Upload training_info.json with huggingface_hub
Browse files- training_info.json +21 -0
training_info.json
ADDED
|
@@ -0,0 +1,21 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"model_name": "Math-MCQ-Generator-v1",
|
| 3 |
+
"base_model": "deepseek-ai/deepseek-math-7b-instruct",
|
| 4 |
+
"training_method": "QLoRA (4-bit quantization)",
|
| 5 |
+
"upload_date": "2025-09-03T22:27:38.364105",
|
| 6 |
+
"dataset_size": 1519,
|
| 7 |
+
"epochs": 5,
|
| 8 |
+
"final_loss": "~0.20",
|
| 9 |
+
"date": "2025-09-03",
|
| 10 |
+
"lora_config": {
|
| 11 |
+
"r": 64,
|
| 12 |
+
"lora_alpha": 128,
|
| 13 |
+
"target_modules": [
|
| 14 |
+
"q_proj",
|
| 15 |
+
"v_proj",
|
| 16 |
+
"k_proj",
|
| 17 |
+
"o_proj"
|
| 18 |
+
],
|
| 19 |
+
"lora_dropout": 0.05
|
| 20 |
+
}
|
| 21 |
+
}
|