Hugging Face's logo

HoangTran223
/
SAM_Variants_ViT_B16_B32_TinyImageNet

Model card Files
xet
 Community
HoangTran223 commited on
Commit
33017ff
·
verified ·
1 Parent(s): de200e6

Upload folder using huggingface_hub

Browse files
This view is limited to 50 files because it contains too many changes.   See raw diff
Files changed (50) hide show
  1. TinyImageNet_ASAM_ViT_rho1_lr0.0001/data.hdf5 +0 -0
  2. TinyImageNet_ASAM_ViT_rho1_lr0.0001/final.model +3 -0
  3. TinyImageNet_ASAM_ViT_rho1_lr0.0001/params.json +97 -0
  4. TinyImageNet_ASAM_ViT_rho1_lr0.0001/terminal.log +24 -0
  5. TinyImageNet_ASAM_ViT_rho1_lr0.0001_run01/data.hdf5 +0 -0
  6. TinyImageNet_ASAM_ViT_rho1_lr0.0001_run01/final.model +3 -0
  7. TinyImageNet_ASAM_ViT_rho1_lr0.0001_run01/params.json +97 -0
  8. TinyImageNet_ASAM_ViT_rho1_lr0.0001_run01/terminal.log +24 -0
  9. TinyImageNet_ESAM_ViT_rho0.075_lr0.0001/data.hdf5 +0 -0
  10. TinyImageNet_ESAM_ViT_rho0.075_lr0.0001/final.model +3 -0
  11. TinyImageNet_ESAM_ViT_rho0.075_lr0.0001/params.json +97 -0
  12. TinyImageNet_ESAM_ViT_rho0.075_lr0.0001/terminal.log +24 -0
  13. TinyImageNet_ESAM_ViT_rho0.075_lr0.0001_run01/data.hdf5 +0 -0
  14. TinyImageNet_ESAM_ViT_rho0.075_lr0.0001_run01/final.model +3 -0
  15. TinyImageNet_ESAM_ViT_rho0.075_lr0.0001_run01/params.json +97 -0
  16. TinyImageNet_ESAM_ViT_rho0.075_lr0.0001_run01/terminal.log +24 -0
  17. TinyImageNet_FisherSAM_ViT_rho0.01_lr0.0001/data.hdf5 +0 -0
  18. TinyImageNet_FisherSAM_ViT_rho0.01_lr0.0001/final.model +3 -0
  19. TinyImageNet_FisherSAM_ViT_rho0.01_lr0.0001/params.json +97 -0
  20. TinyImageNet_FisherSAM_ViT_rho0.01_lr0.0001/terminal.log +24 -0
  21. TinyImageNet_FisherSAM_ViT_rho0.1_lr0.0001/data.hdf5 +0 -0
  22. TinyImageNet_FisherSAM_ViT_rho0.1_lr0.0001/final.model +3 -0
  23. TinyImageNet_FisherSAM_ViT_rho0.1_lr0.0001/params.json +97 -0
  24. TinyImageNet_FisherSAM_ViT_rho0.1_lr0.0001/terminal.log +24 -0
  25. TinyImageNet_FriendlySAM_ViT_rho0.01_lr0.0001/data.hdf5 +0 -0
  26. TinyImageNet_FriendlySAM_ViT_rho0.01_lr0.0001/final.model +3 -0
  27. TinyImageNet_FriendlySAM_ViT_rho0.01_lr0.0001/params.json +97 -0
  28. TinyImageNet_FriendlySAM_ViT_rho0.01_lr0.0001/terminal.log +24 -0
  29. TinyImageNet_GSAM_ViT_rho0.01_lr0.0001/data.hdf5 +0 -0
  30. TinyImageNet_GSAM_ViT_rho0.01_lr0.0001/final.model +3 -0
  31. TinyImageNet_GSAM_ViT_rho0.01_lr0.0001/params.json +97 -0
  32. TinyImageNet_GSAM_ViT_rho0.01_lr0.0001/terminal.log +24 -0
  33. TinyImageNet_MSAM_ViT_rho0.075_lr0.0001/data.hdf5 +0 -0
  34. TinyImageNet_MSAM_ViT_rho0.075_lr0.0001/final.model +3 -0
  35. TinyImageNet_MSAM_ViT_rho0.075_lr0.0001/params.json +97 -0
  36. TinyImageNet_MSAM_ViT_rho0.075_lr0.0001/terminal.log +24 -0
  37. TinyImageNet_MSAM_ViT_rho0.075_lr0.0001_run01/data.hdf5 +0 -0
  38. TinyImageNet_MSAM_ViT_rho0.075_lr0.0001_run01/final.model +3 -0
  39. TinyImageNet_MSAM_ViT_rho0.075_lr0.0001_run01/params.json +97 -0
  40. TinyImageNet_MSAM_ViT_rho0.075_lr0.0001_run01/terminal.log +24 -0
  41. TinyImageNet_MongeSAM_ViT_rho0.075_lr0.0001/data.hdf5 +0 -0
  42. TinyImageNet_MongeSAM_ViT_rho0.075_lr0.0001/final.model +3 -0
  43. TinyImageNet_MongeSAM_ViT_rho0.075_lr0.0001/params.json +97 -0
  44. TinyImageNet_MongeSAM_ViT_rho0.075_lr0.0001/terminal.log +24 -0
  45. TinyImageNet_MongeSAM_ViT_rho0.075_lr0.0001_run01/data.hdf5 +0 -0
  46. TinyImageNet_MongeSAM_ViT_rho0.075_lr0.0001_run01/final.model +3 -0
  47. TinyImageNet_MongeSAM_ViT_rho0.075_lr0.0001_run01/params.json +97 -0
  48. TinyImageNet_MongeSAM_ViT_rho0.075_lr0.0001_run01/terminal.log +24 -0
  49. TinyImageNet_SAM_ViT_rho0.01_lr0.0001/data.hdf5 +0 -0
  50. TinyImageNet_SAM_ViT_rho0.01_lr0.0001/final.model +3 -0
TinyImageNet_ASAM_ViT_rho1_lr0.0001/data.hdf5 ADDED
Binary file (56 kB). View file
 
TinyImageNet_ASAM_ViT_rho1_lr0.0001/final.model ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:43388b81453a8abaecb6e795e80d4cf573e87a09efe6b1fefcd63f7c38b8ac6e
3
+ size 1052309706
TinyImageNet_ASAM_ViT_rho1_lr0.0001/params.json ADDED
@@ -0,0 +1,97 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "widthFactor": 4,
3
+ "ViTPatchsize": 32,
4
+ "ViTSize": "B",
5
+ "pyramidAlpha": 270,
6
+ "pyramidBottleneck": true,
7
+ "model": "ViT",
8
+ "dropout": 0.0,
9
+ "BN": true,
10
+ "depth": 16,
11
+ "label_smoothing": 0.1,
12
+ "ESAMbeta": 0.5,
13
+ "ESAMgamma": 0.7,
14
+ "lookSAMk": 5,
15
+ "lookSAMalpha": 0.7,
16
+ "lookSAMLayerwise": false,
17
+ "optimizer": "ASAM",
18
+ "weightDecay": 0.001,
19
+ "momentum": 0.9,
20
+ "rho": 0.075,
21
+ "samRho": 0.075,
22
+ "asamRho": 1.0,
23
+ "esamRho": 0.075,
24
+ "looksamRho": 0.05,
25
+ "friendlysamRho": 0.075,
26
+ "fishersamRho": 0.5,
27
+ "mongesamRho": 0.075,
28
+ "gsamRho": 0.075,
29
+ "bayesiansamRho": 0.075,
30
+ "msamRho": 0.075,
31
+ "adamwSamRho": 0.075,
32
+ "adamwMsamRho": 0.075,
33
+ "adamwFriendlysamRho": 0.075,
34
+ "sigma": 1.0,
35
+ "lmbda": 0.95,
36
+ "fisherEta": 0.1,
37
+ "gsamAlpha": 0.1,
38
+ "gsamAdaptive": false,
39
+ "bayesMSharpness": 8,
40
+ "bayesDelta": 10.0,
41
+ "bayesBeta1": 0.9,
42
+ "bayesBeta2": 0.999,
43
+ "bayesSInit": 1.0,
44
+ "bayesDamping": 1e-08,
45
+ "bayesEps": 1e-12,
46
+ "nesterov": false,
47
+ "grad_clip_norm": 0.0,
48
+ "dataThreads": 4,
49
+ "dataDir": "/home/hungpv/.datasets",
50
+ "batchSize": 64,
51
+ "dataset": "TinyImageNet",
52
+ "imageSize": 64,
53
+ "flip": true,
54
+ "crop": true,
55
+ "cut": false,
56
+ "cutoutProp": 0.5,
57
+ "randAugment": true,
58
+ "randAugment_magnitude": 10,
59
+ "mixup": false,
60
+ "mixupProp": 0.2,
61
+ "normalize": "std",
62
+ "truncate": false,
63
+ "verbose": false,
64
+ "metrics": [
65
+ "learningRate",
66
+ "loss",
67
+ "accuracy"
68
+ ],
69
+ "logEach": 1,
70
+ "rndSeed": false,
71
+ "deterministic": false,
72
+ "learningRate": 0.0001,
73
+ "lrScheduler": [
74
+ "startRamp",
75
+ "cos"
76
+ ],
77
+ "LRScheduler_step_steps": 3.0,
78
+ "LRScheduler_step_gamma": 0.2,
79
+ "LRScheduler_exp_maxDecay": 0.01,
80
+ "LRScheduler_cos_periods": 0.5,
81
+ "LRScheduler_StartRamp_epochs": 1.0,
82
+ "LRScheduler_WRN_T0": 50.0,
83
+ "LRScheduler_WRN_Tmult": 1.2,
84
+ "rhoScheduler": [
85
+ "const"
86
+ ],
87
+ "rhoScheduler_StartRamp_epochs": 8.0,
88
+ "rhoScheduler_StartJump_epochs": 8.0,
89
+ "saveCheckpoint": true,
90
+ "keepLastCheckpoint": false,
91
+ "checkpointsList": [],
92
+ "saveCheckpointInterval": 0,
93
+ "logDir": "/mnt/hungpv/projects/SAM_Variant/logs",
94
+ "logSubDir": "TinyImageNet_ASAM_ViT_rho1_lr0.0001",
95
+ "epochs": 20,
96
+ "contin": false
97
+ }
TinyImageNet_ASAM_ViT_rho1_lr0.0001/terminal.log ADDED
@@ -0,0 +1,24 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ┏━━━━━━━╸S╺╸T╺╸A╺╸T╺╸S╺━━━━━━━┳━━━━━━━━━━━━━━━━━T╺╸R╺╸A╺╸I╺╸N ━━━━━━━━━━━━━━━━┳━━━━━━━━━━T╺╸E╺╸S╺╸T ━━━━━━━━━━┓
2
+ ┃ ┃ ┃ ┃
3
+ ┃ epoch │ time ┃ learningRate │ loss │ accuracy ┃ loss │ accuracy ┃
4
+ ┠──────────────┼──────────────╂───────────────┼───────────────┼───────────────╂───────────────┼───────────────┨
5
+ ┃ 1 │ 15:08 min ┃ 4.988e-05 │ 4.3452 │ 5.17% ┃ 2.3462 │ 39.43% ┃
6
+ ┃ 2 │ 05:55 min ┃ 9.857e-05 │ 1.1988 │ 69.40% ┃ 0.6596 │ 83.70% ┃
7
+ ┃ 3 │ 05:25 min ┃ 9.615e-05 │ 0.7342 │ 81.34% ┃ 0.5396 │ 86.11% ┃
8
+ ┃ 4 │ 05:25 min ┃ 9.259e-05 │ 0.6144 │ 83.96% ┃ 0.4840 │ 87.68% ┃
9
+ ┃ 5 │ 05:25 min ┃ 8.798e-05 │ 0.5418 │ 85.76% ┃ 0.4530 │ 88.05% ┃
10
+ ┃ 6 │ 05:25 min ┃ 8.244e-05 │ 0.4930 │ 86.97% ┃ 0.4349 │ 88.37% ┃
11
+ ┃ 7 │ 05:25 min ┃ 7.610e-05 │ 0.4560 │ 87.80% ┃ 0.4204 │ 88.77% ┃
12
+ ┃ 8 │ 05:25 min ┃ 6.911e-05 │ 0.4271 │ 88.56% ┃ 0.4071 │ 89.13% ┃
13
+ ┃ 9 │ 05:25 min ┃ 6.166e-05 │ 0.4063 │ 89.15% ┃ 0.3985 │ 89.26% ┃
14
+ ┃ 10 │ 05:24 min ┃ 5.392e-05 │ 0.3829 │ 89.78% ┃ 0.3921 │ 89.48% ┃
15
+ ┃ 11 │ 05:25 min ┃ 4.608e-05 │ 0.3683 │ 90.19% ┃ 0.3877 │ 89.35% ┃
16
+ ┃ 12 │ 05:25 min ┃ 3.834e-05 │ 0.3582 │ 90.48% ┃ 0.3833 │ 89.62% ┃
17
+ ┃ 13 │ 05:26 min ┃ 3.088e-05 │ 0.3458 │ 90.79% ┃ 0.3806 │ 89.71% ┃
18
+ ┃ 14 │ 05:25 min ┃ 2.390e-05 │ 0.3356 │ 91.23% ┃ 0.3787 │ 89.79% ┃
19
+ ┃ 15 │ 05:25 min ┃ 1.756e-05 │ 0.3289 │ 91.34% ┃ 0.3766 │ 89.75% ┃
20
+ ┃ 16 │ 05:25 min ┃ 1.202e-05 │ 0.3231 │ 91.57% ┃ 0.3757 │ 89.79% ┃
21
+ ┃ 17 │ 05:25 min ┃ 7.410e-06 │ 0.3209 │ 91.66% ┃ 0.3747 │ 89.85% ┃
22
+ ┃ 18 │ 05:25 min ┃ 3.853e-06 │ 0.3172 │ 91.67% ┃ 0.3743 │ 89.86% ┃
23
+ ┃ 19 │ 05:25 min ┃ 1.431e-06 │ 0.3147 │ 91.75% ┃ 0.3743 │ 89.84% ┃
24
+ ┃ 20 │ 05:25 min ┃ 2.052e-07 │ 0.3165 │ 91.70% ┃ 0.3743 │ 89.86% ┃
TinyImageNet_ASAM_ViT_rho1_lr0.0001_run01/data.hdf5 ADDED
Binary file (56 kB). View file
 
