---
base_model: unsloth/gemma-3-270m-it
tags:
- text-generation-inference
- transformers
- unsloth
- gemma3_text
- trl
license: apache-2.0
language:
- en
---

# Uploaded  model

- **Developed by:** Nonovogo
- **License:** apache-2.0
- **Finetuned from model :** unsloth/gemma-3-270m-it

Use
```
text = tokenizer.apply_chat_template(
    messages,
    tokenize = False,
    add_generation_prompt = True
).removeprefix('<bos>')

# This forces the model to enter "thinking mode" immediately.
text += "<think>\n"

# 3. Generate
_ = model.generate(
    **tokenizer(text, return_tensors="pt").to("cuda"),
    max_new_tokens=2048,     # Don't let it ramble forever
    
    # --- STABILITY SETTINGS ---
    do_sample=True,         # Enable sampling to break deterministic loops
    temperature=0.1,        # Very low temp (focused) but not zero
    top_p=0.95,             # Standard filtering
    repetition_penalty=1.0, # CRITICAL: Disable penalty (1.0 = no penalty)
    
    streamer=TextStreamer(tokenizer, skip_prompt=True),
    eos_token_id=tokenizer.eos_token_id # Ensure it knows when to stop
)
```
For better output 
- This gemma3_text model was trained 2x faster with [Unsloth](https://github.com/unslothai/unsloth) and Huggingface's TRL library.

[<img src="https://raw.githubusercontent.com/unslothai/unsloth/main/images/unsloth%20made%20with%20love.png" width="200"/>](https://github.com/unslothai/unsloth)