dahara1/llama3.1-8b-Instruct-awq

llama3.1-8bのAWQ量子化版です。
4GB超のGPUメモリがあれば高速に動かす事ができます。

This is the AWQ quantization version of llama3.1-8b.
If you have more than 4GB of GPU memory, you can run it at high speed.　　

量子化時に日本語と中国語を多めに使っているため、hugging-quants/Meta-Llama-3.1-8B-Instruct-AWQ-INT4より日本語データを使って計測したPerplexityが良い事がわかっています
Because Japanese and Chinese are used a lot during quantization, It is known that Perplexity measured using Japanese data is better than hugging-quants/Meta-Llama-3.1-8B-Instruct-AWQ-INT4.

セットアップ(setup)

pip install transformers==4.43.3 autoawq==0.2.6 accelerate==0.33.0

サンプルスクリプト(sample script)

import torch
from transformers import AutoModelForCausalLM, AutoTokenizer, AwqConfig

model_id = "dahara1/llama3.1-8b-Instruct-awq"

quantization_config = AwqConfig(
    bits=4,
    fuse_max_seq_len=512, # Note: Update this as per your use-case
    do_fuse=True,
)

tokenizer = AutoTokenizer.from_pretrained(model_id)
model = AutoModelForCausalLM.from_pretrained(
  model_id,
  torch_dtype=torch.float16,
  low_cpu_mem_usage=True,
  device_map="auto",
  quantization_config=quantization_config
)

prompt = [
  {"role": "system", "content": "あなたは親切で役に立つアシスタントです。常に海賊のように返答してください"},
  {"role": "user", "content": "ディープラーニングとは何ですか？"},
]
inputs = tokenizer.apply_chat_template(
  prompt,
  tokenize=True,
  add_generation_prompt=True,
  return_tensors="pt",
  return_dict=True,
).to("cuda")

outputs = model.generate(**inputs, do_sample=True, max_new_tokens=256)
print(tokenizer.batch_decode(outputs[:, inputs['input_ids'].shape[1]:], skip_special_tokens=True)[0])