TinyImageNet_ASAM_ViT_rho1_lr0.0001_run01/final.model ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:281c4b6f2a9db3d74d8992be7ea32537f0bd908fdb00fba11bb629391c4c95b1
3
+ size 1035057354
TinyImageNet_ASAM_ViT_rho1_lr0.0001_run01/params.json ADDED
@@ -0,0 +1,97 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "widthFactor": 4,
3
+ "ViTPatchsize": 16,
4
+ "ViTSize": "B",
5
+ "pyramidAlpha": 270,
6
+ "pyramidBottleneck": true,
7
+ "model": "ViT",
8
+ "dropout": 0.0,
9
+ "BN": true,
10
+ "depth": 16,
11
+ "label_smoothing": 0.1,
12
+ "ESAMbeta": 0.5,
13
+ "ESAMgamma": 0.7,
14
+ "lookSAMk": 5,
15
+ "lookSAMalpha": 0.7,
16
+ "lookSAMLayerwise": false,
17
+ "optimizer": "ASAM",
18
+ "weightDecay": 0.001,
19
+ "momentum": 0.9,
20
+ "rho": 0.01,
21
+ "samRho": 0.01,
22
+ "asamRho": 1.0,
23
+ "esamRho": 0.075,
24
+ "looksamRho": 0.05,
25
+ "friendlysamRho": 0.075,
26
+ "fishersamRho": 0.1,
27
+ "mongesamRho": 0.075,
28
+ "gsamRho": 0.075,
29
+ "bayesiansamRho": 0.075,
30
+ "msamRho": 0.075,
31
+ "adamwSamRho": 0.075,
32
+ "adamwMsamRho": 0.075,
33
+ "adamwFriendlysamRho": 0.075,
34
+ "sigma": 1.0,
35
+ "lmbda": 0.95,
36
+ "fisherEta": 0.1,
37
+ "gsamAlpha": 0.1,
38
+ "gsamAdaptive": false,
39
+ "bayesMSharpness": 8,
40
+ "bayesDelta": 10.0,
41
+ "bayesBeta1": 0.9,
42
+ "bayesBeta2": 0.999,
43
+ "bayesSInit": 1.0,
44
+ "bayesDamping": 1e-08,
45
+ "bayesEps": 1e-12,
46
+ "nesterov": false,
47
+ "grad_clip_norm": 0.0,
48
+ "dataThreads": 4,
49
+ "dataDir": "/home/hungpv/.datasets",
50
+ "batchSize": 64,
51
+ "dataset": "TinyImageNet",
52
+ "imageSize": 64,
53
+ "flip": true,
54
+ "crop": true,
55
+ "cut": false,
56
+ "cutoutProp": 0.5,
57
+ "randAugment": true,
58
+ "randAugment_magnitude": 10,
59
+ "mixup": false,
60
+ "mixupProp": 0.2,
61
+ "normalize": "std",
62
+ "truncate": false,
63
+ "verbose": false,
64
+ "metrics": [
65
+ "learningRate",
66
+ "loss",
67
+ "accuracy"
68
+ ],
69
+ "logEach": 1,
70
+ "rndSeed": false,
71
+ "deterministic": false,
72
+ "learningRate": 0.0001,
73
+ "lrScheduler": [
74
+ "startRamp",
75
+ "cos"
76
+ ],
77
+ "LRScheduler_step_steps": 3.0,
78
+ "LRScheduler_step_gamma": 0.2,
79
+ "LRScheduler_exp_maxDecay": 0.01,
80
+ "LRScheduler_cos_periods": 0.5,
81
+ "LRScheduler_StartRamp_epochs": 1.0,
82
+ "LRScheduler_WRN_T0": 50.0,
83
+ "LRScheduler_WRN_Tmult": 1.2,
84
+ "rhoScheduler": [
85
+ "const"
86
+ ],
87
+ "rhoScheduler_StartRamp_epochs": 8.0,
88
+ "rhoScheduler_StartJump_epochs": 8.0,
89
+ "saveCheckpoint": true,
90
+ "keepLastCheckpoint": false,
91
+ "checkpointsList": [],
92
+ "saveCheckpointInterval": 0,
93
+ "logDir": "/mnt/hungpv/projects/SAM_Variant/logs",
94
+ "logSubDir": "TinyImageNet_ASAM_ViT_rho1_lr0.0001_run01",
95
+ "epochs": 20,
96
+ "contin": false
97
+ }
TinyImageNet_ASAM_ViT_rho1_lr0.0001_run01/terminal.log ADDED
@@ -0,0 +1,24 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ┏━━━━━━━╸S╺╸T╺╸A╺╸T╺╸S╺━━━━━━━┳━━━━━━━━━━━━━━━━━T╺╸R╺╸A╺╸I╺╸N ━━━━━━━━━━━━━━━━┳━━━━━━━━━━T╺╸E╺╸S╺╸T ━━━━━━━━━━┓
2
+ ┃ ┃ ┃ ┃
3
+ ┃ epoch │ time ┃ learningRate │ loss │ accuracy ┃ loss │ accuracy ┃
4
+ ┠──────────────┼──────────────╂───────────────┼───────────────┼───────────────╂───────────────┼───────────────┨
5
+ ┃ 1 │ 51:26 min ┃ 4.988e-05 │ 4.4725 │ 2.21% ┃ 3.7916 │ 8.80% ┃
6
+ ┃ 2 │ 51:12 min ┃ 9.857e-05 │ 2.2201 │ 42.02% ┃ 0.9877 │ 75.58% ┃
7
+ ┃ 3 │ 51:08 min ┃ 9.615e-05 │ 0.7781 │ 81.00% ┃ 0.5503 │ 87.21% ┃
8
+ ┃ 4 │ 51:12 min ┃ 9.259e-05 │ 0.5495 │ 86.57% ┃ 0.4607 │ 88.97% ┃
9
+ ┃ 5 │ 51:12 min ┃ 8.798e-05 │ 0.4560 │ 88.74% ┃ 0.4179 │ 89.66% ┃
10
+ ┃ 6 │ 51:12 min ┃ 8.244e-05 │ 0.3962 │ 89.98% ┃ 0.3885 │ 90.43% ┃
11
+ ┃ 7 │ 49:34 min ┃ 7.610e-05 │ 0.3544 │ 90.98% ┃ 0.3783 │ 90.34% ┃
12
+ ┃ 8 │ 51:15 min ┃ 6.911e-05 │ 0.3253 │ 91.74% ┃ 0.3635 │ 90.64% ┃
13
+ ┃ 9 │ 51:15 min ┃ 6.166e-05 │ 0.3009 │ 92.35% ┃ 0.3565 │ 90.88% ┃
14
+ ┃ 10 │ 51:15 min ┃ 5.392e-05 │ 0.2783 │ 93.08% ┃ 0.3499 │ 90.95% ┃
15
+ ┃ 11 │ 51:10 min ┃ 4.608e-05 │ 0.2615 │ 93.49% ┃ 0.3491 │ 90.77% ┃
16
+ ┃ 12 │ 51:15 min ┃ 3.834e-05 │ 0.2483 │ 93.88% ┃ 0.3416 │ 91.13% ┃
17
+ ┃ 13 │ 49:38 min ┃ 3.088e-05 │ 0.2364 │ 94.21% ┃ 0.3404 │ 91.11% ┃
18
+ ┃ 14 │ 51:10 min ┃ 2.390e-05 │ 0.2280 │ 94.49% ┃ 0.3402 │ 91.18% ┃
19
+ ┃ 15 │ 51:06 min ┃ 1.756e-05 │ 0.2224 │ 94.64% ┃ 0.3373 │ 91.30% ┃
20
+ ┃ 16 │ 51:11 min ┃ 1.202e-05 │ 0.2175 │ 94.77% ┃ 0.3357 │ 91.28% ┃
21
+ ┃ 17 │ 51:10 min ┃ 7.410e-06 │ 0.2128 │ 94.94% ┃ 0.3353 │ 91.19% ┃
22
+ ┃ 18 │ 51:06 min ┃ 3.853e-06 │ 0.2106 │ 94.99% ┃ 0.3351 │ 91.31% ┃
23
+ ┃ 19 │ 49:35 min ┃ 1.431e-06 │ 0.2105 │ 94.97% ┃ 0.3350 │ 91.28% ┃
24
+ ┃ 20 │ 51:12 min ┃ 2.052e-07 │ 0.2098 │ 95.02% ┃ 0.3350 │ 91.27% ┃
TinyImageNet_ESAM_ViT_rho0.075_lr0.0001/data.hdf5 ADDED
Binary file (56 kB). View file
 
