Feature Extraction
PyTorch
Bioacoustics
ilyassmoummad commited on
Commit
a42b871
1 Parent(s): 063b300

Rename cvt-13-224x224.yaml to cvt-13-224x224.json

Browse files
Files changed (2) hide show
  1. cvt-13-224x224.json +97 -0
  2. cvt-13-224x224.yaml +0 -83
cvt-13-224x224.json ADDED
@@ -0,0 +1,97 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "OUTPUT_DIR": "OUTPUT/",
3
+ "WORKERS": 6,
4
+ "PRINT_FREQ": 500,
5
+ "AMP": {
6
+ "ENABLED": true
7
+ },
8
+ "MODEL": {
9
+ "NAME": "cls_cvt",
10
+ "SPEC": {
11
+ "INIT": "trunc_norm",
12
+ "NUM_STAGES": 3,
13
+ "PATCH_SIZE": [7, 3, 3],
14
+ "PATCH_STRIDE": [4, 2, 2],
15
+ "PATCH_PADDING": [2, 1, 1],
16
+ "DIM_EMBED": [64, 192, 384],
17
+ "NUM_HEADS": [1, 3, 6],
18
+ "DEPTH": [1, 2, 10],
19
+ "MLP_RATIO": [4.0, 4.0, 4.0],
20
+ "ATTN_DROP_RATE": [0.0, 0.0, 0.0],
21
+ "DROP_RATE": [0.0, 0.0, 0.0],
22
+ "DROP_PATH_RATE": [0.0, 0.0, 0.1],
23
+ "QKV_BIAS": [true, true, true],
24
+ "CLS_TOKEN": [false, false, true],
25
+ "POS_EMBED": [false, false, false],
26
+ "QKV_PROJ_METHOD": ["dw_bn", "dw_bn", "dw_bn"],
27
+ "KERNEL_QKV": [3, 3, 3],
28
+ "PADDING_KV": [1, 1, 1],
29
+ "STRIDE_KV": [2, 2, 2],
30
+ "PADDING_Q": [1, 1, 1],
31
+ "STRIDE_Q": [1, 1, 1]
32
+ }
33
+ },
34
+ "AUG": {
35
+ "MIXUP_PROB": 1.0,
36
+ "MIXUP": 0.8,
37
+ "MIXCUT": 1.0,
38
+ "TIMM_AUG": {
39
+ "USE_LOADER": true,
40
+ "RE_COUNT": 1,
41
+ "RE_MODE": "pixel",
42
+ "RE_SPLIT": false,
43
+ "RE_PROB": 0.25,
44
+ "AUTO_AUGMENT": "rand-m9-mstd0.5-inc1",
45
+ "HFLIP": 0.5,
46
+ "VFLIP": 0.0,
47
+ "COLOR_JITTER": 0.4,
48
+ "INTERPOLATION": "bicubic"
49
+ }
50
+ },
51
+ "LOSS": {
52
+ "LABEL_SMOOTHING": 0.1
53
+ },
54
+ "CUDNN": {
55
+ "BENCHMARK": true,
56
+ "DETERMINISTIC": false,
57
+ "ENABLED": true
58
+ },
59
+ "DATASET": {
60
+ "DATASET": "imagenet",
61
+ "DATA_FORMAT": "jpg",
62
+ "ROOT": "DATASET/imagenet/",
63
+ "TEST_SET": "val",
64
+ "TRAIN_SET": "train"
65
+ },
66
+ "TEST": {
67
+ "BATCH_SIZE_PER_GPU": 32,
68
+ "IMAGE_SIZE": [224, 224],
69
+ "MODEL_FILE": "",
70
+ "INTERPOLATION": 3
71
+ },
72
+ "TRAIN": {
73
+ "BATCH_SIZE_PER_GPU": 256,
74
+ "LR": 0.00025,
75
+ "IMAGE_SIZE": [224, 224],
76
+ "BEGIN_EPOCH": 0,
77
+ "END_EPOCH": 300,
78
+ "LR_SCHEDULER": {
79
+ "METHOD": "timm",
80
+ "ARGS": {
81
