diff --git a/TinyImageNet_ASAM_ViT_rho1_lr0.0001/data.hdf5 b/TinyImageNet_ASAM_ViT_rho1_lr0.0001/data.hdf5 new file mode 100644 index 0000000000000000000000000000000000000000..34633a8f2b2047fdab61e5add69cfc511670b4ec Binary files /dev/null and b/TinyImageNet_ASAM_ViT_rho1_lr0.0001/data.hdf5 differ diff --git a/TinyImageNet_ASAM_ViT_rho1_lr0.0001/final.model b/TinyImageNet_ASAM_ViT_rho1_lr0.0001/final.model new file mode 100644 index 0000000000000000000000000000000000000000..09666d21cde6e055c341704a5fc3b6c12ef5c63b --- /dev/null +++ b/TinyImageNet_ASAM_ViT_rho1_lr0.0001/final.model @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:43388b81453a8abaecb6e795e80d4cf573e87a09efe6b1fefcd63f7c38b8ac6e +size 1052309706 diff --git a/TinyImageNet_ASAM_ViT_rho1_lr0.0001/params.json b/TinyImageNet_ASAM_ViT_rho1_lr0.0001/params.json new file mode 100644 index 0000000000000000000000000000000000000000..761e24930435d5ecc92ef048359a49ed77dd4fc8 --- /dev/null +++ b/TinyImageNet_ASAM_ViT_rho1_lr0.0001/params.json @@ -0,0 +1,97 @@ +{ + "widthFactor": 4, + "ViTPatchsize": 32, + "ViTSize": "B", + "pyramidAlpha": 270, + "pyramidBottleneck": true, + "model": "ViT", + "dropout": 0.0, + "BN": true, + "depth": 16, + "label_smoothing": 0.1, + "ESAMbeta": 0.5, + "ESAMgamma": 0.7, + "lookSAMk": 5, + "lookSAMalpha": 0.7, + "lookSAMLayerwise": false, + "optimizer": "ASAM", + "weightDecay": 0.001, + "momentum": 0.9, + "rho": 0.075, + "samRho": 0.075, + "asamRho": 1.0, + "esamRho": 0.075, + "looksamRho": 0.05, + "friendlysamRho": 0.075, + "fishersamRho": 0.5, + "mongesamRho": 0.075, + "gsamRho": 0.075, + "bayesiansamRho": 0.075, + "msamRho": 0.075, + "adamwSamRho": 0.075, + "adamwMsamRho": 0.075, + "adamwFriendlysamRho": 0.075, + "sigma": 1.0, + "lmbda": 0.95, + "fisherEta": 0.1, + "gsamAlpha": 0.1, + "gsamAdaptive": false, + "bayesMSharpness": 8, + "bayesDelta": 10.0, + "bayesBeta1": 0.9, + "bayesBeta2": 0.999, + "bayesSInit": 1.0, + "bayesDamping": 1e-08, + "bayesEps": 1e-12, + "nesterov": false, + "grad_clip_norm": 0.0, + "dataThreads": 4, + "dataDir": "/home/hungpv/.datasets", + "batchSize": 64, + "dataset": "TinyImageNet", + "imageSize": 64, + "flip": true, + "crop": true, + "cut": false, + "cutoutProp": 0.5, + "randAugment": true, + "randAugment_magnitude": 10, + "mixup": false, + "mixupProp": 