TinyImageNet_ESAM_ViT_rho0.075_lr0.0001/final.model ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:0330934be4e4769f381f039cf54f331344e1c89fd68dc06326333495a46a5de5
3
+ size 1052309770
TinyImageNet_ESAM_ViT_rho0.075_lr0.0001/params.json ADDED
@@ -0,0 +1,97 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "widthFactor": 4,
3
+ "ViTPatchsize": 32,
4
+ "ViTSize": "B",
5
+ "pyramidAlpha": 270,
6
+ "pyramidBottleneck": true,
7
+ "model": "ViT",
8
+ "dropout": 0.0,
9
+ "BN": true,
10
+ "depth": 16,
11
+ "label_smoothing": 0.1,
12
+ "ESAMbeta": 0.5,
13
+ "ESAMgamma": 0.7,
14
+ "lookSAMk": 5,
15
+ "lookSAMalpha": 0.7,
16
+ "lookSAMLayerwise": false,
17
+ "optimizer": "ESAM",
18
+ "weightDecay": 0.001,
19
+ "momentum": 0.9,
20
+ "rho": 0.075,
21
+ "samRho": 0.075,
22
+ "asamRho": 1.0,
23
+ "esamRho": 0.075,
24
+ "looksamRho": 0.05,
25
+ "friendlysamRho": 0.075,
26
+ "fishersamRho": 0.5,
27
+ "mongesamRho": 0.075,
28
+ "gsamRho": 0.075,
29
+ "bayesiansamRho": 0.075,
30
+ "msamRho": 0.075,
31
+ "adamwSamRho": 0.075,
32
+ "adamwMsamRho": 0.075,
33
+ "adamwFriendlysamRho": 0.075,
34
+ "sigma": 1.0,
35
+ "lmbda": 0.95,
36
+ "fisherEta": 0.1,
37
+ "gsamAlpha": 0.1,
38
+ "gsamAdaptive": false,
39
+ "bayesMSharpness": 8,
40
+ "bayesDelta": 10.0,
41
+ "bayesBeta1": 0.9,
42
+ "bayesBeta2": 0.999,
43
+ "bayesSInit": 1.0,
44
+ "bayesDamping": 1e-08,
45
+ "bayesEps": 1e-12,
46
+ "nesterov": false,
47
+ "grad_clip_norm": 0.0,
48
+ "dataThreads": 4,
49
+ "dataDir": "/home/hungpv/.datasets",
50
+ "batchSize": 64,
51
+ "dataset": "TinyImageNet",
52
+ "imageSize": 64,
53
+ "flip": true,
54
+ "crop": true,
55
+ "cut": false,
56
+ "cutoutProp": 0.5,
57
+ "randAugment": true,
58
+ "randAugment_magnitude": 10,
59
+ "mixup": false,
60
+ "mixupProp": 0.2,
61
+ "normalize": "std",
62
+ "truncate": false,
63
+ "verbose": false,
64
+ "metrics": [
65
+ "learningRate",
66
+ "loss",
67
+ "accuracy"
68
+ ],
69
+ "logEach": 1,
70
+ "rndSeed": false,
71
+ "deterministic": false,
72
+ "learningRate": 0.0001,
73
+ "lrScheduler": [
74
+ "startRamp",
75
+ "cos"
76
+ ],
77
+ "LRScheduler_step_steps": 3.0,
78
+ "LRScheduler_step_gamma": 0.2,
79
+ "LRScheduler_exp_maxDecay": 0.01,
80
+ "LRScheduler_cos_periods": 0.5,
81
+ "LRScheduler_StartRamp_epochs": 1.0,
82
+ "LRScheduler_WRN_T0": 50.0,
83
+ "LRScheduler_WRN_Tmult": 1.2,
84
+ "rhoScheduler": [
85
+ "const"
86
+ ],
87
+ "rhoScheduler_StartRamp_epochs": 8.0,
88
+ "rhoScheduler_StartJump_epochs": 8.0,
89
+ "saveCheckpoint": true,
90
+ "keepLastCheckpoint": false,
91
+ "checkpointsList": [],
92
+ "saveCheckpointInterval": 0,
93
+ "logDir": "/mnt/hungpv/projects/SAM_Variant/logs",
94
+ "logSubDir": "TinyImageNet_ESAM_ViT_rho0.075_lr0.0001",
95
+ "epochs": 20,
96
+ "contin": false
97
+ }
TinyImageNet_ESAM_ViT_rho0.075_lr0.0001/terminal.log ADDED
@@ -0,0 +1,24 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ┏━━━━━━━╸S╺╸T╺╸A╺╸T╺╸S╺━━━━━━━┳━━━━━━━━━━━━━━━━━T╺╸R╺╸A╺╸I╺╸N ━━━━━━━━━━━━━━━━┳━━━━━━━━━━T╺╸E╺╸S╺╸T ━━━━━━━━━━┓
2
+ ┃ ┃ ┃ ┃
3
+ ┃ epoch │ time ┃ learningRate │ loss │ accuracy ┃ loss │ accuracy ┃
4
+ ┠──────────────┼──────────────╂───────────────┼───────────────┼───────────────╂───────────────┼───────────────┨
5
+ ┃ 1 │ 07:13 min ┃ 4.988e-05 │ 4.6157 │ 0.95% ┃ 4.3750 │ 1.83% ┃
6
+ ┃ 2 │ 05:29 min ┃ 9.857e-05 │ 4.1348 │ 5.88% ┃ 3.0892 │ 25.36% ┃
7
+ ┃ 3 │ 05:29 min ┃ 9.615e-05 │ 1.7850 │ 56.97% ┃ 0.8597 │ 79.93% ┃
8
+ ┃ 4 │ 05:29 min ┃ 9.259e-05 │ 0.9346 │ 77.68% ┃ 0.6752 │ 84.21% ┃
9
+ ┃ 5 │ 05:29 min ┃ 8.798e-05 │ 0.7858 │ 81.36% ┃ 0.6099 │ 85.58% ┃
10
+ ┃ 6 │ 05:29 min ┃ 8.244e-05 │ 0.7125 │ 83.05% ┃ 0.5693 │ 86.36% ┃
11
+ ┃ 7 │ 05:29 min ┃ 7.610e-05 │ 0.6619 │ 84.13% ┃ 0.5423 │ 86.92% ┃
12
+ ┃ 8 │ 05:29 min ┃ 6.911e-05 │ 0.6264 │ 84.97% ┃ 0.5236 │ 87.15% ┃
13
+ ┃ 9 │ 05:29 min ┃ 6.166e-05 │ 0.5986 │ 85.68% ┃ 0.5094 │ 87.71% ┃
14
+ ┃ 10 │ 05:29 min ┃ 5.392e-05 │ 0.5746 │ 86.34% ┃ 0.4975 │ 87.93% ┃
15
+ ┃ 11 │ 05:29 min ┃ 4.608e-05 │ 0.5602 │ 86.65% ┃ 0.4906 │ 88.11% ┃
16
+ ┃ 12 │ 05:29 min ┃ 3.834e-05 │ 0.5490 │ 87.02% ┃ 0.4838 │ 88.09% ┃
17
+ ┃ 13 │ 05:29 min ┃ 3.088e-05 │ 0.5370 │ 87.22% ┃ 0.4793 │ 88.47% ┃
18
+ ┃ 14 │ 05:29 min ┃ 2.390e-05 │ 0.5276 │ 87.46% ┃ 0.4751 │ 88.50% ┃
19
+ ┃ 15 │ 05:29 min ┃ 1.756e-05 │ 0.5229 │ 87.61% ┃ 0.4721 │ 88.49% ┃
20
+ ┃ 16 │ 05:29 min ┃ 1.202e-05 │ 0.5173 │ 87.77% ┃ 0.4703 │ 88.62% ┃
21
+ ┃ 17 │ 05:29 min ┃ 7.410e-06 │ 0.5143 │ 87.85% ┃ 0.4691 │ 88.63% ┃
22
+ ┃ 18 │ 05:29 min ┃ 3.853e-06 │ 0.5099 │ 87.98% ┃ 0.4684 │ 88.61% ┃
23
+ ┃ 19 │ 05:29 min ┃ 1.431e-06 │ 0.5082 │ 87.99% ┃ 0.4680 │ 88.58% ┃
24
+ ┃ 20 │ 05:29 min ┃ 2.052e-07 │ 0.5101 │ 87.88% ┃ 0.4680 │ 88.58% ┃
TinyImageNet_ESAM_ViT_rho0.075_lr0.0001_run01/data.hdf5 ADDED
Binary file (56 kB). View file
 
TinyImageNet_ESAM_ViT_rho0.075_lr0.0001_run01/final.model ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:b8e598e3248fe2f779bf5fff0505f60dee159dc1433b3a569190e160a47d128b
3
+ size 1035057418
TinyImageNet_ESAM_ViT_rho0.075_lr0.0001_run01/params.json ADDED
@@ -0,0 +1,97 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "widthFactor": 4,
3
+ "ViTPatchsize": 16,
4
+ "ViTSize": "B",
5
+ "pyramidAlpha": 270,
6
+ "pyramidBottleneck": true,
7
+ "model": "ViT",
8
+ "dropout": 0.0,
9
+ "BN": true,
10
+ "depth": 16,
11
+ "label_smoothing": 0.1,
12
+ "ESAMbeta": 0.5,
13
+ "ESAMgamma": 0.7,
14
+ "lookSAMk": 5,
15
+ "lookSAMalpha": 0.7,
16
+ "lookSAMLayerwise": false,
17
+ "optimizer": "ESAM",
18
+ "weightDecay": 0.001,
19
+ "momentum": 0.9,
20
+ "rho": 0.01,
21
+ "samRho": 0.01,
22
+ "asamRho": 1.0,
23
+ "esamRho": 0.075,
24
+ "looksamRho": 0.05,
25
+ "friendlysamRho": 0.075,
26
+ "fishersamRho": 0.1,
27
+ "mongesamRho": 0.075,
28
+ "gsamRho": 0.075,
29
+ "bayesiansamRho": 0.075,
30
+ "msamRho": 0.075,
31
+ "adamwSamRho": 0.075,
32
+ "adamwMsamRho": 0.075,
33
+ "adamwFriendlysamRho": 0.075,
34
+ "sigma": 1.0,
35
+ "lmbda": 0.95,
36
+ "fisherEta": 0.1,
37
+ "gsamAlpha": 0.1,
38
+ "gsamAdaptive": false,
39
+ "bayesMSharpness": 8,
40
+ "bayesDelta": 10.0,
41
+ "bayesBeta1": 0.9,
42
+ "bayesBeta2": 0.999,
43
+ "bayesSInit": 1.0,
44
+ "bayesDamping": 1e-08,
45
+ "bayesEps": 1e-12,
46
+ "nesterov": false,
47
+ "grad_clip_norm": 0.0,
48
+ "dataThreads": 4,
49
+ "dataDir": "/home/hungpv/.datasets",
50
+ "batchSize": 64,
51
+ "dataset": "TinyImageNet",
52
+ "imageSize": 64,
53
+ "flip": true,
54
+ "crop": true,
55
+ "cut": false,
56
+ "cutoutProp": 0.5,
57
+ "randAugment": true,
58
+ "randAugment_magnitude": 10,
59
+ "mixup": false,
60
+ "mixupProp": 0.2,
61
+ "normalize": "std",
62
+ "truncate": false,
63
+ "verbose": false,
64
+ "metrics": [
65
+ "learningRate",
66
+ "loss",
67
+ "accuracy"
68
+ ],
69
+ "logEach": 1,
70
+ "rndSeed": false,
71
+ "deterministic": false,
72
+ "learningRate": 0.0001,
73
+ "lrScheduler": [
74
+ "startRamp",
75
+ "cos"
76
+ ],
77
+ "LRScheduler_step_steps": 3.0,
78
+ "LRScheduler_step_gamma": 0.2,
79
+ "LRScheduler_exp_maxDecay": 0.01,
80
+ "LRScheduler_cos_periods": 0.5,
81
+ "LRScheduler_StartRamp_epochs": 1.0,
82
+ "LRScheduler_WRN_T0": 50.0,
83
+ "LRScheduler_WRN_Tmult": 1.2,
84
+ "rhoScheduler": [
85
+ "const"
86
+ ],
87
+ "rhoScheduler_StartRamp_epochs": 8.0,
88
+ "rhoScheduler_StartJump_epochs": 8.0,
89
+ "saveCheckpoint": true,
90
+ "keepLastCheckpoint": false,
91
+ "checkpointsList": [],
92
+ "saveCheckpointInterval": 0,
93
+ "logDir": "/mnt/hungpv/projects/SAM_Variant/logs",
94
+ "logSubDir": "TinyImageNet_ESAM_ViT_rho0.075_lr0.0001_run01",
95
+ "epochs": 20,
96
+ "contin": false
97
+ }
TinyImageNet_ESAM_ViT_rho0.075_lr0.0001_run01/terminal.log ADDED
@@ -0,0 +1,24 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ┏━━━━━━━╸S╺╸T╺╸A╺╸T╺╸S╺━━━━━━━┳━━━━━━━━━━━━━━━━━T╺╸R╺╸A╺╸I╺╸N ━━━━━━━━━━━━━━━━┳━━━━━━━━━━T╺╸E╺╸S╺╸T ━━━━━━━━━━┓
2
+ ┃ ┃ ┃ ┃
3
+ ┃ epoch │ time ┃ learningRate │ loss │ accuracy ┃ loss │ accuracy ┃
4
+ ┠──────────────┼──────────────╂───────────────┼───────────────┼───────────────╂───────────────┼───────────────┨
5
+ ┃ 1 │ 53:44 min ┃ 4.988e-05 │ 4.5522 │ 1.14% ┃ 4.3090 │ 2.98% ┃
6
+ ┃ 2 │ 51:58 min ┃ 9.857e-05 │ 2.9266 │ 28.66% ┃ 0.8639 │ 79.41% ┃
7
+ ┃ 3 │ 51:56 min ┃ 9.615e-05 │ 0.6978 │ 83.23% ┃ 0.4965 │ 88.62% ┃
8
+ ┃ 4 │ 51:57 min ┃ 9.259e-05 │ 0.5044 │ 87.68% ┃ 0.4274 │ 89.98% ┃
9
+ ┃ 5 │ 51:54 min ┃ 8.798e-05 │ 0.4224 │ 89.60% ┃ 0.3921 │ 90.48% ┃
10
+ ┃ 6 │ 50:23 min ┃ 8.244e-05 │ 0.3705 │ 90.82% ┃ 0.3697 │ 90.86% ┃
11
+ ┃ 7 │ 51:53 min ┃ 7.610e-05 │ 0.3315 │ 91.92% ┃ 0.3562 │ 91.17% ┃
12
+ ┃ 8 │ 48:21 min ┃ 6.911e-05 │ 0.3052 │ 92.60% ┃ 0.3452 │ 91.23% ┃
13
+ ┃ 9 │ 52:02 min ┃ 6.166e-05 │ 0.2831 │ 93.18% ┃ 0.3395 │ 91.34% ┃
14
+ ┃ 10 │ 52:02 min ┃ 5.392e-05 │ 0.2629 │ 93.86% ┃ 0.3328 │ 91.53% ┃
15
+ ┃ 11 │ 51:57 min ┃ 4.608e-05 │ 0.2480 │ 94.29% ┃ 0.3309 │ 91.38% ┃
16
+ ┃ 12 │ 52:01 min ┃ 3.834e-05 │ 0.2357 │ 94.64% ┃ 0.3245 │ 91.65% ┃
17
+ ┃ 13 │ 52:02 min ┃ 3.088e-05 │ 0.2265 │ 94.97% ┃ 0.3227 │ 91.58% ┃
18
+ ┃ 14 │ 50:19 min ┃ 2.390e-05 │ 0.2179 │ 95.21% ┃ 0.3219 │ 91.51% ┃
19
+ ┃ 15 │ 51:56 min ┃ 1.756e-05 │ 0.2133 │ 95.31% ┃ 0.3197 │ 91.65% ┃
20
+ ┃ 16 │ 51:56 min ┃ 1.202e-05 │ 0.2093 │ 95.41% ┃ 0.3181 │ 91.61% ┃
21
+ ┃ 17 │ 51:51 min ┃ 7.410e-06 │ 0.2053 │ 95.61% ┃ 0.3178 │ 91.72% ┃
22
+ ┃ 18 │ 51:56 min ┃ 3.853e-06 │ 0.2033 │ 95.64% ┃ 0.3174 │ 91.70% ┃
23
+ ┃ 19 │ 51:56 min ┃ 1.431e-06 │ 0.2027 │ 95.65% ┃ 0.3172 │ 91.68% ┃
24
+ ┃ 20 │ 50:20 min ┃ 2.052e-07 │ 0.2027 │ 95.69% ┃ 0.3172 │ 91.68% ┃
TinyImageNet_FisherSAM_ViT_rho0.01_lr0.0001/data.hdf5 ADDED
Binary file (56 kB). View file
 
