Upload README.md with huggingface_hub
Browse files
README.md
ADDED
|
@@ -0,0 +1,47 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
---
|
| 2 |
+
base_model: meta-llama/Llama-2-7b-hf
|
| 3 |
+
tags:
|
| 4 |
+
- LoRA
|
| 5 |
+
- bittensor
|
| 6 |
+
- gradients
|
| 7 |
+
license: apache-2.0
|
| 8 |
+
---
|
| 9 |
+
|
| 10 |
+
# Submission for task `submission_test_speed_check`
|
| 11 |
+
|
| 12 |
+
- Task ID: `sim-task-readme-speed`
|
| 13 |
+
- Repo: `submission_test_speed_check`
|
| 14 |
+
- Loss: `4.868357499440511`
|
| 15 |
+
- Timestamp: 2025-06-30T23:56:20.301535
|
| 16 |
+
|
| 17 |
+
---
|
| 18 |
+
|
| 19 |
+
### LoRA Parameters
|
| 20 |
+
- Rank: 8
|
| 21 |
+
- Alpha: 16
|
| 22 |
+
- Dropout: 0.1
|
| 23 |
+
- Epochs: 1
|
| 24 |
+
- Learning rate: 1e-4
|
| 25 |
+
- Batch size: 1
|
| 26 |
+
|
| 27 |
+
### Example
|
| 28 |
+
**Instruction:**
|
| 29 |
+
Come si calcola la derivata di \( x^2 \)?
|
| 30 |
+
|
| 31 |
+
**Response:**
|
| 32 |
+
La derivata di \( x^2 \) rispetto a \( x \) è \( 2x \).
|
| 33 |
+
|
| 34 |
+
### Training Info
|
| 35 |
+
This LoRA adapter was trained using a dataset dynamically generated from RAG (FAISS index) based on the task theme.
|
| 36 |
+
The dataset consisted of 35 prompt-response examples related to mathematics and programming.
|
| 37 |
+
|
| 38 |
+
The adapter was trained on CPU in ~7 minutes and uploaded automatically via Hugging Face API.
|
| 39 |
+
|
| 40 |
+
---
|
| 41 |
+
|
| 42 |
+
### Training Curve
|
| 43 |
+
|
| 44 |
+
|
| 45 |
+
### License and Usage
|
| 46 |
+
This adapter inherits the license and intended usage of the base model Meta LLaMA 2 (Meta AI License),
|
| 47 |
+
and is provided here for research purposes only as part of the Bittensor Subnet 56 participation.
|