nilnguyen2k commited on
Commit
be09629
·
verified ·
1 Parent(s): 833814f

Upload checkpoints/cnn_vit_gnn__attention/model_delta_metadata.json with huggingface_hub

Browse files
checkpoints/cnn_vit_gnn__attention/model_delta_metadata.json ADDED
@@ -0,0 +1,129 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "branches": [
3
+ "cnn",
4
+ "vit",
5
+ "gnn"
6
+ ],
7
+ "mode": "attention",
8
+ "seed": 42,
9
+ "config": {
10
+ "root_img": "/kaggle/input/datasets/nilnguyenquoc/mit67-701515/MIT67_701515",
11
+ "root_graph": "/kaggle/input/datasets/nilnguyenquoc/mit67-graphs-yolov8xworld-k3-c025-m24-vis32-iou30",
12
+ "out_dir": "/kaggle/working",
13
+ "hf_upload": true,
14
+ "hf_upload_repo": "nilnguyen2k/mit67_fusion_head",
15
+ "hf_upload_private": false,
16
+ "backbones": {
17
+ "cnn": {
18
+ "repo": "nilnguyen2k/mit67_resnet50_places365",
19
+ "file": "mit_resnet50.pth",
20
+ "model_name": "resnet50"
21
+ },
22
+ "vit": {
23
+ "repo": "nilnguyen2k/mit67_swin_tiny",
24
+ "file": "mit67_swin_tiny.pth",
25
+ "model_name": "swin_tiny_patch4_window7_224"
26
+ },
27
+ "gnn": {
28
+ "repo": "nilnguyen2k/mit67_gnn",
29
+ "file": "mit67_gnn.pth"
30
+ }
31
+ },
32
+ "num_classes": 67,
33
+ "imgsz": 224,
34
+ "extract_batch_size": 64,
35
+ "fit_split": "valid",
36
+ "use_pca": true,
37
+ "pca_dim": 64,
38
+ "pca_max_dim": 192,
39
+ "pca_whiten": true,
40
+ "pca_dim_grid": [
41
+ 32,
42
+ 64,
43
+ 128,
44
+ 192
45
+ ],
46
+ "pca_dim_ablation": [
47
+ 192
48
+ ],
49
+ "pca_fit_split": "train",
50
+ "inner_folds": 5,
51
+ "cv_repeats": 3,
52
+ "standalone_vit_acc": 87.15,
53
+ "n_aug_variants": 8,
54
+ "allow_hflip": false,
55
+ "dims": {
56
+ "cnn": 192,
57
+ "vit": 192,
58
+ "gnn": 192
59
+ },
60
+ "common_dim": 32,
61
+ "hidden_dim": 0,
62
+ "dropout_adapter": 0.4,
63
+ "dropout_head": 0.4,
64
+ "delta_rank": 12,
65
+ "delta_l2_penalty": 0.01,
66
+ "delta_weight_decay": 0.1,
67
+ "use_delta": true,
68
+ "epochs_base": 120,
69
+ "epochs_delta": 30,
70
+ "epochs": 80,
71
+ "batch_size": 64,
72
+ "lr": 0.001,
73
+ "weight_decay": 0.01,
74
+ "patience": 0,
75
+ "mixup_alpha": 0.4,
76
+ "aux_weight": 0.1,
77
+ "residual_base_branch": "vit",
78
+ "delta_scale_init": 1.0,
79
+ "gate_sup_weight": 0.3,
80
+ "label_smoothing": 0.1,
81
+ "log_every": 5,
82
+ "seeds": [
83
+ 42,
84
+ 1337,
85
+ 2024
86
+ ],
87
+ "branch_sets": [
88
+ [
89
+ "vit"
90
+ ],
91
+ [
92
+ "cnn",
93
+ "vit"
94
+ ],
95
+ [
96
+ "vit",
97
+ "gnn"
98
+ ],
99
+ [
100
+ "cnn",
101
+ "vit",
102
+ "gnn"
103
+ ]
104
+ ],
105
+ "modes": [
106
+ "concat",
107
+ "attention"
108
+ ],
109
+ "gnn_arch": {
110
+ "vocab_size": 256,
111
+ "embedding_dim": 32,
112
+ "box_features": 7,
113
+ "extra_node_features": 2,
114
+ "edge_dim": 3,
115
+ "crop_size": 32,
116
+ "visual_embed_dim": 16,
117
+ "hidden_channels": 112,
118
+ "heads": 4
119
+ }
120
+ },
121
+ "variant": "delta",
122
+ "metrics": {
123
+ "accuracy": 87.93677204658901,
124
+ "precision_macro": 88.33280695473897,
125
+ "recall_macro": 85.84124321486036,
126
+ "f1_macro": 86.6769042526632,
127
+ "f1_weighted": 87.80023336396476
128
+ }
129
+ }