TinyImageNet_FisherSAM_ViT_rho0.01_lr0.0001/final.model ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:9e6220c14d5cc65441d221b0e5bf9eb8d79d68282db3cdbe1e9cda7bd67eb3db
3
+ size 1052309770
TinyImageNet_FisherSAM_ViT_rho0.01_lr0.0001/params.json ADDED
@@ -0,0 +1,97 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "widthFactor": 4,
3
+ "ViTPatchsize": 32,
4
+ "ViTSize": "B",
5
+ "pyramidAlpha": 270,
6
+ "pyramidBottleneck": true,
7
+ "model": "ViT",
8
+ "dropout": 0.0,
9
+ "BN": true,
10
+ "depth": 16,
11
+ "label_smoothing": 0.1,
12
+ "ESAMbeta": 0.5,
13
+ "ESAMgamma": 0.7,
14
+ "lookSAMk": 5,
15
+ "lookSAMalpha": 0.7,
16
+ "lookSAMLayerwise": false,
17
+ "optimizer": "FisherSAM",
18
+ "weightDecay": 0.001,
19
+ "momentum": 0.9,
20
+ "rho": 0.01,
21
+ "samRho": 0.01,
22
+ "asamRho": 1.0,
23
+ "esamRho": 0.075,
24
+ "looksamRho": 0.05,
25
+ "friendlysamRho": 0.01,
26
+ "fishersamRho": 0.01,
27
+ "mongesamRho": 0.075,
28
+ "gsamRho": 0.01,
29
+ "bayesiansamRho": 0.075,
30
+ "msamRho": 0.075,
31
+ "adamwSamRho": 0.075,
32
+ "adamwMsamRho": 0.075,
33
+ "adamwFriendlysamRho": 0.075,
34
+ "sigma": 1.0,
35
+ "lmbda": 0.95,
36
+ "fisherEta": 0.1,
37
+ "gsamAlpha": 0.1,
38
+ "gsamAdaptive": false,
39
+ "bayesMSharpness": 8,
40
+ "bayesDelta": 10.0,
41
+ "bayesBeta1": 0.9,
42
+ "bayesBeta2": 0.999,
43
+ "bayesSInit": 1.0,
44
+ "bayesDamping": 1e-08,
45
+ "bayesEps": 1e-12,
46
+ "nesterov": false,
47
+ "grad_clip_norm": 0.0,
48
+ "dataThreads": 4,
49
+ "dataDir": "/home/hungpv/.datasets",
50
+ "batchSize": 64,
51
+ "dataset": "TinyImageNet",
52
+ "imageSize": 64,
53
+ "flip": true,
54
+ "crop": true,
55
+ "cut": false,
56
+ "cutoutProp": 0.5,
57
+ "randAugment": true,
58
+ "randAugment_magnitude": 10,
59
+ "mixup": false,
60
+ "mixupProp": 0.2,
61
+ "normalize": "std",
62
+ "truncate": false,
63
+ "verbose": false,
64
+ "metrics": [
65
+ "learningRate",
66
+ "loss",
67
+ "accuracy"
68
+ ],
69
+ "logEach": 1,
70
+ "rndSeed": false,
71
+ "deterministic": false,
72
+ "learningRate": 0.0001,
73
+ "lrScheduler": [
74
+ "startRamp",
75
+ "cos"
76
+ ],
77
+ "LRScheduler_step_steps": 3.0,
78
+ "LRScheduler_step_gamma": 0.2,
79
+ "LRScheduler_exp_maxDecay": 0.01,
80
+ "LRScheduler_cos_periods": 0.5,
81
+ "LRScheduler_StartRamp_epochs": 1.0,
82
+ "LRScheduler_WRN_T0": 50.0,
83
+ "LRScheduler_WRN_Tmult": 1.2,
84
+ "rhoScheduler": [
85
+ "const"
86
+ ],
87
+ "rhoScheduler_StartRamp_epochs": 8.0,
88
+ "rhoScheduler_StartJump_epochs": 8.0,
89
+ "saveCheckpoint": true,
90
+ "keepLastCheckpoint": false,
91
+ "checkpointsList": [],
92
+ "saveCheckpointInterval": 0,
93
+ "logDir": "/mnt/hungpv/projects/SAM_Variant/logs",
94
+ "logSubDir": "TinyImageNet_FisherSAM_ViT_rho0.01_lr0.0001",
95
+ "epochs": 20,
96
+ "contin": false
97
+ }
TinyImageNet_FisherSAM_ViT_rho0.01_lr0.0001/terminal.log ADDED
@@ -0,0 +1,24 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ┏━━━━━━━╸S╺╸T╺╸A╺╸T╺╸S╺━━━━━━━┳━━━━━━━━━━━━━━━━━T╺╸R╺╸A╺╸I╺╸N ━━━━━━━━━━━━━━━━┳━━━━━━━━━━T╺╸E╺╸S╺╸T ━━━━━━━━━━┓
2
+ ┃ ┃ ┃ ┃
3
+ ┃ epoch │ time ┃ learningRate │ loss │ accuracy ┃ loss │ accuracy ┃
4
+ ┠──────────────┼──────────────╂───────────────┼───────────────┼───────────────╂───────────────┼───────────────┨
5
+ ┃ 1 │ 14:09 min ┃ 4.988e-05 │ 4.4779 │ 2.81% ┃ 3.1837 │ 20.95% ┃
6
+ ┃ 2 │ 12:56 min ┃ 9.857e-05 │ 1.3964 │ 64.50% ┃ 0.6817 │ 83.22% ┃
7
+ ┃ 3 │ 12:54 min ┃ 9.615e-05 │ 0.7532 │ 80.85% ┃ 0.5447 │ 86.20% ┃
8
+ ┃ 4 │ 12:49 min ┃ 9.259e-05 │ 0.6257 │ 83.73% ┃ 0.4867 │ 87.41% ┃
9
+ ┃ 5 │ 12:56 min ┃ 8.798e-05 │ 0.5528 │ 85.56% ┃ 0.4560 │ 88.12% ┃
10
+ ┃ 6 │ 12:56 min ┃ 8.244e-05 │ 0.5053 │ 86.71% ┃ 0.4357 │ 88.30% ┃
11
+ ┃ 7 │ 12:55 min ┃ 7.610e-05 │ 0.4692 │ 87.55% ┃ 0.4223 │ 88.93% ┃
12
+ ┃ 8 │ 12:50 min ┃ 6.911e-05 │ 0.4416 │ 88.33% ┃ 0.4103 │ 88.99% ┃
13
+ ┃ 9 │ 12:54 min ┃ 6.166e-05 │ 0.4218 │ 88.86% ┃ 0.4009 │ 89.21% ┃
14
+ ┃ 10 │ 12:55 min ┃ 5.392e-05 │ 0.3999 │ 89.48% ┃ 0.3937 │ 89.42% ┃
15
+ ┃ 11 │ 12:54 min ┃ 4.608e-05 │ 0.3863 │ 89.85% ┃ 0.3891 │ 89.46% ┃
16
+ ┃ 12 │ 12:53 min ┃ 3.834e-05 │ 0.3769 │ 90.14% ┃ 0.3843 │ 89.66% ┃
17
+ ┃ 13 │ 12:56 min ┃ 3.088e-05 │ 0.3656 │ 90.46% ┃ 0.3819 │ 89.68% ┃
18
+ ┃ 14 │ 12:56 min ┃ 2.390e-05 │ 0.3563 │ 90.80% ┃ 0.3795 │ 89.76% ┃
19
+ ┃ 15 │ 12:49 min ┃ 1.756e-05 │ 0.3502 │ 90.87% ┃ 0.3772 │ 89.72% ┃
20
+ ┃ 16 │ 12:55 min ┃ 1.202e-05 │ 0.3446 │ 91.11% ┃ 0.3765 │ 89.76% ┃
21
+ ┃ 17 │ 12:54 min ┃ 7.410e-06 │ 0.3426 │ 91.18% ┃ 0.3755 │ 89.84% ┃
22
+ ┃ 18 │ 12:54 min ┃ 3.853e-06 │ 0.3392 │ 91.26% ┃ 0.3753 │ 89.88% ┃
23
+ ┃ 19 │ 12:49 min ┃ 1.431e-06 │ 0.3372 │ 91.28% ┃ 0.3752 │ 89.81% ┃
24
+ ┃ 20 │ 12:56 min ┃ 2.052e-07 │ 0.3387 │ 91.26% ┃ 0.3752 │ 89.82% ┃
TinyImageNet_FisherSAM_ViT_rho0.1_lr0.0001/data.hdf5 ADDED
Binary file (56 kB). View file
 
TinyImageNet_FisherSAM_ViT_rho0.1_lr0.0001/final.model ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:cd56d668ce71acd516bb6e51596c8baf4400ad20ecfeabe60485c0a8012de4c3
3
+ size 1035057418
TinyImageNet_FisherSAM_ViT_rho0.1_lr0.0001/params.json ADDED
@@ -0,0 +1,97 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "widthFactor": 4,
3
+ "ViTPatchsize": 16,
4
+ "ViTSize": "B",
5
+ "pyramidAlpha": 270,
6
+ "pyramidBottleneck": true,
7
+ "model": "ViT",
8
+ "dropout": 0.0,
9
+ "BN": true,
10
+ "depth": 16,
11
+ "label_smoothing": 0.1,
12
+ "ESAMbeta": 0.5,
13
+ "ESAMgamma": 0.7,
14
+ "lookSAMk": 5,
15
+ "lookSAMalpha": 0.7,
16
+ "lookSAMLayerwise": false,
17
+ "optimizer": "FisherSAM",
18
+ "weightDecay": 0.001,
19
+ "momentum": 0.9,
20
+ "rho": 0.01,
21
+ "samRho": 0.01,
22
+ "asamRho": 1.0,
23
+ "esamRho": 0.075,
24
+ "looksamRho": 0.05,
25
+ "friendlysamRho": 0.075,
26
+ "fishersamRho": 0.1,
27
+ "mongesamRho": 0.075,
28
+ "gsamRho": 0.075,
29
+ "bayesiansamRho": 0.075,
30
+ "msamRho": 0.075,
31
+ "adamwSamRho": 0.075,
32
+ "adamwMsamRho": 0.075,
33
+ "adamwFriendlysamRho": 0.075,
34
+ "sigma": 1.0,
35
+ "lmbda": 0.95,
36
+ "fisherEta": 0.1,
37
+ "gsamAlpha": 0.1,
38
+ "gsamAdaptive": false,
39
+ "bayesMSharpness": 8,
40
+ "bayesDelta": 10.0,
41
+ "bayesBeta1": 0.9,
42
+ "bayesBeta2": 0.999,
43
+ "bayesSInit": 1.0,
44
+ "bayesDamping": 1e-08,
45
+ "bayesEps": 1e-12,
46
+ "nesterov": false,
47
+ "grad_clip_norm": 0.0,
48
+ "dataThreads": 4,
49
+ "dataDir": "/home/hungpv/.datasets",
50
+ "batchSize": 64,
51
+ "dataset": "TinyImageNet",
52
+ "imageSize": 64,
53
+ "flip": true,
54
+ "crop": true,
55
+ "cut": false,
56
+ "cutoutProp": 0.5,
57
+ "randAugment": true,
58
+ "randAugment_magnitude": 10,
59
+ "mixup": false,
60
+ "mixupProp": 0.2,
61
+ "normalize": "std",
62
+ "truncate": false,
63
+ "verbose": false,
64
+ "metrics": [
65
+ "learningRate",
66
+ "loss",
67
+ "accuracy"
68
+ ],
69
+ "logEach": 1,
70
+ "rndSeed": false,
71
+ "deterministic": false,
72
+ "learningRate": 0.0001,
73
+ "lrScheduler": [
74
+ "startRamp",
75
+ "cos"
76
+ ],
77
+ "LRScheduler_step_steps": 3.0,
78
+ "LRScheduler_step_gamma": 0.2,
79
+ "LRScheduler_exp_maxDecay": 0.01,
80
+ "LRScheduler_cos_periods": 0.5,
81
+ "LRScheduler_StartRamp_epochs": 1.0,
82
+ "LRScheduler_WRN_T0": 50.0,
83
+ "LRScheduler_WRN_Tmult": 1.2,
84
+ "rhoScheduler": [
85
+ "const"
86
+ ],
87
+ "rhoScheduler_StartRamp_epochs": 8.0,
88
+ "rhoScheduler_StartJump_epochs": 8.0,
89
+ "saveCheckpoint": true,
90
+ "keepLastCheckpoint": false,
91
+ "checkpointsList": [],
92
+ "saveCheckpointInterval": 0,
93
+ "logDir": "/mnt/hungpv/projects/SAM_Variant/logs",
94
+ "logSubDir": "TinyImageNet_FisherSAM_ViT_rho0.1_lr0.0001",
95
+ "epochs": 20,
96
+ "contin": false
97
+ }
TinyImageNet_FisherSAM_ViT_rho0.1_lr0.0001/terminal.log ADDED
@@ -0,0 +1,24 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ┏━━━━━━━╸S╺╸T╺╸A╺╸T╺╸S╺━━━━━━━┳━━━━━━━━━━━━━━━━━T╺╸R╺╸A╺╸I╺╸N ━━━━━━━━━━━━━━━━┳━━━━━━━━━━T╺╸E╺╸S╺╸T ━━━━━━━━━━┓
2
+ ┃ ┃ ┃ ┃
3
+ ┃ epoch │ time ┃ learningRate │ loss │ accuracy ┃ loss │ accuracy ┃
4
+ ┠──────────────┼──────────────╂───────────────┼───────────────┼───────────────╂───────────────┼───────────────┨
5
+ ┃ 1 │ 53:23 min ┃ 4.988e-05 │ 4.5657 │ 1.25% ┃ 4.2563 │ 3.66% ┃
6
+ ┃ 2 │ 52:14 min ┃ 9.857e-05 │ 2.6575 │ 34.31% ┃ 0.8503 │ 79.59% ┃
7
+ ┃ 3 │ 52:08 min ┃ 9.615e-05 │ 0.7120 │ 82.78% ┃ 0.5081 │ 88.13% ┃
8
+ ┃ 4 │ 52:13 min ┃ 9.259e-05 │ 0.5254 │ 87.04% ┃ 0.4358 │ 89.64% ┃
9
+ ┃ 5 │ 52:12 min ┃ 8.798e-05 │ 0.4458 │ 88.96% ┃ 0.3995 │ 90.05% ┃
10
+ ┃ 6 │ 50:22 min ┃ 8.244e-05 │ 0.3959 │ 90.06% ┃ 0.3753 │ 90.60% ┃
11
+ ┃ 7 │ 52:11 min ┃ 7.610e-05 │ 0.3591 │ 91.03% ┃ 0.3616 │ 90.93% ┃
12
+ ┃ 8 │ 52:14 min ┃ 6.911e-05 │ 0.3339 │ 91.67% ┃ 0.3500 │ 91.12% ┃
13
+ ┃ 9 │ 52:10 min ┃ 6.166e-05 │ 0.3135 │ 92.16% ┃ 0.3431 │ 91.25% ┃
14
+ ┃ 10 │ 52:15 min ┃ 5.392e-05 │ 0.2944 │ 92.78% ┃ 0.3361 │ 91.41% ┃
15
+ ┃ 11 │ 52:15 min ┃ 4.608e-05 │ 0.2808 │ 93.16% ┃ 0.3339 │ 91.30% ┃
16
+ ┃ 12 │ 50:39 min ┃ 3.834e-05 │ 0.2687 │ 93.43% ┃ 0.3284 │ 91.62% ┃
17
+ ┃ 13 │ 52:10 min ┃ 3.088e-05 │ 0.2600 │ 93.68% ┃ 0.3259 │ 91.51% ┃
18
+ ┃ 14 │ 52:14 min ┃ 2.390e-05 │ 0.2521 │ 94.01% ┃ 0.3243 │ 91.61% ┃
19
+ ┃ 15 │ 52:10 min ┃ 1.756e-05 │ 0.2479 │ 94.09% ┃ 0.3221 │ 91.70% ┃
20
+ ┃ 16 │ 52:15 min ┃ 1.202e-05 │ 0.2435 │ 94.19% ┃ 0.3207 │ 91.68% ┃
21
+ ┃ 17 │ 52:15 min ┃ 7.410e-06 │ 0.2403 │ 94.34% ┃ 0.3201 │ 91.72% ┃
22
+ ┃ 18 │ 50:39 min ┃ 3.853e-06 │ 0.2383 │ 94.34% ┃ 0.3198 │ 91.72% ┃
23
+ ┃ 19 │ 52:09 min ┃ 1.431e-06 │ 0.2379 │ 94.33% ┃ 0.3197 │ 91.75% ┃
24
+ ┃ 20 │ 52:14 min ┃ 2.052e-07 │ 0.2380 │ 94.45% ┃ 0.3197 │ 91.73% ┃
TinyImageNet_FriendlySAM_ViT_rho0.01_lr0.0001/data.hdf5 ADDED
Binary file (56 kB). View file
 
