nilnguyen2k commited on
Commit
2089fc4
·
verified ·
1 Parent(s): 68178c9

Upload checkpoints/vit_gnn_attention/model_delta_metadata.json with huggingface_hub

Browse files
checkpoints/vit_gnn_attention/model_delta_metadata.json ADDED
@@ -0,0 +1,134 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "branches": [
3
+ "vit",
4
+ "gnn"
5
+ ],
6
+ "mode": "attention",
7
+ "seed": 1337,
8
+ "config": {
9
+ "root_img": "/kaggle/input/datasets/nilnguyenquoc/mit67-701515/MIT67_701515",
10
+ "root_graph": "/kaggle/input/datasets/nilnguyenquoc/mit67-graphs-yolov8xworld-k3-c025-m24-vis32-iou30",
11
+ "out_dir": "/kaggle/working",
12
+ "hf_upload": true,
13
+ "hf_upload_repo": "nilnguyen2k/mit67_fusion_summarize",
14
+ "hf_upload_private": false,
15
+ "backbones": {
16
+ "cnn": {
17
+ "repo": "nilnguyen2k/mit67_resnet50_places365",
18
+ "file": "mit_resnet50.pth",
19
+ "model_name": "resnet50"
20
+ },
21
+ "vit": {
22
+ "repo": "nilnguyen2k/mit67_swin_small",
23
+ "file": "mit67_swin_small.pth",
24
+ "model_name": "swin_small_patch4_window7_224"
25
+ },
26
+ "gnn": {
27
+ "repo": "nilnguyen2k/mit67_gnn",
28
+ "file": "mit67_gnn.pth"
29
+ }
30
+ },
31
+ "num_classes": 67,
32
+ "imgsz": 224,
33
+ "extract_batch_size": 64,
34
+ "fit_split": "valid",
35
+ "use_pca": true,
36
+ "pca_dim": 64,
37
+ "pca_max_dim": 192,
38
+ "pca_whiten": true,
39
+ "pca_dim_grid": [
40
+ 32,
41
+ 64,
42
+ 128,
43
+ 192
44
+ ],
45
+ "pca_dim_ablation": [
46
+ 192
47
+ ],
48
+ "pca_fit_split": "train",
49
+ "inner_folds": 5,
50
+ "cv_repeats": 3,
51
+ "standalone_vit_acc": 87.15,
52
+ "n_aug_variants": 8,
53
+ "allow_hflip": false,
54
+ "dims": {
55
+ "cnn": 192,
56
+ "vit": 192,
57
+ "gnn": 192
58
+ },
59
+ "common_dim": 32,
60
+ "hidden_dim": 0,
61
+ "dropout_adapter": 0.4,
62
+ "dropout_head": 0.4,
63
+ "delta_rank": 12,
64
+ "delta_l2_penalty": 0.01,
65
+ "delta_weight_decay": 0.1,
66
+ "use_delta": true,
67
+ "epochs_base": 120,
68
+ "epochs_delta": 30,
69
+ "epochs": 80,
70
+ "batch_size": 64,
71
+ "lr": 0.001,
72
+ "weight_decay": 0.01,
73
+ "patience": 0,
74
+ "mixup_alpha": 0.4,
75
+ "aux_weight": 0.1,
76
+ "residual_base_branch": "vit",
77
+ "delta_scale_init": 1.0,
78
+ "gate_sup_weight": 0.3,
79
+ "label_smoothing": 0.1,
80
+ "log_every": 5,
81
+ "seeds": [
82
+ 42,
83
+ 1337,
84
+ 2024
85
+ ],
86
+ "branch_sets": [
87
+ [
88
+ "cnn"
89
+ ],
90
+ [
91
+ "vit"
92
+ ],
93
+ [
94
+ "gnn"
95
+ ],
96
+ [
97
+ "cnn",
98
+ "vit"
99
+ ],
100
+ [
101
+ "vit",
102
+ "gnn"
103
+ ],
104
+ [
105
+ "cnn",
106
+ "vit",
107
+ "gnn"
108
+ ]
109
+ ],
110
+ "modes": [
111
+ "concat",
112
+ "attention"
113
+ ],
114
+ "gnn_arch": {
115
+ "vocab_size": 256,
116
+ "embedding_dim": 32,
117
+ "box_features": 7,
118
+ "extra_node_features": 2,
119
+ "edge_dim": 3,
120
+ "crop_size": 32,
121
+ "visual_embed_dim": 16,
122
+ "hidden_channels": 112,
123
+ "heads": 4
124
+ }
125
+ },
126
+ "variant": "delta",
127
+ "metrics": {
128
+ "accuracy": 89.64226289517471,
129
+ "precision_macro": 89.3758688997361,
130
+ "recall_macro": 88.05600374943889,
131
+ "f1_macro": 88.4438917728843,
132
+ "f1_weighted": 89.56939598503399
133
+ }
134
+ }