---
language:
- en
license: apache-2.0
tags:
- text-generation
- instruct
- manim
- lora
- gguf
datasets:
- ArunKr/verified-data-manim
base_model: HuggingFaceTB/SmolLM-135M-Instruct
library_name: transformers
pipeline_tag: text-generation
---

# gemma-3-270m-it-web-agent - Fine-tuned

This repository contains three variants of the model:

- **LoRA adapters** → [ArunKr/gemma-3-270m-it-web-agent-lora](https://huggingface.co/ArunKr/gemma-3-270m-it-web-agent-lora)
- **Merged FP16 weights** → [ArunKr/gemma-3-270m-it-web-agent-16bit](https://huggingface.co/ArunKr/gemma-3-270m-it-web-agent-16bit)
- **GGUF quantizations** → [ArunKr/gemma-3-270m-it-web-agent-gguf](https://huggingface.co/ArunKr/gemma-3-270m-it-web-agent-gguf)

### Training
- Base model: `unsloth/gemma-3-270m-it`
- Dataset: `ArunKr/gui_grounding_dataset-100`
- Method: LoRA fine-tuning with [Unsloth](https://github.com/unslothai/unsloth)

### Quantizations
We provide `f16`, `bf16`, `f32`, and `q8_0` GGUF files for llama.cpp / Ollama.

### Usage Example
```python
from transformers import AutoModelForCausalLM, AutoTokenizer

tok = AutoTokenizer.from_pretrained("ArunKr/gemma-3-270m-it-web-agent-16bit")
model = AutoModelForCausalLM.from_pretrained("ArunKr/gemma-3-270m-it-web-agent-16bit")
print(model.generate(**tok("Hello", return_tensors="pt")))
```

### Ollama Example

```bash
ollama run ArunKr/SmolLM-135M-Instruct-manim-gguf:<file_name>.gguf
```

www.ollama.com