README.md

metadata

base_model: unsloth/gemma-3-4b-it-unsloth-bnb-4bit
library_name: peft
model_name: Gemma-3-4B-it-Medical-LoRA
tags:
  - generated_from_trainer
  - trl
  - unsloth
  - sft
licence: license

Model Card for Gemma-3-4B-it-Medical-LoRA

This model is a fine-tuned version of unsloth/gemma-3-4b-it-unsloth-bnb-4bit. It has been trained using TRL.

Quick start

from transformers import pipeline

question = "If you had a time machine, but could only go to the past or the future once and never return, which would you choose and why?"
generator = pipeline("text-generation", model="None", device="cuda")
output = generator([{"role": "user", "content": question}], max_new_tokens=128, return_full_text=False)[0]
print(output["generated_text"])

Training procedure

This model was trained with SFT.

Inference Usage

import torch
from transformers import AutoModelForCausalLM, AutoTokenizer, pipeline
from peft import PeftModel

device = "cuda" if torch.cuda.is_available() else "cpu"


# Define model and LoRA adapter paths
base_model_name = "google/gemma-3-1b-it"
lora_adapter_name = "danhtran2mind/Gemma-3-1B-Instruct-Vi-Medical-LoRA"

# Load tokenizer
tokenizer = AutoTokenizer.from_pretrained(base_model_name)

# Load base model with optimized settings
model = AutoModelForCausalLM.from_pretrained(
    base_model_name,
    torch_dtype=torch.float16,  # Use FP16 for efficiency
    device_map=device,
    trust_remote_code=True
)

# Apply LoRA adapter
model = PeftModel.from_pretrained(model, lora_adapter_name)

# Set model to evaluation mode
model.eval()

# Define the question
question = ("Khi nghi ngờ bị loét dạ dày tá tràng nên đến khoa nào "
          "tại bệnh viện để thăm khám?")

seed = 42
torch.manual_seed(seed)
if torch.cuda.is_available():
    torch.cuda.manual_seed(seed)
    torch.cuda.manual_seed_all(seed)

# Create text generation pipeline
generator = pipeline(
    "text-generation",
    model=model,
    tokenizer=tokenizer,
    torch_dtype=torch.float16,
    device_map=device,
    max_new_tokens=2048,
    num_return_sequences=1,
    do_sample=True,
    temperature=0.7,
    top_p=0.9,
    top_k=64,
)

# Format input for the pipeline
input_prompt = [{"role": "user", "content": question}]

# Generate response
output = generator(input_prompt, return_full_text=False)[0]

# Print the generated text
print(output["generated_text"])

Framework versions

• PEFT 0.14.0
• TRL: 0.19.0
• Transformers: 4.52.4
• Pytorch: 2.6.0+cu124
• Datasets: 3.6.0
• Tokenizers: 0.21.1

Citations

Cite TRL as:

@misc{vonwerra2022trl,
    title        = {{TRL: Transformer Reinforcement Learning}},
    author       = {Leandro von Werra and Younes Belkada and Lewis Tunstall and Edward Beeching and Tristan Thrush and Nathan Lambert and Shengyi Huang and Kashif Rasul and Quentin Gallou{\'e}dec},
    year         = 2020,
    journal      = {GitHub repository},
    publisher    = {GitHub},
    howpublished = {\url{https://github.com/huggingface/trl}}
}