Hugging Face's logo

nilnguyen2k
/
mit67_fusion_summarize

Image Classification
PyTorch
scene-recognition
multimodal
Model card Files
xet
 Community
nilnguyen2k commited on
Commit
8a42059
·
verified ·
1 Parent(s): 1ea17d5

Upload checkpoints/cnn/model_no_delta_metadata.json with huggingface_hub

Browse files
Files changed (1) hide show
  1. checkpoints/cnn/model_no_delta_metadata.json +133 -0
checkpoints/cnn/model_no_delta_metadata.json ADDED
@@ -0,0 +1,133 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "branches": [
3
+ "cnn"
4
+ ],
5
+ "mode": null,
6
+ "seed": 42,
7
+ "config": {
8
+ "root_img": "/kaggle/input/datasets/nilnguyenquoc/mit67-701515/MIT67_701515",
9
+ "root_graph": "/kaggle/input/datasets/nilnguyenquoc/mit67-graphs-yolov8xworld-k3-c025-m24-vis32-iou30",
10
+ "out_dir": "/kaggle/working",
11
+ "hf_upload": true,
12
+ "hf_upload_repo": "nilnguyen2k/mit67_fusion_summarize",
13
+ "hf_upload_private": false,
14
+ "backbones": {
15
+ "cnn": {
16
+ "repo": "nilnguyen2k/mit67_resnet50_places365",
17
+ "file": "mit_resnet50.pth",
18
+ "model_name": "resnet50"
19
+ },
20
+ "vit": {
21
+ "repo": "nilnguyen2k/mit67_swin_small",
22
+ "file": "mit67_swin_small.pth",
23
+ "model_name": "swin_small_patch4_window7_224"
24
+ },
25
+ "gnn": {
26
+ "repo": "nilnguyen2k/mit67_gnn",
27
+ "file": "mit67_gnn.pth"
28
+ }
29
+ },
30
+ "num_classes": 67,
31
+ "imgsz": 224,
32
+ "extract_batch_size": 64,
33
+ "fit_split": "valid",
34
+ "use_pca": true,
35
+ "pca_dim": 64,
36
+ "pca_max_dim": 192,
37
+ "pca_whiten": true,
38
+ "pca_dim_grid": [
39
+ 32,
40
+ 64,
41
+ 128,
42
+ 192
43
+ ],
44
+ "pca_dim_ablation": [
45
+ 192
46
+ ],
47
+ "pca_fit_split": "train",
48
+ "inner_folds": 5,
49
+ "cv_repeats": 3,
50
+ "standalone_vit_acc": 87.15,
51
+ "n_aug_variants": 8,
52
+ "allow_hflip": false,
53
+ "dims": {
54
+ "cnn": 192,
55
+ "vit": 192,
56
+ "gnn": 192
57
+ },
58
+ "common_dim": 32,
59
+ "hidden_dim": 0,
60
+ "dropout_adapter": 0.4,
61
+ "dropout_head": 0.4,
62
+ "delta_rank": 12,
63
+ "delta_l2_penalty": 0.01,
64
+ "delta_weight_decay": 0.1,
65
+ "use_delta": true,
66
+ "epochs_base": 120,
67
+ "epochs_delta": 30,
68
+ "epochs": 80,
69
+ "batch_size": 64,
70
+ "lr": 0.001,
71
+ "weight_decay": 0.01,
72
+ "patience": 0,
73
+ "mixup_alpha": 0.4,
74
+ "aux_weight": 0.1,
75
+ "residual_base_branch": "vit",
76
+ "delta_scale_init": 1.0,
77
+ "gate_sup_weight": 0.3,
78
+ "label_smoothing": 0.1,
79
+ "log_every": 5,
80
+ "seeds": [
81
+ 42,
82
+ 1337,
83
+ 2024
84
+ ],
85
+ "branch_sets": [
86
+ [
87
+ "cnn"
88
+ ],
89
+ [
90
+ "vit"
91
+ ],
92
+ [
93
+ "gnn"
94
+ ],
95
+ [
96
+ "cnn",
97
+ "vit"
98
+ ],
99
+ [
100
+ "vit",
101
+ "gnn"
102
+ ],
103
+ [
104
+ "cnn",
105
+ "vit",
106
+ "gnn"
107
+ ]
108
+ ],
109
+ "modes": [
110
+ "concat",
111
+ "attention"
112
+ ],
113
+ "gnn_arch": {
114
+ "vocab_size": 256,
115
+ "embedding_dim": 32,
116
+ "box_features": 7,
117
+ "extra_node_features": 2,
118
+ "edge_dim": 3,
119
+ "crop_size": 32,
120
+ "visual_embed_dim": 16,
121
+ "hidden_channels": 112,
122
+ "heads": 4
123
+ }
124
+ },
125
+ "variant": "no_delta",
126
+ "metrics": {
127
+ "accuracy": 82.61231281198003,
128
+ "precision_macro": 82.3686140685749,
129
+ "recall_macro": 80.0241332593511,
130
+ "f1_macro": 80.53716083990115,
131
+ "f1_weighted": 82.35023900626118
132
+ }
133
+ }