0.2, + "normalize": "std", + "truncate": false, + "verbose": false, + "metrics": [ + "learningRate", + "loss", + "accuracy" + ], + "logEach": 1, + "rndSeed": false, + "deterministic": false, + "learningRate": 0.0001, + "lrScheduler": [ + "startRamp", + "cos" + ], + "LRScheduler_step_steps": 3.0, + "LRScheduler_step_gamma": 0.2, + "LRScheduler_exp_maxDecay": 0.01, + "LRScheduler_cos_periods": 0.5, + "LRScheduler_StartRamp_epochs": 1.0, + "LRScheduler_WRN_T0": 50.0, + "LRScheduler_WRN_Tmult": 1.2, + "rhoScheduler": [ + "const" + ], + "rhoScheduler_StartRamp_epochs": 8.0, + "rhoScheduler_StartJump_epochs": 8.0, + "saveCheckpoint": true, + "keepLastCheckpoint": false, + "checkpointsList": [], + "saveCheckpointInterval": 0, + "logDir": "/mnt/hungpv/projects/SAM_Variant/logs", + "logSubDir": "TinyImageNet_ASAM_ViT_rho1_lr0.0001", + "epochs": 20, + "contin": false +} \ No newline at end of file diff --git a/TinyImageNet_ASAM_ViT_rho1_lr0.0001/terminal.log b/TinyImageNet_ASAM_ViT_rho1_lr0.0001/terminal.log new file mode 100644 index 0000000000000000000000000000000000000000..6a76e7e808dffefe95025c7e3db0ba16ef477e1d --- /dev/null +++ b/TinyImageNet_ASAM_ViT_rho1_lr0.0001/terminal.log @@ -0,0 +1,24 @@ +┏━━━━━━━╸S╺╸T╺╸A╺╸T╺╸S╺━━━━━━━┳━━━━━━━━━━━━━━━━━T╺╸R╺╸A╺╸I╺╸N ━━━━━━━━━━━━━━━━┳━━━━━━━━━━T╺╸E╺╸S╺╸T ━━━━━━━━━━┓ +┃ ┃ ┃ ┃ +┃ epoch │ time ┃ learningRate │ loss │ accuracy ┃ loss │ accuracy ┃ +┠──────────────┼──────────────╂───────────────┼───────────────┼───────────────╂───────────────┼───────────────┨ +┃ 1 │ 15:08 min ┃ 4.988e-05 │ 4.3452 │ 5.17% ┃ 2.3462 │ 39.43% ┃ +┃ 2 │ 05:55 min ┃ 9.857e-05 │ 1.1988 │ 69.40% ┃ 0.6596 │ 83.70% ┃ +┃ 3 │ 05:25 min ┃ 9.615e-05 │ 0.7342 │ 81.34% ┃ 0.5396 │ 86.11% ┃ +┃ 4 │ 05:25 min ┃ 9.259e-05 │ 0.6144 │ 83.96% ┃ 0.4840 │ 87.68% ┃ +┃ 5 │ 05:25 min ┃ 8.798e-05 │ 0.5418 │ 85.76% ┃ 0.4530 │ 88.05% ┃ +┃ 6 │ 05:25 min ┃ 8.244e-05 │ 0.4930 │ 86.97% ┃ 0.4349 │ 88.37% ┃ +┃ 7 │ 05:25 min ┃ 7.610e-05 │ 0.4560 │ 87.80% ┃ 0.4204 │ 88.77% ┃ +┃ 8 │ 05:25 min ┃ 6.911e-05 │ 0.4271 │ 88.56% ┃ 0.4071 │ 89.13% ┃ +┃ 9 │ 05:25 min ┃ 6.166e-05 │ 0.4063 │ 89.15% ┃ 0.3985 │ 89.26% ┃ +┃ 10 │ 05:24 min ┃ 5.392e-05 │ 0.3829 │ 89.78% ┃ 0.3921 │ 89.48% ┃ +┃ 11 │ 05:25 min ┃ 4.608e-05 │ 0.3683 │ 90.19% ┃ 0.3877 │ 89.35% ┃ +┃ 12 │ 05:25 min ┃ 3.834e-05 │ 0.3582 │ 90.48% ┃ 0.3833 │ 89.62% ┃ +┃ 13 │ 05:26 min ┃ 3.088e-05 │ 0.3458 │ 90.79% ┃ 0.3806 │ 89.71% ┃ +┃ 14 │ 05:25 min ┃ 2.390e-05 │ 0.3356 │ 91.23% ┃ 0.3787 │ 89.79% ┃ +┃ 15 │ 05:25 min ┃ 1.756e-05 │ 0.3289 │ 91.34% ┃ 0.3766 │ 89.75% ┃ +┃ 16 │ 05:25 min ┃ 1.202e-05 │ 0.3231 │ 91.57% ┃ 0.3757 │ 89.79% ┃ +┃ 17 │ 05:25 min ┃ 7.410e-06 │ 0.3209 │ 91.66% ┃ 0.3747 │ 89.85% ┃ +┃ 18 │ 05:25 min ┃ 3.853e-06 │ 0.3172 │ 91.67% ┃ 0.3743 │ 89.86% ┃ +┃ 19 │ 05:25 min ┃ 1.431e-06 │ 0.3147 │ 91.75% ┃ 0.3743 │ 89.84% ┃ +┃ 20 │ 05:25 min ┃ 2.052e-07 │ 0.3165 │ 91.70% ┃ 0.3743 │ 89.86% ┃ diff --git a/TinyImageNet_ASAM_ViT_rho1_lr0.0001_run01/data.hdf5 b/TinyImageNet_ASAM_ViT_rho1_lr0.0001_run01/data.hdf5 new file mode 100644 index 0000000000000000000000000000000000000000..85050f9674d699eb15eb9a2ce69708b9e7921707 Binary files /dev/null and b/TinyImageNet_ASAM_ViT_rho1_lr0.0001_run01/data.hdf5 differ diff --git a/TinyImageNet_ASAM_ViT_rho1_lr0.0001_run01/final.model b/TinyImageNet_ASAM_ViT_rho1_lr0.0001_run01/final.model new file mode 100644 index 0000000000000000000000000000000000000000..222413be1e649938ff8606d50072fde6c9532251 --- /dev/null +++ b/TinyImageNet_ASAM_ViT_rho1_lr0.0001_run01/final.model @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:281c4b6f2a9db3d74d8992be7ea32537f0bd908fdb00fba11bb629391c4c95b1 +size 1035057354 diff --git a/TinyImageNet_ASAM_ViT_rho1_lr0.0001_run01/params.json b/TinyImageNet_ASAM_ViT_rho1_lr0.0001_run01/params.json new file mode 100644 index 0000000000000000000000000000000000000000..6667ca6d5861b07a80a6c2995da18e059f392942 --- /dev/null +++ b/TinyImageNet_ASAM_ViT_rho1_lr0.0001_run01/params.json @@ -0,0 +1,97 @@ +{ + "widthFactor": 4, + "ViTPatchsize": 16, + "ViTSize": "B", + "pyramidAlpha": 270, + "pyramidBottleneck": true, + "model": "ViT", + "dropout": 0.0, + "BN": true, + "depth": 16, + "label_smoothing": 0.1, + "ESAMbeta": 0.5, + "ESAMgamma": 0.7, + "lookSAMk": 5, + "lookSAMalpha": 0.7, + "lookSAMLayerwise": false, + "optimizer": "ASAM", + "weightDecay": 0.001, + "momentum": 0.9, + "rho": 0.01, + "samRho": 0.01, + "asamRho": 1.0, + "esamRho": 0.075, + "looksamRho": 0.05, + "friendlysamRho": 0.075, + "fishersamRho": 0.1, + "mongesamRho": 0.075, + "gsamRho": 0.075, + "bayesiansamRho": 0.075, + "msamRho": 0.075, + "adamwSamRho": 0.075, + "adamwMsamRho": 0.075, + "adamwFriendlysamRho": 0.075, + "sigma": 1.0, + "lmbda": 0.95, + "fisherEta": 