AbstractPhil commited on
Commit
c1a9009
·
verified ·
1 Parent(s): 61f9d02

Update metrics - Run 20251013_004438

Browse files
Files changed (1) hide show
  1. best_model.json +34 -20
best_model.json CHANGED
@@ -1,27 +1,41 @@
1
  {
2
- "model_name": "David-fully_shared-weighted_sum",
3
- "run_id": "20251012_235237",
4
- "timestamp": "2025-10-13T00:38:47.151984",
5
- "best_val_acc": 66.52333333333333,
6
- "best_epoch": 9,
7
- "final_train_acc": 63.87124655359788,
8
- "final_train_loss": 3.9703641328395016,
9
  "scale_accuracies": {
10
- "256": 65.962,
11
- "512": 66.43466666666667
 
 
 
 
 
 
 
12
  },
13
  "architecture": {
14
- "preset": "small_fast",
15
- "sharing_mode": "fully_shared",
16
- "fusion_mode": "weighted_sum",
17
  "scales": [
 
18
  256,
19
- 512
 
 
 
 
 
 
20
  ],
21
  "feature_dim": 512,
22
  "num_classes": 1000,
23
- "use_belly": false,
24
- "belly_expand": 2.0
25
  },
26
  "training": {
27
  "dataset": "AbstractPhil/imagenet-clip-features-orderly",
@@ -39,10 +53,10 @@
39
  "scheduler": "cosine_restarts"
40
  },
41
  "files": {
42
- "weights_safetensors": "weights/David-fully_shared-weighted_sum/20251012_235237/best_model_acc66.52.safetensors",
43
- "weights_pytorch": "weights/David-fully_shared-weighted_sum/20251012_235237/best_model.pth",
44
- "config": "weights/David-fully_shared-weighted_sum/20251012_235237/david_config.json",
45
- "training_config": "weights/David-fully_shared-weighted_sum/20251012_235237/train_config.json",
46
- "tensorboard": "runs/David-fully_shared-weighted_sum/20251012_235237/"
47
  }
48
  }
 
1
  {
2
+ "model_name": "David-decoupled-deep_efficiency",
3
+ "run_id": "20251013_004438",
4
+ "timestamp": "2025-10-13T00:49:36.453661",
5
+ "best_val_acc": 58.398666666666664,
6
+ "best_epoch": 0,
7
+ "final_train_acc": 51.657460216609806,
8
+ "final_train_loss": 4.900288952114611,
9
  "scale_accuracies": {
10
+ "128": 58.398666666666664,
11
+ "256": 67.03133333333334,
12
+ "384": 69.55466666666666,
13
+ "448": 70.34333333333333,
14
+ "512": 70.83533333333334,
15
+ "576": 71.29266666666666,
16
+ "640": 71.60266666666666,
17
+ "768": 72.03333333333333,
18
+ "896": 72.252
19
  },
20
  "architecture": {
21
+ "preset": "gated_expert_team",
22
+ "sharing_mode": "decoupled",
23
+ "fusion_mode": "deep_efficiency",
24
  "scales": [
25
+ 128,
26
  256,
27
+ 384,
28
+ 448,
29
+ 512,
30
+ 576,
31
+ 640,
32
+ 768,
33
+ 896
34
  ],
35
  "feature_dim": 512,
36
  "num_classes": 1000,
37
+ "use_belly": true,
38
+ "belly_expand": 4
39
  },
40
  "training": {
41
  "dataset": "AbstractPhil/imagenet-clip-features-orderly",
 
53
  "scheduler": "cosine_restarts"
54
  },
55
  "files": {
56
+ "weights_safetensors": "weights/David-decoupled-deep_efficiency/20251013_004438/best_model_acc58.40.safetensors",
57
+ "weights_pytorch": "weights/David-decoupled-deep_efficiency/20251013_004438/best_model.pth",
58
+ "config": "weights/David-decoupled-deep_efficiency/20251013_004438/david_config.json",
59
+ "training_config": "weights/David-decoupled-deep_efficiency/20251013_004438/train_config.json",
60
+ "tensorboard": "runs/David-decoupled-deep_efficiency/20251013_004438/"
61
  }
62
  }