TinyImageNet_FriendlySAM_ViT_rho0.01_lr0.0001/final.model ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:3149cf5bd989bcf9c3d204a93be8bde2bd89b1aa65d349ecf8925f860aea6e86
3
+ size 1403071210
TinyImageNet_FriendlySAM_ViT_rho0.01_lr0.0001/params.json ADDED
@@ -0,0 +1,97 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "widthFactor": 4,
3
+ "ViTPatchsize": 32,
4
+ "ViTSize": "B",
5
+ "pyramidAlpha": 270,
6
+ "pyramidBottleneck": true,
7
+ "model": "ViT",
8
+ "dropout": 0.0,
9
+ "BN": true,
10
+ "depth": 16,
11
+ "label_smoothing": 0.1,
12
+ "ESAMbeta": 0.5,
13
+ "ESAMgamma": 0.7,
14
+ "lookSAMk": 5,
15
+ "lookSAMalpha": 0.7,
16
+ "lookSAMLayerwise": false,
17
+ "optimizer": "FriendlySAM",
18
+ "weightDecay": 0.001,
19
+ "momentum": 0.9,
20
+ "rho": 0.01,
21
+ "samRho": 0.01,
22
+ "asamRho": 1.0,
23
+ "esamRho": 0.075,
24
+ "looksamRho": 0.05,
25
+ "friendlysamRho": 0.01,
26
+ "fishersamRho": 0.01,
27
+ "mongesamRho": 0.075,
28
+ "gsamRho": 0.01,
29
+ "bayesiansamRho": 0.075,
30
+ "msamRho": 0.075,
31
+ "adamwSamRho": 0.075,
32
+ "adamwMsamRho": 0.075,
33
+ "adamwFriendlysamRho": 0.075,
34
+ "sigma": 1.0,
35
+ "lmbda": 0.95,
36
+ "fisherEta": 0.1,
37
+ "gsamAlpha": 0.1,
38
+ "gsamAdaptive": false,
39
+ "bayesMSharpness": 8,
40
+ "bayesDelta": 10.0,
41
+ "bayesBeta1": 0.9,
42
+ "bayesBeta2": 0.999,
43
+ "bayesSInit": 1.0,
44
+ "bayesDamping": 1e-08,
45
+ "bayesEps": 1e-12,
46
+ "nesterov": false,
47
+ "grad_clip_norm": 0.0,
48
+ "dataThreads": 4,
49
+ "dataDir": "/home/hungpv/.datasets",
50
+ "batchSize": 64,
51
+ "dataset": "TinyImageNet",
52
+ "imageSize": 64,
53
+ "flip": true,
54
+ "crop": true,
55
+ "cut": false,
56
+ "cutoutProp": 0.5,
57
+ "randAugment": true,
58
+ "randAugment_magnitude": 10,
59
+ "mixup": false,
60
+ "mixupProp": 0.2,
61
+ "normalize": "std",
62
+ "truncate": false,
63
+ "verbose": false,
64
+ "metrics": [
65
+ "learningRate",
66
+ "loss",
67
+ "accuracy"
68
+ ],
69
+ "logEach": 1,
70
+ "rndSeed": false,
71
+ "deterministic": false,
72
+ "learningRate": 0.0001,
73
+ "lrScheduler": [
74
+ "startRamp",
75
+ "cos"
76
+ ],
77
+ "LRScheduler_step_steps": 3.0,
78
+ "LRScheduler_step_gamma": 0.2,
79
+ "LRScheduler_exp_maxDecay": 0.01,
80
+ "LRScheduler_cos_periods": 0.5,
81
+ "LRScheduler_StartRamp_epochs": 1.0,
82
+ "LRScheduler_WRN_T0": 50.0,
83
+ "LRScheduler_WRN_Tmult": 1.2,
84
+ "rhoScheduler": [
85
+ "const"
86
+ ],
87
+ "rhoScheduler_StartRamp_epochs": 8.0,
88
+ "rhoScheduler_StartJump_epochs": 8.0,
89
+ "saveCheckpoint": true,
90
+ "keepLastCheckpoint": false,
91
+ "checkpointsList": [],
92
+ "saveCheckpointInterval": 0,
93
+ "logDir": "/mnt/hungpv/projects/SAM_Variant/logs",
94
+ "logSubDir": "TinyImageNet_FriendlySAM_ViT_rho0.01_lr0.0001",
95
+ "epochs": 20,
96
+ "contin": false
97
+ }
TinyImageNet_FriendlySAM_ViT_rho0.01_lr0.0001/terminal.log ADDED
@@ -0,0 +1,24 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ┏━━━━━━━╸S╺╸T╺╸A╺╸T╺╸S╺━━━━━━━┳━━━━━━━━━━━━━━━━━T╺╸R╺╸A╺╸I╺╸N ━━━━━━━━━━━━━━━━┳━━━━━━━━━━T╺╸E╺╸S╺╸T ━━━━━━━━━━┓
2
+ ┃ ┃ ┃ ┃
3
+ ┃ epoch │ time ┃ learningRate │ loss │ accuracy ┃ loss │ accuracy ┃
4
+ ┠──────────────┼──────────────╂───────────────┼───────────────┼───────────────╂───────────────┼───────────────┨
5
+ ┃ 1 │ 13:46 min ┃ 4.988e-05 │ 4.4879 │ 2.64% ┃ 3.2766 │ 19.22% ┃
6
+ ┃ 2 │ 12:35 min ┃ 9.857e-05 │ 1.4256 │ 63.77% ┃ 0.6835 │ 83.11% ┃
7
+ ┃ 3 │ 12:34 min ┃ 9.615e-05 │ 0.7552 │ 80.83% ┃ 0.5452 │ 86.15% ┃
8
+ ┃ 4 │ 12:29 min ┃ 9.259e-05 │ 0.6267 │ 83.71% ┃ 0.4870 │ 87.44% ┃
9
+ ┃ 5 │ 12:34 min ┃ 8.798e-05 │ 0.5533 │ 85.57% ┃ 0.4559 │ 88.09% ┃
10
+ ┃ 6 │ 12:34 min ┃ 8.244e-05 │ 0.5055 │ 86.72% ┃ 0.4356 │ 88.37% ┃
11
+ ┃ 7 │ 12:35 min ┃ 7.610e-05 │ 0.4694 │ 87.55% ┃ 0.4219 │ 88.88% ┃
12
+ ┃ 8 │ 12:29 min ┃ 6.911e-05 │ 0.4417 │ 88.30% ┃ 0.4099 │ 88.96% ┃
13
+ ┃ 9 │ 12:34 min ┃ 6.166e-05 │ 0.4218 │ 88.86% ┃ 0.4008 │ 89.23% ┃
14
+ ┃ 10 │ 12:34 min ┃ 5.392e-05 │ 0.3999 │ 89.47% ┃ 0.3935 │ 89.36% ┃
15
+ ┃ 11 │ 12:34 min ┃ 4.608e-05 │ 0.3863 │ 89.82% ┃ 0.3889 │ 89.49% ┃
16
+ ┃ 12 │ 12:28 min ┃ 3.834e-05 │ 0.3768 │ 90.17% ┃ 0.3841 │ 89.68% ┃
17
+ ┃ 13 │ 12:34 min ┃ 3.088e-05 │ 0.3655 │ 90.44% ┃ 0.3817 │ 89.69% ┃
18
+ ┃ 14 │ 12:34 min ┃ 2.390e-05 │ 0.3562 │ 90.80% ┃ 0.3792 │ 89.78% ┃
19
+ ┃ 15 │ 12:34 min ┃ 1.756e-05 │ 0.3501 │ 90.87% ┃ 0.3769 │ 89.73% ┃
20
+ ┃ 16 │ 12:29 min ┃ 1.202e-05 │ 0.3445 │ 91.10% ┃ 0.3762 │ 89.80% ┃
21
+ ┃ 17 │ 12:34 min ┃ 7.410e-06 │ 0.3425 │ 91.22% ┃ 0.3753 │ 89.81% ┃
22
+ ┃ 18 │ 12:34 min ┃ 3.853e-06 │ 0.3391 │ 91.27% ┃ 0.3750 │ 89.83% ┃
23
+ ┃ 19 │ 12:34 min ┃ 1.431e-06 │ 0.3371 │ 91.30% ┃ 0.3749 │ 89.84% ┃
24
+ ┃ 20 │ 12:29 min ┃ 2.052e-07 │ 0.3386 │ 91.25% ┃ 0.3749 │ 89.85% ┃
TinyImageNet_GSAM_ViT_rho0.01_lr0.0001/data.hdf5 ADDED
Binary file (56 kB). View file
 
