mirror of
https://github.com/ParisNeo/lollms-webui.git
synced 2025-04-07 02:46:45 +00:00
upgraded
This commit is contained in:
saloui
parent
7748b8368d
commit
5c114ad82c
@ -73,6 +73,7 @@ class LLMBackend:
|
||||
"""
|
||||
models_dir = Path('./models')/config["backend"] # replace with the actual path to the models folder
|
||||
return [f.name for f in models_dir.glob(LLMBackend.file_extension)]
|
||||
|
||||
@staticmethod
|
||||
def get_available_models():
|
||||
# Create the file path relative to the child class's directory
|
||||
|
||||
1
backends/backend_template/.gitignore
vendored
Normal file
1
backends/backend_template/.gitignore
vendored
Normal file
@ -0,0 +1 @@
|
||||
config_local.yaml
|
||||
@ -17,16 +17,21 @@ from typing import Callable
|
||||
from api.backend import LLMBackend
|
||||
import yaml
|
||||
from ctransformers import AutoModelForCausalLM
|
||||
from api.config import load_config
|
||||
import re
|
||||
|
||||
__author__ = "parisneo"
|
||||
__github__ = "https://github.com/nomic-ai/gpt4all-ui"
|
||||
__copyright__ = "Copyright 2023, "
|
||||
__license__ = "Apache 2.0"
|
||||
|
||||
backend_name = "GPTJ"
|
||||
backend_name = "CustomBackend"
|
||||
|
||||
class GPTJ(LLMBackend):
|
||||
file_extension='*.bin'
|
||||
class CustomBackend(LLMBackend):
|
||||
# Define what is the extension of the model files supported by your backend
|
||||
# Only applicable for local models for remote models like gpt4 and others, you can keep it empty
|
||||
# and reimplement your own list_models method
|
||||
file_extension='*.bin'
|
||||
def __init__(self, config:dict) -> None:
|
||||
"""Builds a LLAMACPP backend
|
||||
|
||||
@ -34,33 +39,19 @@ class GPTJ(LLMBackend):
|
||||
config (dict): The configuration file
|
||||
"""
|
||||
super().__init__(config, False)
|
||||
if 'gpt2' in self.config['model']:
|
||||
model_type='gpt2'
|
||||
elif 'gptj' in self.config['model']:
|
||||
model_type='gptj'
|
||||
elif 'gpt_neox' in self.config['model']:
|
||||
model_type='gpt_neox'
|
||||
elif 'dolly-v2' in self.config['model']:
|
||||
model_type='dolly-v2'
|
||||
elif 'starcoder' in self.config['model']:
|
||||
model_type='starcoder'
|
||||
elif 'llama' in self.config['model'] or 'wizardLM' in self.config['model']:
|
||||
model_type='llama'
|
||||
elif 'mpt' in self.config['model']:
|
||||
model_type='mpt'
|
||||
else:
|
||||
print("The model you are using is not supported by this backend")
|
||||
return
|
||||
|
||||
|
||||
if self.config["use_avx2"]:
|
||||
self.model = AutoModelForCausalLM.from_pretrained(
|
||||
f"./models/c_transformers/{self.config['model']}", model_type=model_type
|
||||
)
|
||||
# The local config can be used to store personal information that shouldn't be shared like chatgpt Key
|
||||
# or other personal information
|
||||
# This file is never commited to the repository as it is ignored by .gitignore
|
||||
self._local_config_file_path = Path(__file__).parent/"config_local.yaml"
|
||||
if self._local_config_file_path.exists:
|
||||
self.config = load_config(self._local_config_file_path)
|
||||
else:
|
||||
self.model = AutoModelForCausalLM.from_pretrained(
|
||||
f"./models/c_transformers/{self.config['model']}", model_type=model_type, lib = "avx"
|
||||
)
|
||||
self.config = {
|
||||
#Put your default configurations here
|
||||
}
|
||||
|
||||
# Do your initialization stuff
|
||||
|
||||
def tokenize(self, prompt):
|
||||
"""
|
||||
@ -105,19 +96,11 @@ class GPTJ(LLMBackend):
|
||||
self.model.reset()
|
||||
tokens = self.model.tokenize(prompt)
|
||||
count = 0
|
||||
for tok in self.model.generate(
|
||||
tokens,
|
||||
top_k=self.config['top_k'],
|
||||
top_p=self.config['top_p'],
|
||||
temperature=self.config['temperature'],
|
||||
repetition_penalty=self.config['repeat_penalty'],
|
||||
seed=self.config['seed'],
|
||||
batch_size=1,
|
||||
threads = self.config['n_threads'],
|
||||
reset=True,
|
||||
):
|
||||
|
||||
|
||||
generated_text = """
|
||||
This is an empty backend that shows how you can build your own backend.
