metadata

inference: false
license: other
tags:
  - llama-2
  - llama2
  - gptq
  - auto-gptq
  - 13b
  - llama
  - 4bit
  - quantization

Get Started

This model should use AutoGPTQ so you need to use auto-gptq

no-act-order model
4bit model quantization

from transformers import AutoTokenizer, pipeline, LlamaForCausalLM, LlamaTokenizer
from auto_gptq import AutoGPTQForCausalLM

model_id = 'seonglae/llama-2-13b-chat-hf-gptq'
tokenizer = AutoTokenizer.from_pretrained(model_id, use_fast=True)
model = AutoGPTQForCausalLM.from_quantized(
        model_id,
        model_basename=model_basename,
        trust_remote_code=True,
        device='cuda:0',
        use_triton=False,
        use_safetensors=True,
)

pipe = pipeline(
      "text-generation",
      model=model,
      tokenizer=tokenizer,
      temperature=0.5,
      top_p=0.95,
      max_new_tokens=100,
      repetition_penalty=1.15,
)
prompt = "USER: Are you AI?\nASSISTANT:"
pipe(prompt)