README.md

---
base_model: meta-llama/Llama-2-7b-hf
tags:
- LoRA
- bittensor
- gradients
license: apache-2.0
---

# Submission for task `submission_test_speed_check`

- Task ID: `sim-task-readme-speed`
- Repo: `submission_test_speed_check`
- Loss: `4.868357499440511`
- Timestamp: 2025-06-30T23:56:20.301535

---

### LoRA Parameters
- Rank: 8
- Alpha: 16
- Dropout: 0.1
- Epochs: 1
- Learning rate: 1e-4
- Batch size: 1

### Example
**Instruction:**  
Come si calcola la derivata di \( x^2 \)?  

**Response:**  
La derivata di \( x^2 \) rispetto a \( x \) è \( 2x \).

### Training Info
This LoRA adapter was trained using a dataset dynamically generated from RAG (FAISS index) based on the task theme.  
The dataset consisted of 35 prompt-response examples related to mathematics and programming.

The adapter was trained on CPU in ~7 minutes and uploaded automatically via Hugging Face API.

---

### Training Curve
![Loss Curve](loss_curve.png)

### License and Usage
This adapter inherits the license and intended usage of the base model Meta LLaMA 2 (Meta AI License),  
and is provided here for research purposes only as part of the Bittensor Subnet 56 participation.