|
||||
Find it in backends
|
||||
"""
|
||||
for tok in re.split(r' |\n', generated_text):
|
||||
if count >= n_predict or self.model.is_eos_token(tok):
|
||||
break
|
||||
word = self.model.detokenize(tok)
|
||||
@ -126,8 +109,6 @@ class GPTJ(LLMBackend):
|
||||
break
|
||||
output += word
|
||||
count += 1
|
||||
|
||||
|
||||
except Exception as ex:
|
||||
print(ex)
|
||||
return output
|
||||
|
||||
@ -13,6 +13,7 @@ class Install:
|
||||
print("-------------- Template backend -------------------------------")
|
||||
print("This is the first time you are using this backend.")
|
||||
print("Installing ...")
|
||||
# Example of installing py torche
|
||||
"""
|
||||
try:
|
||||
print("Checking pytorch")
|
||||
@ -31,11 +32,11 @@ class Install:
|
||||
requirements_file = current_dir / "requirements.txt"
|
||||
subprocess.run(["pip", "install", "--upgrade", "--no-cache-dir", "-r", str(requirements_file)])
|
||||
|
||||
# Create ther models folder
|
||||
models_folder = Path("./models/c_transformers")
|
||||
# Create the models folder
|
||||
models_folder = Path(f"./models/{Path(__file__).parent.stem}")
|
||||
models_folder.mkdir(exist_ok=True, parents=True)
|
||||
|
||||
#Create the install file
|
||||
#Create the install file (a file that is used to insure the installation was done correctly)
|
||||
with open(install_file,"w") as f:
|
||||
f.write("ok")
|
||||
print("Installed successfully")
|
||||
|
||||
@ -1,52 +1,7 @@
|
||||
- LLAMA: 'true'
|
||||
description: GGML format model files for the original LLaMa
|
||||
filename: llama-7b.ggmlv3.q4_0.bin
|
||||
license: Non commercial
|
||||
owner_link: https://huggingface.co/TheBloke/
|
||||
owner: TheBloke
|
||||
server: https://huggingface.co/TheBloke/LLaMa-7B-GGML/resolve/main/
|
||||
sha256: ec2f2d1f0dfb73b72a4cbac7fa121abbe04c37ab327125a38248f930c0f09ddf
|
||||
- LLAMA: 'true'
|
||||
description: GGML format model files for Wizard LM 7B model
|
||||
filename: wizardLM-7B.ggmlv3.q4_0.bin
|
||||
license: Non commercial
|
||||
owner_link: https://huggingface.co/TheBloke/
|
||||
owner: TheBloke
|
||||
server: https://huggingface.co/TheBloke/wizardLM-7B-GGML/resolve/main/
|
||||
sha256: ea35e30a7c140485b856d0919284ce59e4ca47c1b8af037ea8b7ba05ef291c43
|
||||
- LLAMA: 'true'
|
||||
description: GGML format model files for Wizard LM 7B model
|
||||
filename: koala-7b.ggml.unquantized.pr613.bin
|
||||
license: Non commercial
|
||||
owner_link: https://huggingface.co/TheBloke/
|
||||
owner: TheBloke
|
||||
server: https://huggingface.co/TheBloke/koala-7b-ggml-unquantized/resolve/main/
|
||||
sha256: c478ceced3b38800cb768225b1e759a32c9e89bd33606fb38eeff3b811e28371
|
||||
|
||||
|
||||
|
||||
|
||||
- MPT-7B: 'true'
|
||||
description: MPT-7B
|
||||
filename: mpt-7b.ggmlv3.q5_1.bin
|
||||
license: Apache-2.0
|
||||
owner_link: https://huggingface.co/TheBloke/
|
||||
owner: TheBloke
|
||||
server: https://huggingface.co/TheBloke/MPT-7B-GGML/resolve/main/
|
||||
sha256: c947c38405921a199c603fed2ed63486811777ba370bb51c40c3132e5cfe84e5
|
||||
- MPT-7B-Instruct: 'true'
|
||||
description: MPT-7B-Instruct
|
||||
filename: mpt-7b-instruct.ggmlv3.q5_1.bin
|
||||
license: Apache-2.0
|
||||
owner_link: https://huggingface.co/TheBloke/
|
||||
owner: TheBloke
|
||||
server: https://huggingface.co/TheBloke/MPT-7B-Instruct-GGML/resolve/main/
|
||||
sha256: a4d17a39ac277d48a3d55aa74b36a4e6e1b891b58937a838243fad549d26c686
|
||||
- MPT-7B-Storywriter: 'true'
|
||||
description: MPT-7B-Storywriter
|
||||
filename: mpt-7b-storywriter.ggmlv3.q5_1.bin
|
||||
license: Apache-2.0
|
||||
owner_link: https://huggingface.co/TheBloke/
|
||||
owner: TheBloke
|
||||
server: https://huggingface.co/TheBloke/MPT-7B-Storywriter-GGML/resolve/main/
|
||||
sha256: 3b7dd7aa7508cc8cb4e262fe4b93214826f38d18d04059075e05837457f5402
|
||||
- description: The description of the model
|
||||
filename: the file to be loaded
|
||||
license: The licence
|
||||
owner_link: https://link_to_the_owner_web_page
|
||||
owner: Owner_name
|
||||
server: https://Path_to_the_server_to_load_from
|
||||
sha256: The Hash code of the file
|
||||
Loading…
x
Reference in New Issue
Block a user