Add training results
Browse files- training_results.json +37 -0
training_results.json
ADDED
@@ -0,0 +1,37 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
{
|
2 |
+
"experiment_summary": {
|
3 |
+
"timestamp": "2025-05-26T20:36:55.416522",
|
4 |
+
"model_name": "Qwen/Qwen3-0.6B-Base",
|
5 |
+
"use_context_training": false,
|
6 |
+
"use_context_evaluation": false,
|
7 |
+
"quantization": "8bit",
|
8 |
+
"quantize_only": false,
|
9 |
+
"max_samples": null
|
10 |
+
},
|
11 |
+
"results": {
|
12 |
+
"baseline_accuracy": 0.6966019417475728,
|
13 |
+
"fine_tuned_accuracy": 0.7067961165048544,
|
14 |
+
"improvement": 0.010194174757281571,
|
15 |
+
"improvement_percentage": "+1.02%"
|
16 |
+
},
|
17 |
+
"training_config": {
|
18 |
+
"timestamp": "2025-05-26T17:27:49.001547",
|
19 |
+
"model_name": "Qwen/Qwen3-0.6B-Base",
|
20 |
+
"data_path": "data/mcqa_training_format.tsv",
|
21 |
+
"use_context": false,
|
22 |
+
"eval_with_context": false,
|
23 |
+
"epochs": 5,
|
24 |
+
"batch_size": 2,
|
25 |
+
"learning_rate": 5e-05,
|
26 |
+
"test_size": 0.1,
|
27 |
+
"max_samples": null,
|
28 |
+
"output_dir": "./qwen-mcqa-ft",
|
29 |
+
"baseline_only": false,
|
30 |
+
"eval_only": false,
|
31 |
+
"quantize_only": false,
|
32 |
+
"quantization": "8bit",
|
33 |
+
"device": "cuda",
|
34 |
+
"torch_version": "2.5.1+cu124"
|
35 |
+
},
|
36 |
+
"success": true
|
37 |
+
}
|