README.md

metadata

license: apache-2.0
datasets:
  - shibing624/alpaca-zh
language:
  - zh
tags:
  - LoRA
  - LLaMA
  - Alpaca
  - PEFT
  - int8

Model Card for llama-7b-alpaca-zh-20k

Uses

Direct Use

from peft import PeftModel
from transformers import GenerationConfig, LlamaForCausalLM, LlamaTokenizer


max_memory = {i: "15GIB" for i in range(torch.cuda.device_count())}
tokenizer = LlamaTokenizer.from_pretrained(base_model)
model = LlamaForCausalLM.from_pretrained(
    base_model,
    load_in_8bit=True,
    torch_dtype=torch.float16,
    device_map="auto"
    max_memory=max_memory
)
model = PeftModel.from_pretrained(
    model,
    lora_weights,
    torch_dtype=torch.float16,
    max_memory=max_memory
)