TinyImageNet_GSAM_ViT_rho0.01_lr0.0001/final.model ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:9e83b3f1029f585242c97fbc10ad47b5034880b3ae1c82f4e8b1d2e52104584a
3
+ size 1753832650
TinyImageNet_GSAM_ViT_rho0.01_lr0.0001/params.json ADDED
@@ -0,0 +1,97 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "widthFactor": 4,
3
+ "ViTPatchsize": 32,
4
+ "ViTSize": "B",
5
+ "pyramidAlpha": 270,
6
+ "pyramidBottleneck": true,
7
+ "model": "ViT",
8
+ "dropout": 0.0,
9
+ "BN": true,
10
+ "depth": 16,
11
+ "label_smoothing": 0.1,
12
+ "ESAMbeta": 0.5,
13
+ "ESAMgamma": 0.7,
14
+ "lookSAMk": 5,
15
+ "lookSAMalpha": 0.7,
16
+ "lookSAMLayerwise": false,
17
+ "optimizer": "GSAM",
18
+ "weightDecay": 0.001,
19
+ "momentum": 0.9,
20
+ "rho": 0.01,
21
+ "samRho": 0.01,
22
+ "asamRho": 1.0,
23
+ "esamRho": 0.075,
24
+ "looksamRho": 0.05,
25
+ "friendlysamRho": 0.01,
26
+ "fishersamRho": 0.01,
27
+ "mongesamRho": 0.075,
28
+ "gsamRho": 0.01,
29
+ "bayesiansamRho": 0.075,
30
+ "msamRho": 0.075,
31
+ "adamwSamRho": 0.075,
32
+ "adamwMsamRho": 0.075,
33
+ "adamwFriendlysamRho": 0.075,
34
+ "sigma": 1.0,
35
+ "lmbda": 0.95,
36
+ "fisherEta": 0.1,
37
+ "gsamAlpha": 0.1,
38
+ "gsamAdaptive": false,
39
+ "bayesMSharpness": 8,
40
+ "bayesDelta": 10.0,
41
+ "bayesBeta1": 0.9,
42
+ "bayesBeta2": 0.999,
43
+ "bayesSInit": 1.0,
44
+ "bayesDamping": 1e-08,
45
+ "bayesEps": 1e-12,
46
+ "nesterov": false,
47
+ "grad_clip_norm": 0.0,
48
+ "dataThreads": 4,
49
+ "dataDir": "/home/hungpv/.datasets",
50
+ "batchSize": 64,
51
+ "dataset": "TinyImageNet",
52
+ "imageSize": 64,
53
+ "flip": true,
54
+ "crop": true,
55
+ "cut": false,
56
+ "cutoutProp": 0.5,
57
+ "randAugment": true,
58
+ "randAugment_magnitude": 10,
59
+ "mixup": false,
60
+ "mixupProp": 0.2,
61
+ "normalize": "std",
62
+ "truncate": false,
63
+ "verbose": false,
64
+ "metrics": [
65
+ "learningRate",
66
+ "loss",
67
+ "accuracy"
68
+ ],
69
+ "logEach": 1,
70
+ "rndSeed": false,
71
+ "deterministic": false,
72
+ "learningRate": 0.0001,
73
+ "lrScheduler": [
74
+ "startRamp",
75
+ "cos"
76
+ ],
77
+ "LRScheduler_step_steps": 3.0,
78
+ "LRScheduler_step_gamma": 0.2,
79
+ "LRScheduler_exp_maxDecay": 0.01,
80
+ "LRScheduler_cos_periods": 0.5,
81
+ "LRScheduler_StartRamp_epochs": 1.0,
82
+ "LRScheduler_WRN_T0": 50.0,
83
+ "LRScheduler_WRN_Tmult": 1.2,
84
+ "rhoScheduler": [
85
+ "const"
86
+ ],
87
+ "rhoScheduler_StartRamp_epochs": 8.0,
88
+ "rhoScheduler_StartJump_epochs": 8.0,
89
+ "saveCheckpoint": true,
90
+ "keepLastCheckpoint": false,
91
+ "checkpointsList": [],
92
+ "saveCheckpointInterval": 0,
93
+ "logDir": "/mnt/hungpv/projects/SAM_Variant/logs",
94
+ "logSubDir": "TinyImageNet_GSAM_ViT_rho0.01_lr0.0001",
95
+ "epochs": 20,
96
+ "contin": false
97
+ }
TinyImageNet_GSAM_ViT_rho0.01_lr0.0001/terminal.log ADDED
@@ -0,0 +1,24 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ┏━━━━━━━╸S╺╸T╺╸A╺╸T╺╸S╺━━━━━━━┳━━━━━━━━━━━━━━━━━T╺╸R╺╸A╺╸I╺╸N ━━━━━━━━━━━━━━━━┳━━━━━━━━━━T╺╸E╺╸S╺╸T ━━━━━━━━━━┓
2
+ ┃ ┃ ┃ ┃
3
+ ┃ epoch │ time ┃ learningRate │ loss │ accuracy ┃ loss │ accuracy ┃
4
+ ┠──────────────┼──────────────╂───────────────┼───────────────┼───────────────╂───────────────┼───────────────┨
5
+ ┃ 1 │ 14:24 min ┃ 4.988e-05 │ 4.4898 │ 2.64% ┃ 3.2798 │ 19.23% ┃
6
+ ┃ 2 │ 13:15 min ┃ 9.857e-05 │ 1.4294 │ 63.69% ┃ 0.6861 │ 83.01% ┃
7
+ ┃ 3 │ 13:09 min ┃ 9.615e-05 │ 0.7579 │ 80.76% ┃ 0.5472 │ 86.17% ┃
8
+ ┃ 4 │ 13:16 min ┃ 9.259e-05 │ 0.6296 │ 83.66% ┃ 0.4887 │ 87.45% ┃
9
+ ┃ 5 │ 13:14 min ┃ 8.798e-05 │ 0.5566 │ 85.49% ┃ 0.4577 │ 88.06% ┃
10
+ ┃ 6 │ 13:14 min ┃ 8.244e-05 │ 0.5090 │ 86.65% ┃ 0.4369 │ 88.31% ┃
11
+ ┃ 7 │ 13:09 min ┃ 7.610e-05 │ 0.4731 │ 87.46% ┃ 0.4236 │ 88.81% ┃
12
+ ┃ 8 │ 13:14 min ┃ 6.911e-05 │ 0.4456 │ 88.23% ┃ 0.4116 │ 88.98% ┃
13
+ ┃ 9 │ 13:14 min ┃ 6.166e-05 │ 0.4259 │ 88.77% ┃ 0.4026 │ 89.19% ┃
14
+ ┃ 10 │ 13:14 min ┃ 5.392e-05 │ 0.4042 │ 89.38% ┃ 0.3950 │ 89.39% ┃
15
+ ┃ 11 │ 13:16 min ┃ 4.608e-05 │ 0.3907 │ 89.74% ┃ 0.3903 │ 89.42% ┃
16
+ ┃ 12 │ 13:14 min ┃ 3.834e-05 │ 0.3814 │ 90.05% ┃ 0.3856 │ 89.62% ┃
17
+ ┃ 13 │ 13:14 min ┃ 3.088e-05 │ 0.3702 │ 90.36% ┃ 0.3832 │ 89.66% ┃
18
+ ┃ 14 │ 13:13 min ┃ 2.390e-05 │ 0.3609 │ 90.67% ┃ 0.3807 │ 89.77% ┃
19
+ ┃ 15 │ 13:09 min ┃ 1.756e-05 │ 0.3550 │ 90.78% ┃ 0.3783 │ 89.59% ┃
20
+ ┃ 16 │ 13:14 min ┃ 1.202e-05 │ 0.3493 │ 91.01% ┃ 0.3776 │ 89.72% ┃
21
+ ┃ 17 │ 13:14 min ┃ 7.410e-06 │ 0.3474 │ 91.07% ┃ 0.3767 │ 89.76% ┃
22
+ ┃ 18 │ 12:25 min ┃ 3.853e-06 │ 0.3440 │ 91.16% ┃ 0.3765 │ 89.78% ┃
23
+ ┃ 19 │ 13:14 min ┃ 1.431e-06 │ 0.3421 │ 91.17% ┃ 0.3763 │ 89.80% ┃
24
+ ┃ 20 │ 13:14 min ┃ 2.052e-07 │ 0.3436 │ 91.15% ┃ 0.3763 │ 89.81% ┃
TinyImageNet_MSAM_ViT_rho0.075_lr0.0001/data.hdf5 ADDED
Binary file (56 kB). View file
 
TinyImageNet_MSAM_ViT_rho0.075_lr0.0001/final.model ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:bb6a44e93cfae55dfb62dd23e14cebb14a7120a63ab64ce8e42ccf684a74cade
3
+ size 701548454
TinyImageNet_MSAM_ViT_rho0.075_lr0.0001/params.json ADDED
@@ -0,0 +1,97 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "widthFactor": 4,
3
+ "ViTPatchsize": 32,
4
+ "ViTSize": "B",
5
+ "pyramidAlpha": 270,
6
+ "pyramidBottleneck": true,
7
+ "model": "ViT",
8
+ "dropout": 0.0,
9
+ "BN": true,
10
+ "depth": 16,
11
+ "label_smoothing": 0.1,
12
+ "ESAMbeta": 0.5,
13
+ "ESAMgamma": 0.7,
14
+ "lookSAMk": 5,
15
+ "lookSAMalpha": 0.7,
16
+ "lookSAMLayerwise": false,
17
+ "optimizer": "MSAM",
18
+ "weightDecay": 0.001,
19
+ "momentum": 0.9,
20
+ "rho": 0.075,
21
+ "samRho": 0.075,
22
+ "asamRho": 1.0,
23
+ "esamRho": 0.075,
24
+ "looksamRho": 0.05,
25
+ "friendlysamRho": 0.075,
26
+ "fishersamRho": 0.5,
27
+ "mongesamRho": 0.075,
28
+ "gsamRho": 0.075,
29
+ "bayesiansamRho": 0.075,
30
+ "msamRho": 0.075,
31
+ "adamwSamRho": 0.075,
32
+ "adamwMsamRho": 0.075,
33
+ "adamwFriendlysamRho": 0.075,
34
+ "sigma": 1.0,
35
+ "lmbda": 0.95,
36
+ "fisherEta": 0.1,
37
+ "gsamAlpha": 0.1,
38
+ "gsamAdaptive": false,
39
+ "bayesMSharpness": 8,
40
+ "bayesDelta": 10.0,
41
+ "bayesBeta1": 0.9,
42
+ "bayesBeta2": 0.999,
43
+ "bayesSInit": 1.0,
44
+ "bayesDamping": 1e-08,
45
+ "bayesEps": 1e-12,
46
+ "nesterov": false,
47
+ "grad_clip_norm": 0.0,
48
+ "dataThreads": 4,
49
+ "dataDir": "/home/hungpv/.datasets",
50
+ "batchSize": 64,
51
+ "dataset": "TinyImageNet",
52
+ "imageSize": 64,
53
+ "flip": true,
54
+ "crop": true,
55
+ "cut": false,
56
+ "cutoutProp": 0.5,
57
+ "randAugment": true,
58
+ "randAugment_magnitude": 10,
59
+ "mixup": false,
60
+ "mixupProp": 0.2,
61
+ "normalize": "std",
62
+ "truncate": false,
63
+ "verbose": false,
64
+ "metrics": [
65
+ "learningRate",
66
+ "loss",
67
+ "accuracy"
68
+ ],
69
+ "logEach": 1,
70
+ "rndSeed": false,
71
+ "deterministic": false,
72
+ "learningRate": 0.0001,
73
+ "lrScheduler": [
74
+ "startRamp",
75
+ "cos"
76
+ ],
77
+ "LRScheduler_step_steps": 3.0,
78
+ "LRScheduler_step_gamma": 0.2,
79
+ "LRScheduler_exp_maxDecay": 0.01,
80
+ "LRScheduler_cos_periods": 0.5,
81
+ "LRScheduler_StartRamp_epochs": 1.0,
82
+ "LRScheduler_WRN_T0": 50.0,
83
+ "LRScheduler_WRN_Tmult": 1.2,
84
+ "rhoScheduler": [
85
+ "const"
86
+ ],
87
+ "rhoScheduler_StartRamp_epochs": 8.0,
88
+ "rhoScheduler_StartJump_epochs": 8.0,
89
+ "saveCheckpoint": true,
90
+ "keepLastCheckpoint": false,
91
+ "checkpointsList": [],
92
+ "saveCheckpointInterval": 0,
93
+ "logDir": "/mnt/hungpv/projects/SAM_Variant/logs",
94
+ "logSubDir": "TinyImageNet_MSAM_ViT_rho0.075_lr0.0001",
95
+ "epochs": 20,
96
+ "contin": false
97
+ }
TinyImageNet_MSAM_ViT_rho0.075_lr0.0001/terminal.log ADDED
@@ -0,0 +1,24 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ┏━━━━━━━╸S╺╸T╺╸A╺╸T╺╸S╺━━━━━━━┳━━━━━━━━━━━━━━━━━T╺╸R╺╸A╺╸I╺╸N ━━━━━━━━━━━━━━━━┳━━━━━━━━━━T╺╸E╺╸S╺╸T ━━━━━━━━━━┓
2
+ ┃ ┃ ┃ ┃
3
+ ┃ epoch │ time ┃ learningRate │ loss │ accuracy ┃ loss │ accuracy ┃
4
+ ┠──────────────┼──────────────╂───────────────┼───────────────┼───────────────╂───────────────┼───────────────┨
5
+ ┃ 1 │ 04:16 min ┃ 4.988e-05 │ 4.2781 │ 4.77% ┃ 2.4957 │ 35.53% ┃
6
+ ┃ 2 │ 03:00 min ┃ 9.857e-05 │ 1.2220 │ 68.81% ┃ 0.6547 │ 83.88% ┃
7
+ ┃ 3 │ 03:04 min ┃ 9.615e-05 │ 0.7458 │ 81.02% ┃ 0.5378 │ 86.17% ┃
8
+ ┃ 4 │ 03:01 min ┃ 9.259e-05 │ 0.6224 │ 83.70% ┃ 0.4808 │ 87.50% ┃
9
+ ┃ 5 │ 03:01 min ┃ 8.798e-05 │ 0.5484 │ 85.54% ┃ 0.4513 │ 88.13% ┃
10
+ ┃ 6 │ 03:01 min ┃ 8.244e-05 │ 0.5001 │ 86.67% ┃ 0.4340 │ 88.58% ┃
11
+ ┃ 7 │ 03:01 min ┃ 7.610e-05 │ 0.4615 │ 87.59% ┃ 0.4191 │ 88.78% ┃
12
+ ┃ 8 │ 03:10 min ┃ 6.911e-05 │ 0.4321 │ 88.31% ┃ 0.4073 │ 88.97% ┃
13
+ ┃ 9 │ 03:02 min ┃ 6.166e-05 │ 0.4116 │ 88.86% ┃ 0.3996 │ 89.16% ┃
14
+ ┃ 10 │ 03:01 min ┃ 5.392e-05 │ 0.3880 │ 89.56% ┃ 0.3953 │ 89.44% ┃
15
+ ┃ 11 │ 03:01 min ┃ 4.608e-05 │ 0.3740 │ 89.91% ┃ 0.3910 │ 89.38% ┃
16
+ ┃ 12 │ 03:01 min ┃ 3.834e-05 │ 0.3619 │ 90.25% ┃ 0.3871 │ 89.51% ┃
17
+ ┃ 13 │ 03:03 min ┃ 3.088e-05 │ 0.3492 │ 90.62% ┃ 0.3844 │ 89.57% ┃
18
+ ┃ 14 │ 03:01 min ┃ 2.390e-05 │ 0.3399 │ 90.92% ┃ 0.3819 │ 89.65% ┃
19
+ ┃ 15 │ 03:01 min ┃ 1.756e-05 │ 0.3326 │ 91.11% ┃ 0.3806 │ 89.60% ┃
20
+ ┃ 16 │ 03:01 min ┃ 1.202e-05 │ 0.3265 │ 91.33% ┃ 0.3801 │ 89.65% ┃
21
+ ┃ 17 │ 03:04 min ┃ 7.410e-06 │ 0.3239 │ 91.43% ┃ 0.3794 │ 89.73% ┃
22
+ ┃ 18 │ 03:01 min ┃ 3.853e-06 │ 0.3210 │ 91.46% ┃ 0.3788 │ 89.71% ┃
23
+ ┃ 19 │ 03:01 min ┃ 1.431e-06 │ 0.3178 │ 91.59% ┃ 0.3790 │ 89.68% ┃
24
+ ┃ 20 │ 03:05 min ┃ 2.052e-07 │ 0.3200 │ 91.48% ┃ 0.3789 │ 89.70% ┃
TinyImageNet_MSAM_ViT_rho0.075_lr0.0001_run01/data.hdf5 ADDED
Binary file (56 kB). View file
 
