results in share-able format
printf "Usage: ./bench.sh [n_threads] [encoder-only]\n"
if [ -z "$1" ]; then
n_threads=4
else
n_threads=$1
fi
encoder_only=0
if [ -z "$2" ]; then
encoder_only=0
else
encoder_only=$2
fi
models=( \
"tiny" "tiny-q5_0" "tiny-q5_1" "tiny-q8_0" \
"base" "base-q5_0" "base-q5_1" "base-q8_0" \
"small" "small-q5_0" "small-q5_1" "small-q8_0" \
"medium" "medium-q5_0" "medium-q5_1" "medium-q8_0" \
"large" "large-q5_0" "large-q5_1" "large-q8_0" \
)
if [ "$encoder_only" -eq 0 ]; then
printf "\n"
printf "Running memcpy benchmark\n"
printf "\n"
./bench -w 1 -t $n_threads 2>&1
printf "\n"
printf "Running ggml_mul_mat benchmark with $n_threads threads\n"
printf "\n"
./bench -w 2 -t $n_threads 2>&1
printf "\n"
printf "Running benchmark for all models\n"
printf "This can take a while!\n"
printf "\n"
fi
printf "| %6s | %6s | %12s | %9s | %3s | %7s | %7s | %7s | %7s |\n" "CPU" "OS" "Config" "Model" "Th" "Enc." "Dec." "PP" "Commit"
printf "| %6s | %6s | %12s | %9s | %3s | %7s | %7s | %7s | %7s |\n" "---" "---" "---" "---" "---" "---" "---" "---" "---"
for model in "${models[@]}"; do
# actual run
# store stderr output in a variable in order to parse it later
output=$(./bench -m ./models/ggml-$model.bin -t $n_threads 2>&1)
ret=$?
# parse the output:
encode_time=$(echo "$output" | grep "encode time" | awk '{print $11}')
decode_time=$(echo "$output" | grep "decode time" | awk '{print $11}')
prompt_time=$(echo "$output" | grep "prompt time" | awk '{print $11}')
system_info=$(echo "$output" | grep "system_info")
n_threads=$(echo "$output" | grep "system_info" | awk '{print $4}')
# floor to milliseconds
#encode_time=${encode_time%.*}
#decode_time=${decode_time%.*}
#prompt_time=${prompt_time%.*}
config=""
if [[ $system_info == *"AVX2 = 1"* ]]; then
config="$config AVX2"
fi
if [[ $system_info == *"NEON = 1"* ]]; then
config="$config NEON"
fi
if [[ $system_info == *"BLAS = 1"* ]]; then
config="$config BLAS"
fi
if [[ $system_info == *"COREML = 1"* ]]; then
config="$config COREML"
fi
commit=$(git rev-parse --short HEAD)
if [ $ret -eq 0 ]; then
printf "| <todo> | <todo> | %12s | %9s | %3s | %7s | %7s | %7s | %7s |\n" "$config" "$model" "$n_threads" "$encode_time" "$decode_time" "$prompt_time" "$commit"
fi
done