+ "sched": "cosine",
82
+ "warmup_epochs": 5,
83
+ "warmup_lr": 0.000001,
84
+ "min_lr": 0.00001,
85
+ "cooldown_epochs": 10,
86
+ "decay_rate": 0.1
87
+ }
88
+ },
89
+ "OPTIMIZER": "adamW",
90
+ "WD": 0.05,
91
+ "WITHOUT_WD_LIST": ["bn", "bias", "ln"],
92
+ "SHUFFLE": true
93
+ },
94
+ "DEBUG": {
95
+ "DEBUG": false
96
+ }
97
+ }
cvt-13-224x224.yaml DELETED
@@ -1,83 +0,0 @@
1
- OUTPUT_DIR: 'OUTPUT/'
2
- WORKERS: 6
3
- PRINT_FREQ: 500
4
- AMP:
5
- ENABLED: true
6
-
7
- MODEL:
8
- NAME: cls_cvt
9
- SPEC:
10
- INIT: 'trunc_norm'
11
- NUM_STAGES: 3
12
- PATCH_SIZE: [7, 3, 3]
13
- PATCH_STRIDE: [4, 2, 2]
14
- PATCH_PADDING: [2, 1, 1]
15
- DIM_EMBED: [64, 192, 384]
16
- NUM_HEADS: [1, 3, 6]
17
- DEPTH: [1, 2, 10]
18
- MLP_RATIO: [4.0, 4.0, 4.0]
19
- ATTN_DROP_RATE: [0.0, 0.0, 0.0]
20
- DROP_RATE: [0.0, 0.0, 0.0]
21
- DROP_PATH_RATE: [0.0, 0.0, 0.1]
22
- QKV_BIAS: [True, True, True]
23
- CLS_TOKEN: [False, False, True]
24
- POS_EMBED: [False, False, False]
25
- QKV_PROJ_METHOD: ['dw_bn', 'dw_bn', 'dw_bn']
26
- KERNEL_QKV: [3, 3, 3]
27
- PADDING_KV: [1, 1, 1]
28
- STRIDE_KV: [2, 2, 2]
29
- PADDING_Q: [1, 1, 1]
30
- STRIDE_Q: [1, 1, 1]
31
- AUG:
32
- MIXUP_PROB: 1.0
33
- MIXUP: 0.8
34
- MIXCUT: 1.0
35
- TIMM_AUG:
36
- USE_LOADER: true
37
- RE_COUNT: 1
38
- RE_MODE: pixel
39
- RE_SPLIT: false
40
- RE_PROB: 0.25
41
- AUTO_AUGMENT: rand-m9-mstd0.5-inc1
42
- HFLIP: 0.5
43
- VFLIP: 0.0
44
- COLOR_JITTER: 0.4
45
- INTERPOLATION: bicubic
46
- LOSS:
47
- LABEL_SMOOTHING: 0.1
48
- CUDNN:
49
- BENCHMARK: true
50
- DETERMINISTIC: false
51
- ENABLED: true
52
- DATASET:
53
- DATASET: 'imagenet'
54
- DATA_FORMAT: 'jpg'
55
- ROOT: 'DATASET/imagenet/'
56
- TEST_SET: 'val'
57
- TRAIN_SET: 'train'
58
- TEST:
59
- BATCH_SIZE_PER_GPU: 32
60
- IMAGE_SIZE: [224, 224]
61
- MODEL_FILE: ''
62
- INTERPOLATION: 3
63
- TRAIN:
64
- BATCH_SIZE_PER_GPU: 256
65
- LR: 0.00025
66
- IMAGE_SIZE: [224, 224]
67
- BEGIN_EPOCH: 0
68
- END_EPOCH: 300
69
- LR_SCHEDULER:
70
- METHOD: 'timm'
71
- ARGS:
72
- sched: 'cosine'
73
- warmup_epochs: 5
74
- warmup_lr: 0.000001
75
- min_lr: 0.00001
76
- cooldown_epochs: 10
77
- decay_rate: 0.1
78
- OPTIMIZER: adamW
79
- WD: 0.05
80
- WITHOUT_WD_LIST: ['bn', 'bias', 'ln']
81
- SHUFFLE: true
82
- DEBUG:
83
- DEBUG: false