TinyImageNet_MSAM_ViT_rho0.075_lr0.0001_run01/final.model ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:4c0406d8a275e5f666dbad6956740d983faf42badeda1c536108e2db5ae6bc59
3
+ size 690046886
TinyImageNet_MSAM_ViT_rho0.075_lr0.0001_run01/params.json ADDED
@@ -0,0 +1,97 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "widthFactor": 4,
3
+ "ViTPatchsize": 16,
4
+ "ViTSize": "B",
5
+ "pyramidAlpha": 270,
6
+ "pyramidBottleneck": true,
7
+ "model": "ViT",
8
+ "dropout": 0.0,
9
+ "BN": true,
10
+ "depth": 16,
11
+ "label_smoothing": 0.1,
12
+ "ESAMbeta": 0.5,
13
+ "ESAMgamma": 0.7,
14
+ "lookSAMk": 5,
15
+ "lookSAMalpha": 0.7,
16
+ "lookSAMLayerwise": false,
17
+ "optimizer": "MSAM",
18
+ "weightDecay": 0.001,
19
+ "momentum": 0.9,
20
+ "rho": 0.01,
21
+ "samRho": 0.01,
22
+ "asamRho": 1.0,
23
+ "esamRho": 0.075,
24
+ "looksamRho": 0.05,
25
+ "friendlysamRho": 0.075,
26
+ "fishersamRho": 0.1,
27
+ "mongesamRho": 0.075,
28
+ "gsamRho": 0.075,
29
+ "bayesiansamRho": 0.075,
30
+ "msamRho": 0.075,
31
+ "adamwSamRho": 0.075,
32
+ "adamwMsamRho": 0.075,
33
+ "adamwFriendlysamRho": 0.075,
34
+ "sigma": 1.0,
35
+ "lmbda": 0.95,
36
+ "fisherEta": 0.1,
37
+ "gsamAlpha": 0.1,
38
+ "gsamAdaptive": false,
39
+ "bayesMSharpness": 8,
40
+ "bayesDelta": 10.0,
41
+ "bayesBeta1": 0.9,
42
+ "bayesBeta2": 0.999,
43
+ "bayesSInit": 1.0,
44
+ "bayesDamping": 1e-08,
45
+ "bayesEps": 1e-12,
46
+ "nesterov": false,
47
+ "grad_clip_norm": 0.0,
48
+ "dataThreads": 4,
49
+ "dataDir": "/home/hungpv/.datasets",
50
+ "batchSize": 64,
51
+ "dataset": "TinyImageNet",
52
+ "imageSize": 64,
53
+ "flip": true,
54
+ "crop": true,
55
+ "cut": false,
56
+ "cutoutProp": 0.5,
57
+ "randAugment": true,
58
+ "randAugment_magnitude": 10,
59
+ "mixup": false,
60
+ "mixupProp": 0.2,
61
+ "normalize": "std",
62
+ "truncate": false,
63
+ "verbose": false,
64
+ "metrics": [
65
+ "learningRate",
66
+ "loss",
67
+ "accuracy"
68
+ ],
69
+ "logEach": 1,
70
+ "rndSeed": false,
71
+ "deterministic": false,
72
+ "learningRate": 0.0001,
73
+ "lrScheduler": [
74
+ "startRamp",
75
+ "cos"
76
+ ],
77
+ "LRScheduler_step_steps": 3.0,
78
+ "LRScheduler_step_gamma": 0.2,
79
+ "LRScheduler_exp_maxDecay": 0.01,
80
+ "LRScheduler_cos_periods": 0.5,
81
+ "LRScheduler_StartRamp_epochs": 1.0,
82
+ "LRScheduler_WRN_T0": 50.0,
83
+ "LRScheduler_WRN_Tmult": 1.2,
84
+ "rhoScheduler": [
85
+ "const"
86
+ ],
87
+ "rhoScheduler_StartRamp_epochs": 8.0,
88
+ "rhoScheduler_StartJump_epochs": 8.0,
89
+ "saveCheckpoint": true,
90
+ "keepLastCheckpoint": false,
91
+ "checkpointsList": [],
92
+ "saveCheckpointInterval": 0,
93
+ "logDir": "/mnt/hungpv/projects/SAM_Variant/logs",
94
+ "logSubDir": "TinyImageNet_MSAM_ViT_rho0.075_lr0.0001_run01",
95
+ "epochs": 20,
96
+ "contin": false
97
+ }
TinyImageNet_MSAM_ViT_rho0.075_lr0.0001_run01/terminal.log ADDED
@@ -0,0 +1,24 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ┏━━━━━━━╸S╺╸T╺╸A╺╸T╺╸S╺━━━━━━━┳━━━━━━━━━━━━━━━━━T╺╸R╺╸A╺╸I╺╸N ━━━━━━━━━━━━━━━━┳━━━━━━━━━━T╺╸E╺╸S╺╸T ━━━━━━━━━━┓
2
+ ┃ ┃ ┃ ┃
3
+ ┃ epoch │ time ┃ learningRate │ loss │ accuracy ┃ loss │ accuracy ┃
4
+ ┠──────────────┼──────────────╂───────────────┼───────────────┼───────────────╂───────────────┼───────────────┨
5
+ ┃ 1 │ 37:28 min ┃ 4.988e-05 │ 4.4369 │ 2.38% ┃ 3.7020 │ 9.98% ┃
6
+ ┃ 2 │ 36:04 min ┃ 9.857e-05 │ 2.2456 │ 40.90% ┃ 1.1276 │ 71.09% ┃
7
+ ┃ 3 │ 36:02 min ┃ 9.615e-05 │ 0.8574 │ 78.75% ┃ 0.5902 │ 86.32% ┃
8
+ ┃ 4 │ 36:15 min ┃ 9.259e-05 │ 0.5768 │ 85.98% ┃ 0.4834 │ 88.49% ┃
9
+ ┃ 5 │ 34:54 min ┃ 8.798e-05 │ 0.4718 │ 88.47% ┃ 0.4307 │ 89.42% ┃
10
+ ┃ 6 │ 34:31 min ┃ 8.244e-05 │ 0.4079 │ 89.73% ┃ 0.3970 │ 90.19% ┃
11
+ ┃ 7 │ 36:07 min ┃ 7.610e-05 │ 0.3633 │ 90.81% ┃ 0.3857 │ 90.23% ┃
12
+ ┃ 8 │ 36:17 min ┃ 6.911e-05 │ 0.3324 │ 91.58% ┃ 0.3707 │ 90.49% ┃
13
+ ┃ 9 │ 35:59 min ┃ 6.166e-05 │ 0.3069 │ 92.24% ┃ 0.3639 │ 90.64% ┃
14
+ ┃ 10 │ 36:18 min ┃ 5.392e-05 │ 0.2838 │ 92.97% ┃ 0.3571 │ 90.76% ┃
15
+ ┃ 11 │ 36:06 min ┃ 4.608e-05 │ 0.2670 │ 93.37% ┃ 0.3555 │ 90.62% ┃
16
+ ┃ 12 │ 25:48 min ┃ 3.834e-05 │ 0.2526 │ 93.77% ┃ 0.3480 │ 91.06% ┃
17
+ ┃ 13 │ 12:42 min ┃ 3.088e-05 │ 0.2410 │ 94.12% ┃ 0.3470 │ 90.95% ┃
18
+ ┃ 14 │ 12:42 min ┃ 2.390e-05 │ 0.2317 │ 94.41% ┃ 0.3470 │ 91.00% ┃
19
+ ┃ 15 │ 12:42 min ┃ 1.756e-05 │ 0.2263 │ 94.54% ┃ 0.3439 │ 90.97% ┃
20
+ ┃ 16 │ 12:42 min ┃ 1.202e-05 │ 0.2209 │ 94.70% ┃ 0.3427 │ 91.01% ┃
21
+ ┃ 17 │ 12:42 min ┃ 7.410e-06 │ 0.2165 │ 94.85% ┃ 0.3419 │ 91.02% ┃
22
+ ┃ 18 │ 12:43 min ┃ 3.853e-06 │ 0.2139 │ 94.89% ┃ 0.3418 │ 91.00% ┃
23
+ ┃ 19 │ 12:42 min ┃ 1.431e-06 │ 0.2137 │ 94.91% ┃ 0.3417 │ 91.04% ┃
24
+ ┃ 20 │ 12:42 min ┃ 2.052e-07 │ 0.2129 │ 94.95% ┃ 0.3416 │ 91.10% ┃
TinyImageNet_MongeSAM_ViT_rho0.075_lr0.0001/data.hdf5 ADDED
Binary file (56 kB). View file
 
