Modelpublic

lukasdrews/Gemma-4-E2B-IT-SFT-RLVR-Medical

sourceHugging Faceapache-2.0updated 5mo agoView on Hugging Face

2likes46downloads

Model Card

Gemma-4-E2B-IT-SFT-RLVR-Medical

Gemma-4-E2B-it fine-tuned on PubMedQA using SFT and RLVR.<br> Also check out the training code on GitHub.<br> Quantized models are available here.

Setup

python

#!pip install transformers, torch, accelerate
from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("lukasdrews/Gemma-4-E2B-IT-SFT-RLVR-Medical")
model = AutoModelForCausalLM.from_pretrained("lukasdrews/Gemma-4-E2B-IT-SFT-RLVR-Medical")
messages = [
    {
        "role": "user",
        "content": [
            {"type": "text", "text": "Do GEC produce and bear factor H under complement attack?"}
        ]
    },
]
inputs = tokenizer.apply_chat_template(
    messages,
    add_generation_prompt=True,
    tokenize=True,
    return_dict=True,
    return_tensors="pt",
).to(model.device)

outputs = model.generate(**inputs, max_new_tokens=1024)
print(tokenizer.decode(outputs[0][inputs["input_ids"].shape[-1]:]))

Benchmarks

Model	Quantization	PubMedQA<br>(In-Domain)	MedQA-USMLE<br>(Zero-Shot Transfer)
Gemma-4-E2B-it (base model)	-	58.10 %	29.54 %
Gemma-4-E2B-it + SFT + RLVR	-	73.10 %	43.05 %
Gemma-4-E2B-it + SFT + RLVR	Q8_0	72.40 %	43.00 %
Gemma-4-E2B-it + SFT + RLVR	Q6_K	72.10 %	42.18 %
Gemma-4-E2B-it + SFT + RLVR	Q5KM	72.00 %	38.88 %
Gemma-4-E2B-it + SFT + RLVR	Q4KM	71.80 %	38.88 %