---
license: apache-2.0
base_model: Cagatayd/llama3.2-1B-Instruct-Egitim
tags:
- llama3
- fine-tuned
- docker
- lora
datasets:
- MattCoddity/dockerNLcommands
---

# Fine-tuned Llama 3.2 1B for Docker Commands

This model is a fine-tuned version of [Cagatayd/llama3.2-1B-Instruct-Egitim](https://huggingface.co/Cagatayd/llama3.2-1B-Instruct-Egitim) on the [MattCoddity/dockerNLcommands](https://huggingface.co/datasets/MattCoddity/dockerNLcommands) dataset.

## Training Details

- **Base Model:** Cagatayd/llama3.2-1B-Instruct-Egitim
- **Dataset:** MattCoddity/dockerNLcommands
- **Training Method:** LoRA (Low-Rank Adaptation)
- **Quantization:** 8-bit
- **LoRA Rank:** 32
- **Training Steps:** 60
- **Final Accuracy:** 80.69%

## Usage

```python
from transformers import AutoModelForCausalLM, AutoTokenizer

model = AutoModelForCausalLM.from_pretrained("psssd-cat/llm-base")
tokenizer = AutoTokenizer.from_pretrained("psssd-cat/llm-base")

# Example usage
text = "Find all running docker containers"
inputs = tokenizer(text, return_tensors="pt")
outputs = model.generate(**inputs, max_new_tokens=100)
print(tokenizer.decode(outputs[0], skip_special_tokens=True))
```

## Training Configuration

- Learning Rate: 3e-5
- Batch Size: 2 per device
- Gradient Accumulation Steps: 4
- FP16: Enabled
- Warmup Steps: 5

## Model Performance

- Evaluation Loss: 1.147
- Mean Token Accuracy: 80.69%
- Training Loss: 2.094

---

Generated with [Claude Code](https://claude.com/claude-code)