0.1, + "gsamAlpha": 0.1, + "gsamAdaptive": false, + "bayesMSharpness": 8, + "bayesDelta": 10.0, + "bayesBeta1": 0.9, + "bayesBeta2": 0.999, + "bayesSInit": 1.0, + "bayesDamping": 1e-08, + "bayesEps": 1e-12, + "nesterov": false, + "grad_clip_norm": 0.0, + "dataThreads": 4, + "dataDir": "/home/hungpv/.datasets", + "batchSize": 64, + "dataset": "TinyImageNet", + "imageSize": 64, + "flip": true, + "crop": true, + "cut": false, + "cutoutProp": 0.5, + "randAugment": true, + "randAugment_magnitude": 10, + "mixup": false, + "mixupProp": 0.2, + "normalize": "std", + "truncate": false, + "verbose": false, + "metrics": [ + "learningRate", + "loss", + "accuracy" + ], + "logEach": 1, + "rndSeed": false, + "deterministic": false, + "learningRate": 0.0001, + "lrScheduler": [ + "startRamp", + "cos" + ], + "LRScheduler_step_steps": 3.0, + "LRScheduler_step_gamma": 0.2, + "LRScheduler_exp_maxDecay": 0.01, + "LRScheduler_cos_periods": 0.5, + "LRScheduler_StartRamp_epochs": 1.0, + "LRScheduler_WRN_T0": 50.0, + "LRScheduler_WRN_Tmult": 1.2, + "rhoScheduler": [ + "const" + ], + "rhoScheduler_StartRamp_epochs": 8.0, + "rhoScheduler_StartJump_epochs": 8.0, + "saveCheckpoint": true, + "keepLastCheckpoint": false, + "checkpointsList": [], + "saveCheckpointInterval": 0, + "logDir": "/mnt/hungpv/projects/SAM_Variant/logs", + "logSubDir": "TinyImageNet_ASAM_ViT_rho1_lr0.0001_run01", + "epochs": 20, + "contin": false +} \ No newline at end of file diff --git a/TinyImageNet_ASAM_ViT_rho1_lr0.0001_run01/terminal.log b/TinyImageNet_ASAM_ViT_rho1_lr0.0001_run01/terminal.log new file mode 100644 index 0000000000000000000000000000000000000000..24d74d4a441557df60934c02318630322525e396 --- /dev/null +++ b/TinyImageNet_ASAM_ViT_rho1_lr0.0001_run01/terminal.log @@ -0,0 +1,24 @@ +┏━━━━━━━╸S╺╸T╺╸A╺╸T╺╸S╺━━━━━━━┳━━━━━━━━━━━━━━━━━T╺╸R╺╸A╺╸I╺╸N ━━━━━━━━━━━━━━━━┳━━━━━━━━━━T╺╸E╺╸S╺╸T ━━━━━━━━━━┓ +┃ ┃ ┃ ┃ +┃ epoch │ time ┃ learningRate │ loss │ accuracy ┃ loss │ accuracy ┃ +┠──────────────┼──────────────╂───────────────┼───────────────┼───────────────╂───────────────┼───────────────┨ +┃ 1 │ 51:26 min ┃ 4.988e-05 │ 4.4725 │ 2.21% ┃ 3.7916 │ 8.80% ┃ +┃ 2 │ 51:12 min ┃ 9.857e-05 │ 2.2201 │ 42.02% ┃ 0.9877 │ 75.58% ┃ +┃ 3 │ 51:08 min ┃ 9.615e-05 │ 0.7781 │ 81.00% ┃ 0.5503 │ 87.21% ┃ +┃ 4 │ 51:12 min ┃ 9.259e-05 │ 0.5495 │ 86.57% ┃ 0.4607 │ 