budecosystem/genz-13b-v2-4bit

GenZ 13B v2 4bit

The instruction finetuned model with 4K input length. The model is finetuned on top of pretrained LLaMa2

Inference

from transformers import LlamaForCausalLM, LlamaTokenizer
from auto_gptq import AutoGPTQForCausalLM, BaseQuantizeConfig

base_model = 'budecosystem/genz-13b-v2-4bit'

tokenizer = LlamaTokenizer.from_pretrained(base_model)

model = AutoGPTQForCausalLM.from_quantized(model_name_or_path=base_model,
        model_basename="gptq_model-4bit-128g",
        use_safetensors=True,
        trust_remote_code=True)

prompt = """A chat between a curious user and an artificial intelligence assistant. The assistant gives helpful, detailed, and polite answers to the user's questions.
USER: who are you? ASSISTANT: """
inputs = tokenizer(prompt, return_tensors="pt")
sample = model.generate(**inputs, max_length=128)
print(tokenizer.decode(sample[0]))

Use following prompt template

A chat between a curious user and an artificial intelligence assistant. The assistant gives helpful, detailed, and polite answers to the user's questions. USER: Hi, how are you? ASSISTANT: 

Finetuning

python finetune.py
   --model_name meta-llama/Llama-2-13b
   --data_path dataset.json
   --output_dir output
   --trust_remote_code
   --prompt_column instruction
   --response_column output

Check the GitHub for the code -> GenZ