mirror of https://github.com/arc53/DocsGPT
You cannot select more than 25 topics
Topics must start with a letter or number, can include dashes ('-') and can be up to 35 characters long.
69 lines
2.1 KiB
Python
69 lines
2.1 KiB
Python
from application.llm.base import BaseLLM
|
|
|
|
|
|
class HuggingFaceLLM(BaseLLM):
|
|
|
|
def __init__(
|
|
self,
|
|
api_key=None,
|
|
user_api_key=None,
|
|
llm_name="Arc53/DocsGPT-7B",
|
|
q=False,
|
|
*args,
|
|
**kwargs,
|
|
):
|
|
global hf
|
|
|
|
from langchain.llms import HuggingFacePipeline
|
|
|
|
if q:
|
|
import torch
|
|
from transformers import (
|
|
AutoModelForCausalLM,
|
|
AutoTokenizer,
|
|
pipeline,
|
|
BitsAndBytesConfig,
|
|
)
|
|
|
|
tokenizer = AutoTokenizer.from_pretrained(llm_name)
|
|
bnb_config = BitsAndBytesConfig(
|
|
load_in_4bit=True,
|
|
bnb_4bit_use_double_quant=True,
|
|
bnb_4bit_quant_type="nf4",
|
|
bnb_4bit_compute_dtype=torch.bfloat16,
|
|
)
|
|
model = AutoModelForCausalLM.from_pretrained(
|
|
llm_name, quantization_config=bnb_config
|
|
)
|
|
else:
|
|
from transformers import AutoModelForCausalLM, AutoTokenizer, pipeline
|
|
|
|
tokenizer = AutoTokenizer.from_pretrained(llm_name)
|
|
model = AutoModelForCausalLM.from_pretrained(llm_name)
|
|
|
|
super().__init__(*args, **kwargs)
|
|
self.api_key = api_key
|
|
self.user_api_key = user_api_key
|
|
pipe = pipeline(
|
|
"text-generation",
|
|
model=model,
|
|
tokenizer=tokenizer,
|
|
max_new_tokens=2000,
|
|
device_map="auto",
|
|
eos_token_id=tokenizer.eos_token_id,
|
|
)
|
|
hf = HuggingFacePipeline(pipeline=pipe)
|
|
|
|
def _raw_gen(self, baseself, model, messages, stream=False, **kwargs):
|
|
context = messages[0]["content"]
|
|
user_question = messages[-1]["content"]
|
|
prompt = f"### Instruction \n {user_question} \n ### Context \n {context} \n ### Answer \n"
|
|
|
|
result = hf(prompt)
|
|
|
|
return result.content
|
|
|
|
def _raw_gen_stream(self, baseself, model, messages, stream=True, **kwargs):
|
|
|
|
raise NotImplementedError("HuggingFaceLLM Streaming is not implemented yet.")
|