TinyImageNet_MongeSAM_ViT_rho0.075_lr0.0001/final.model ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:2c72611fc23f45bed32a9e92e634dec2e4b7b6010ba0fd64b37e249b4fe97fb5
3
+ size 1052309706
TinyImageNet_MongeSAM_ViT_rho0.075_lr0.0001/params.json ADDED
@@ -0,0 +1,97 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "widthFactor": 4,
3
+ "ViTPatchsize": 32,
4
+ "ViTSize": "B",
5
+ "pyramidAlpha": 270,
6
+ "pyramidBottleneck": true,
7
+ "model": "ViT",
8
+ "dropout": 0.0,
9
+ "BN": true,
10
+ "depth": 16,
11
+ "label_smoothing": 0.1,
12
+ "ESAMbeta": 0.5,
13
+ "ESAMgamma": 0.7,
14
+ "lookSAMk": 5,
15
+ "lookSAMalpha": 0.7,
16
+ "lookSAMLayerwise": false,
17
+ "optimizer": "MongeSAM",
18
+ "weightDecay": 0.001,
19
+ "momentum": 0.9,
20
+ "rho": 0.075,
21
+ "samRho": 0.075,
22
+ "asamRho": 1.0,
23
+ "esamRho": 0.075,
24
+ "looksamRho": 0.05,
25
+ "friendlysamRho": 0.075,
26
+ "fishersamRho": 0.5,
27
+ "mongesamRho": 0.075,
28
+ "gsamRho": 0.075,
29
+ "bayesiansamRho": 0.075,
30
+ "msamRho": 0.075,
31
+ "adamwSamRho": 0.075,
32
+ "adamwMsamRho": 0.075,
33
+ "adamwFriendlysamRho": 0.075,
34
+ "sigma": 1.0,
35
+ "lmbda": 0.95,
36
+ "fisherEta": 0.1,
37
+ "gsamAlpha": 0.1,
38
+ "gsamAdaptive": false,
39
+ "bayesMSharpness": 8,
40
+ "bayesDelta": 10.0,
41
+ "bayesBeta1": 0.9,
42
+ "bayesBeta2": 0.999,
43
+ "bayesSInit": 1.0,
44
+ "bayesDamping": 1e-08,
45
+ "bayesEps": 1e-12,
46
+ "nesterov": false,
47
+ "grad_clip_norm": 0.0,
48
+ "dataThreads": 4,
49
+ "dataDir": "/home/hungpv/.datasets",
50
+ "batchSize": 64,
51
+ "dataset": "TinyImageNet",
52
+ "imageSize": 64,
53
+ "flip": true,
54
+ "crop": true,
55
+ "cut": false,
56
+ "cutoutProp": 0.5,
57
+ "randAugment": true,
58
+ "randAugment_magnitude": 10,
59
+ "mixup": false,
60
+ "mixupProp": 0.2,
61
+ "normalize": "std",
62
+ "truncate": false,
63
+ "verbose": false,
64
+ "metrics": [
65
+ "learningRate",
66
+ "loss",
67
+ "accuracy"
68
+ ],
69
+ "logEach": 1,
70
+ "rndSeed": false,
71
+ "deterministic": false,
72
+ "learningRate": 0.0001,
73
+ "lrScheduler": [
74
+ "startRamp",
75
+ "cos"
76
+ ],
77
+ "LRScheduler_step_steps": 3.0,
78
+ "LRScheduler_step_gamma": 0.2,
79
+ "LRScheduler_exp_maxDecay": 0.01,
80
+ "LRScheduler_cos_periods": 0.5,
81
+ "LRScheduler_StartRamp_epochs": 1.0,
82
+ "LRScheduler_WRN_T0": 50.0,
83
+ "LRScheduler_WRN_Tmult": 1.2,
84
+ "rhoScheduler": [
85
+ "const"
86
+ ],
87
+ "rhoScheduler_StartRamp_epochs": 8.0,
88
+ "rhoScheduler_StartJump_epochs": 8.0,
89
+ "saveCheckpoint": true,
90
+ "keepLastCheckpoint": false,
91
+ "checkpointsList": [],
92
+ "saveCheckpointInterval": 0,
93
+ "logDir": "/mnt/hungpv/projects/SAM_Variant/logs",
94
+ "logSubDir": "TinyImageNet_MongeSAM_ViT_rho0.075_lr0.0001",
95
+ "epochs": 20,
96
+ "contin": false
97
+ }
TinyImageNet_MongeSAM_ViT_rho0.075_lr0.0001/terminal.log ADDED
@@ -0,0 +1,24 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ┏━━━━━━━╸S╺╸T╺╸A╺╸T╺╸S╺━━━━━━━┳━━━━━━━━━━━━━━━━━T╺╸R╺╸A╺╸I╺╸N ━━━━━━━━━━━━━━━━┳━━━━━━━━━━T╺╸E╺╸S╺╸T ━━━━━━━━━━┓
2
+ ┃ ┃ ┃ ┃
3
+ ┃ epoch │ time ┃ learningRate │ loss │ accuracy ┃ loss │ accuracy ┃
4
+ ┠──────────────┼──────────────╂───────────────┼───────────────┼───────────────╂───────────────┼───────────────┨
5
+ ┃ 1 │ 06:47 min ┃ 4.988e-05 │ 4.4699 │ 2.98% ┃ 3.0921 │ 22.68% ┃
6
+ ┃ 2 │ 05:21 min ┃ 9.857e-05 │ 1.3560 │ 65.49% ┃ 0.6739 │ 83.48% ┃
7
+ ┃ 3 │ 05:22 min ┃ 9.615e-05 │ 0.7467 │ 81.03% ┃ 0.5441 │ 86.12% ┃
8
+ ┃ 4 │ 05:22 min ┃ 9.259e-05 │ 0.6203 │ 83.81% ┃ 0.4857 │ 87.53% ┃
9
+ ┃ 5 │ 05:22 min ┃ 8.798e-05 │ 0.5459 │ 85.71% ┃ 0.4536 │ 87.92% ┃
10
+ ┃ 6 │ 05:22 min ┃ 8.244e-05 │ 0.4971 │ 86.87% ┃ 0.4348 │ 88.35% ┃
11
+ ┃ 7 │ 05:22 min ┃ 7.610e-05 │ 0.4597 │ 87.77% ┃ 0.4216 │ 88.84% ┃
12
+ ┃ 8 │ 05:22 min ┃ 6.911e-05 │ 0.4311 │ 88.51% ┃ 0.4081 │ 88.96% ┃
13
+ ┃ 9 │ 05:22 min ┃ 6.166e-05 │ 0.4108 │ 89.09% ┃ 0.3989 │ 89.09% ┃
14
+ ┃ 10 │ 05:22 min ┃ 5.392e-05 │ 0.3880 │ 89.73% ┃ 0.3924 │ 89.52% ┃
15
+ ┃ 11 │ 05:22 min ┃ 4.608e-05 │ 0.3737 │ 90.09% ┃ 0.3877 │ 89.44% ┃
16
+ ┃ 12 │ 05:22 min ┃ 3.834e-05 │ 0.3640 │ 90.38% ┃ 0.3828 │ 89.62% ┃
17
+ ┃ 13 │ 05:22 min ┃ 3.088e-05 │ 0.3522 │ 90.69% ┃ 0.3806 │ 89.66% ┃
18
+ ┃ 14 │ 05:22 min ┃ 2.390e-05 │ 0.3426 │ 91.07% ┃ 0.3781 │ 89.72% ┃
19
+ ┃ 15 │ 05:22 min ┃ 1.756e-05 │ 0.3359 │ 91.21% ┃ 0.3756 │ 89.85% ┃
20
+ ┃ 16 │ 05:22 min ┃ 1.202e-05 │ 0.3304 │ 91.47% ┃ 0.3751 │ 89.79% ┃
21
+ ┃ 17 │ 05:22 min ┃ 7.410e-06 │ 0.3283 │ 91.54% ┃ 0.3740 │ 89.83% ┃
22
+ ┃ 18 │ 05:22 min ┃ 3.853e-06 │ 0.3247 │ 91.53% ┃ 0.3738 │ 89.86% ┃
23
+ ┃ 19 │ 05:22 min ┃ 1.431e-06 │ 0.3227 │ 91.61% ┃ 0.3737 │ 89.86% ┃
24
+ ┃ 20 │ 05:22 min ┃ 2.052e-07 │ 0.3242 │ 91.57% ┃ 0.3737 │ 89.85% ┃
TinyImageNet_MongeSAM_ViT_rho0.075_lr0.0001_run01/data.hdf5 ADDED
Binary file (56 kB). View file
 
TinyImageNet_MongeSAM_ViT_rho0.075_lr0.0001_run01/final.model ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:6339dfb07c1f3c690bdf81c411bc85f7b363999ed11a48b6e3de596810a6e75c
3
+ size 1035057354
TinyImageNet_MongeSAM_ViT_rho0.075_lr0.0001_run01/params.json ADDED
@@ -0,0 +1,97 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "widthFactor": 4,
3
+ "ViTPatchsize": 16,
4
+ "ViTSize": "B",
5
+ "pyramidAlpha": 270,
6
+ "pyramidBottleneck": true,
7
+ "model": "ViT",
8
+ "dropout": 0.0,
9
+ "BN": true,
10
+ "depth": 16,
11
+ "label_smoothing": 0.1,
12
+ "ESAMbeta": 0.5,
13
+ "ESAMgamma": 0.7,
14
+ "lookSAMk": 5,
15
+ "lookSAMalpha": 0.7,
16
+ "lookSAMLayerwise": false,
17
+ "optimizer": "MongeSAM",
18
+ "weightDecay": 0.001,
19
+ "momentum": 0.9,
20
+ "rho": 0.01,
21
+ "samRho": 0.01,
22
+ "asamRho": 1.0,
23
+ "esamRho": 0.075,
24
+ "looksamRho": 0.05,
25
+ "friendlysamRho": 0.075,
26
+ "fishersamRho": 0.1,
27
+ "mongesamRho": 0.075,
28
+ "gsamRho": 0.075,
29
+ "bayesiansamRho": 0.075,
30
+ "msamRho": 0.075,
31
+ "adamwSamRho": 0.075,
32
+ "adamwMsamRho": 0.075,
33
+ "adamwFriendlysamRho": 0.075,
34
+ "sigma": 1.0,
35
+ "lmbda": 0.95,
36
+ "fisherEta": 0.1,
37
+ "gsamAlpha": 0.1,
38
+ "gsamAdaptive": false,
39
+ "bayesMSharpness": 8,
40
+ "bayesDelta": 10.0,
41
+ "bayesBeta1": 0.9,
42
+ "bayesBeta2": 0.999,
43
+ "bayesSInit": 1.0,
44
+ "bayesDamping": 1e-08,
45
+ "bayesEps": 1e-12,
46
+ "nesterov": false,
47
+ "grad_clip_norm": 0.0,
48
+ "dataThreads": 4,
49
+ "dataDir": "/home/hungpv/.datasets",
50
+ "batchSize": 64,
51
+ "dataset": "TinyImageNet",
52
+ "imageSize": 64,
53
+ "flip": true,
54
+ "crop": true,
55
+ "cut": false,
56
+ "cutoutProp": 0.5,
57
+ "randAugment": true,
58
+ "randAugment_magnitude": 10,
59
+ "mixup": false,
60
+ "mixupProp": 0.2,
61
+ "normalize": "std",
62
+ "truncate": false,
63
+ "verbose": false,
64
+ "metrics": [
65
+ "learningRate",
66
+ "loss",
67
+ "accuracy"
68
+ ],
69
+ "logEach": 1,
70
+ "rndSeed": false,
71
+ "deterministic": false,
72
+ "learningRate": 0.0001,
73
+ "lrScheduler": [
74
+ "startRamp",
75
+ "cos"
76
+ ],
77
+ "LRScheduler_step_steps": 3.0,
78
+ "LRScheduler_step_gamma": 0.2,
79
+ "LRScheduler_exp_maxDecay": 0.01,
80
+ "LRScheduler_cos_periods": 0.5,
81
+ "LRScheduler_StartRamp_epochs": 1.0,
82
+ "LRScheduler_WRN_T0": 50.0,
83
+ "LRScheduler_WRN_Tmult": 1.2,
84
+ "rhoScheduler": [
85
+ "const"
86
+ ],
87
+ "rhoScheduler_StartRamp_epochs": 8.0,
88
+ "rhoScheduler_StartJump_epochs": 8.0,
89
+ "saveCheckpoint": true,
90
+ "keepLastCheckpoint": false,
91
+ "checkpointsList": [],
92
+ "saveCheckpointInterval": 0,
93
+ "logDir": "/mnt/hungpv/projects/SAM_Variant/logs",
94
+ "logSubDir": "TinyImageNet_MongeSAM_ViT_rho0.075_lr0.0001_run01",
95
+ "epochs": 20,
96
+ "contin": false
97
+ }
TinyImageNet_MongeSAM_ViT_rho0.075_lr0.0001_run01/terminal.log ADDED
@@ -0,0 +1,24 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ┏━━━━━━━╸S╺╸T╺╸A╺╸T╺╸S╺━━━━━━━┳━━━━━━━━━━━━━━━━━T╺╸R╺╸A╺╸I╺╸N ━━━━━━━━━━━━━━━━┳━━━━━━━━━━T╺╸E╺╸S╺╸T ━━━━━━━━━━┓
2
+ ┃ ┃ ┃ ┃
3
+ ┃ epoch │ time ┃ learningRate │ loss │ accuracy ┃ loss │ accuracy ┃
4
+ ┠──────────────┼──────────────╂───────────────┼───────────────┼───────────────╂───────────────┼───────────────┨
5
+ ┃ 1 │ 44:55 min ┃ 4.988e-05 │ 4.4800 │ 2.10% ┃ 3.8151 │ 8.30% ┃
6
+ ┃ 2 │ 50:58 min ┃ 9.857e-05 │ 2.2448 │ 41.36% ┃ 1.0051 │ 75.02% ┃
7
+ ┃ 3 │ 51:47 min ┃ 9.615e-05 │ 0.7864 │ 80.85% ┃ 0.5522 │ 87.27% ┃
8
+ ┃ 4 │ 51:52 min ┃ 9.259e-05 │ 0.5521 │ 86.51% ┃ 0.4635 │ 88.78% ┃
9
+ ┃ 5 │ 51:52 min ┃ 8.798e-05 │ 0.4574 │ 88.73% ┃ 0.4195 │ 89.57% ┃
10
+ ┃ 6 │ 51:46 min ┃ 8.244e-05 │ 0.3973 │ 89.93% ┃ 0.3898 │ 90.35% ┃
11
+ ┃ 7 │ 51:10 min ┃ 7.610e-05 │ 0.3552 │ 91.00% ┃ 0.3795 │ 90.26% ┃
12
+ ┃ 8 │ 50:58 min ┃ 6.911e-05 │ 0.3264 │ 91.71% ┃ 0.3635 │ 90.81% ┃
13
+ ┃ 9 │ 51:51 min ┃ 6.166e-05 │ 0.3016 │ 92.39% ┃ 0.3571 │ 90.86% ┃
14
+ ┃ 10 │ 51:47 min ┃ 5.392e-05 │ 0.2789 │ 93.05% ┃ 0.3500 │ 90.82% ┃
15
+ ┃ 11 │ 51:52 min ┃ 4.608e-05 │ 0.2624 │ 93.48% ┃ 0.3486 │ 90.88% ┃
16
+ ┃ 12 │ 51:52 min ┃ 3.834e-05 │ 0.2490 │ 93.85% ┃ 0.3423 │ 91.12% ┃
17
+ ┃ 13 │ 51:06 min ┃ 3.088e-05 │ 0.2373 │ 94.22% ┃ 0.3399 │ 91.16% ┃
18
+ ┃ 14 │ 50:59 min ┃ 2.390e-05 │ 0.2286 │ 94.49% ┃ 0.3399 │ 91.23% ┃
19
+ ┃ 15 │ 51:52 min ┃ 1.756e-05 │ 0.2232 │ 94.59% ┃ 0.3370 │ 91.31% ┃
20
+ ┃ 16 │ 51:51 min ┃ 1.202e-05 │ 0.2180 │ 94.80% ┃ 0.3352 │ 91.30% ┃
21
+ ┃ 17 │ 51:49 min ┃ 7.410e-06 │ 0.2133 │ 94.96% ┃ 0.3351 │ 91.32% ┃
22
+ ┃ 18 │ 51:52 min ┃ 3.853e-06 │ 0.2112 │ 95.01% ┃ 0.3349 │ 91.27% ┃
23
+ ┃ 19 │ 51:51 min ┃ 1.431e-06 │ 0.2112 │ 94.95% ┃ 0.3347 │ 91.35% ┃
24
+ ┃ 20 │ 50:16 min ┃ 2.052e-07 │ 0.2104 │ 95.02% ┃ 0.3346 │ 91.37% ┃
TinyImageNet_SAM_ViT_rho0.01_lr0.0001/data.hdf5 ADDED
Binary file (56 kB). View file
 
TinyImageNet_SAM_ViT_rho0.01_lr0.0001/final.model ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:c6a6c848957188694ee3935910378532f4dbe2d5d624a103115d73fdcce1f6fd
3
+ size 1035057418