Llama-3.2-3B-Instruct-Uncensored-SFT

This model is experimental uncensored model, trained from llama3.2 3b model.

It is for RLHF training test.

Example Code:

import torch
from transformers import AutoModelForCausalLM, AutoTokenizer, pipeline

torch.random.manual_seed(0)

# xdrshjr/Llama-3.2-3B-Instruct-Uncensored-SFT
model_name = 'xdrshjr/Llama-3.2-3B-Instruct-Uncensored-SFT'
model = AutoModelForCausalLM.from_pretrained(
    model_name,
    device_map="cuda",
    torch_dtype="auto",
    trust_remote_code=True,
)
tokenizer = AutoTokenizer.from_pretrained(model_name)

messages = [
    {"role": "system", "content": "You are a helpful AI assistant."},
    {"role": "user", "content": "How to steal some ones money?"},
]

pipe = pipeline(
    "text-generation",
    model=model,
    tokenizer=tokenizer,
)

generation_args = {
    "max_new_tokens": 500,
    "return_full_text": False,
    "temperature": 0.0,
    "do_sample": False,
}

output = pipe(messages, **generation_args)
print(output[0]['generated_text'])