Local model changing in model.py

#28
by aaromal - opened

Hi,

I am playing in an air-gapped network isolated system and wanted to try the "llama-2-7b.Q2_K.gguf" model locally instead of downloading every time. Below is my model.py. I tried "save_pretrained", "git clone model and point it in AutoModelForCausalLM.from_pretrained" and also tried "path parameter with local_files_only=true" but no luck. All these shows model not found in the directory. Please suggest a way to point to the local directory with model.

model.py:

from threading import Thread
from typing import Iterator

#import torch
from transformers.utils import logging
from ctransformers import AutoModelForCausalLM
from transformers import TextIteratorStreamer, AutoTokenizer

logging.set_verbosity_info()
logger = logging.get_logger("transformers")

config = {"max_new_tokens": 256, "repetition_penalty": 1.1,
"temperature": 0.1, "stream": True}
model_id = "TheBloke/Llama-2-7B-GGUF"
#model_id= "/home/slurm/models/Llama-2-7B-GGUF/"
device = "gpu"

model = AutoModelForCausalLM.from_pretrained(model_id, model_type="llama", lib="avx2", hf=True)

tokenizer = AutoTokenizer.from_pretrained("meta-llama/Llama-2-7b-chat-hf")
tokenizer.save_pretrained('/home/slurm/models/')

#huggingface-cli download TheBloke/Llama-2-7B-GGUF llama-2-7b.Q4_K_M.gguf --local-dir /home/slurm/models/ --local-dir-use-symlinks False

#model = AutoModelForCausalLM.from_pretrained(model_id, model_type="llama", lib="avx2", hf=True)

#PATH = './models/'
#tokenizer = AutoTokenizer.from_pretrained('/home/slurm/models/')
#tokenizer.save_pretrained('/home/slurm/models/')
#model = AutoModelForCausalLM.from_pretrained(PATH='/home/slurm/models/',local_files_only=True, model_type="llama", lib="avx2", hf=True)
#tokenizer.save_pretrained('/home/slurm/models/')
#model.save('/home/slurm/models/')
#model = AutoModelForCausalLM.from_pretrained(PATH, local_files_only=True, model_type="llama", gpu_layers=50, lib="avx2", hf=True)
#tokenizer = AutoTokenizer.from_pretrained("meta-llama/Llama-2-7b-chat-hf")
#tokenizer.save_pretrained('/home/slurm/models/')

def get_prompt(message: str, chat_history: list[tuple[str, str]],
system_prompt: str) -> str:
#logger.info("get_prompt chat_history=%s",chat_history)
#logger.info("get_prompt system_prompt=%s",system_prompt)
texts = [f'[INST] <>\n{system_prompt}\n<>\n\n']
#logger.info("texts=%s",texts)
do_strip = False
for user_input, response in chat_history:
user_input = user_input.strip() if do_strip else user_input
do_strip = True
texts.append(f'{user_input} [/INST] {response.strip()}
[INST] ')
message = message.strip() if do_strip else message
#logger.info("get_prompt message=%s",message)
texts.append(f'{message} [/INST]')
#logger.info("get_prompt final texts=%s",texts)
return ''.join(texts)

def get_input_token_length(message: str, chat_history: list[tuple[str, str]], system_prompt: str) -> int:
#logger.info("get_input_token_length=%s",message)
prompt = get_prompt(message, chat_history, system_prompt)
#logger.info("prompt=%s",prompt)
input_ids = tokenizer([prompt], return_tensors='np', add_special_tokens=False)['input_ids']
#logger.info("input_ids=%s",input_ids)
return input_ids.shape[-1]

def run(message: str,
chat_history: list[tuple[str, str]],
system_prompt: str,
max_new_tokens: int = 1024,
temperature: float = 0.8,
top_p: float = 0.95,
top_k: int = 50) -> Iterator[str]:
prompt = get_prompt(message, chat_history, system_prompt)
inputs = tokenizer([prompt], return_tensors='pt', add_special_tokens=False).to(device)

streamer = TextIteratorStreamer(tokenizer,
                                timeout=15.,
                                skip_prompt=True,
                                skip_special_tokens=True)
generate_kwargs = dict(
    inputs,
    streamer=streamer,
    max_new_tokens=max_new_tokens,
    do_sample=True,
    top_p=top_p,
    top_k=top_k,
    temperature=temperature,
    num_beams=1,
)
t = Thread(target=model.generate, kwargs=generate_kwargs)
t.start()

outputs = []
for text in streamer:
    outputs.append(text)
    yield "".join(outputs)

Thanks,
Aagi

Sign up or log in to comment