GaMS-Beta/gemma-3-4b-fine-tuned-slo-VLM

GEMMA-3-4b fine-tuned on Slovenian text-image data

The model is based on google/gemma-3-4b-it and was fine-tuned using SFT trainer.

How to use it:

from transformers import AutoProcessor, Gemma3ForConditionalGeneration
import torch


model_id = "GaMS-Beta/gemma-3-4b-fine-tuned-slo-VLM"
model = Gemma3ForConditionalGeneration.from_pretrained(
    model_id, device_map="auto"
).eval()

processor = AutoProcessor.from_pretrained(model_id)


messages = [
    {
        "role": "system",
        "content": [{"type": "text", "text": ""}]
    },
    {
        "role": "user",
        "content": [
           {"type": "image", "image": "https://sensa.metropolitan.si/media/cache/upload/Photo/2025/02/07/france-preseren_ofmivtT_biggalleryimage.jpg"}, 
           {"type": "text", "text": "Kdo je na sliki?"}
        ]
    }
]

print(processor.apply_chat_template(messages, tokenize=False))

inputs = processor.apply_chat_template(
    messages, add_generation_prompt=True, tokenize=True,
    return_dict=True, return_tensors="pt"
).to(model.device, dtype=torch.bfloat16)


input_len = inputs["input_ids"].shape[-1]

with torch.inference_mode():
    generation = model.generate(**inputs, max_new_tokens=500, do_sample=False)
    generation = generation[0][input_len:]

decoded = processor.decode(generation, skip_special_tokens=True)
print(decoded)