88.97% ┃ +┃ 5 │ 51:12 min ┃ 8.798e-05 │ 0.4560 │ 88.74% ┃ 0.4179 │ 89.66% ┃ +┃ 6 │ 51:12 min ┃ 8.244e-05 │ 0.3962 │ 89.98% ┃ 0.3885 │ 90.43% ┃ +┃ 7 │ 49:34 min ┃ 7.610e-05 │ 0.3544 │ 90.98% ┃ 0.3783 │ 90.34% ┃ +┃ 8 │ 51:15 min ┃ 6.911e-05 │ 0.3253 │ 91.74% ┃ 0.3635 │ 90.64% ┃ +┃ 9 │ 51:15 min ┃ 6.166e-05 │ 0.3009 │ 92.35% ┃ 0.3565 │ 90.88% ┃ +┃ 10 │ 51:15 min ┃ 5.392e-05 │ 0.2783 │ 93.08% ┃ 0.3499 │ 90.95% ┃ +┃ 11 │ 51:10 min ┃ 4.608e-05 │ 0.2615 │ 93.49% ┃ 0.3491 │ 90.77% ┃ +┃ 12 │ 51:15 min ┃ 3.834e-05 │ 0.2483 │ 93.88% ┃ 0.3416 │ 91.13% ┃ +┃ 13 │ 49:38 min ┃ 3.088e-05 │ 0.2364 │ 94.21% ┃ 0.3404 │ 91.11% ┃ +┃ 14 │ 51:10 min ┃ 2.390e-05 │ 0.2280 │ 94.49% ┃ 0.3402 │ 91.18% ┃ +┃ 15 │ 51:06 min ┃ 1.756e-05 │ 0.2224 │ 94.64% ┃ 0.3373 │ 91.30% ┃ +┃ 16 │ 51:11 min ┃ 1.202e-05 │ 0.2175 │ 94.77% ┃ 0.3357 │ 91.28% ┃ +┃ 17 │ 51:10 min ┃ 7.410e-06 │ 0.2128 │ 94.94% ┃ 0.3353 │ 91.19% ┃ +┃ 18 │ 51:06 min ┃ 3.853e-06 │ 0.2106 │ 94.99% ┃ 0.3351 │ 91.31% ┃ +┃ 19 │ 49:35 min ┃ 1.431e-06 │ 0.2105 │ 94.97% ┃ 0.3350 │ 91.28% ┃ +┃ 20 │ 51:12 min ┃ 2.052e-07 │ 0.2098 │ 95.02% ┃ 0.3350 │ 91.27% ┃ diff --git a/TinyImageNet_ESAM_ViT_rho0.075_lr0.0001/data.hdf5 b/TinyImageNet_ESAM_ViT_rho0.075_lr0.0001/data.hdf5 new file mode 100644 index 0000000000000000000000000000000000000000..ecafe96a356c74789b35227f625af3e562264473 Binary files /dev/null and b/TinyImageNet_ESAM_ViT_rho0.075_lr0.0001/data.hdf5 differ diff --git a/TinyImageNet_ESAM_ViT_rho0.075_lr0.0001/final.model b/TinyImageNet_ESAM_ViT_rho0.075_lr0.0001/final.model new file mode 100644 index 0000000000000000000000000000000000000000..6b160709a7bd613a6d532398e596a5a8d09f536b --- /dev/null +++ b/TinyImageNet_ESAM_ViT_rho0.075_lr0.0001/final.model @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:0330934be4e4769f381f039cf54f331344e1c89fd68dc06326333495a46a5de5 +size 1052309770 diff --git a/TinyImageNet_ESAM_ViT_rho0.075_lr0.0001/params.json b/TinyImageNet_ESAM_ViT_rho0.075_lr0.0001/params.json new file mode 100644 index 0000000000000000000000000000000000000000..240d830ec730ad02a7428dbeb05ed6d6bd1bf341 --- /dev/null +++ b/TinyImageNet_ESAM_ViT_rho0.075_lr0.0001/params.json @@ -0,0 +1,97 @@ +{ + "widthFactor": 4, + "ViTPatchsize": 32, + "ViTSize": "B", + "pyramidAlpha": 270, + "pyramidBottleneck": true, + "model": "ViT", + "dropout": 0.0, + "BN": true, + "depth": 16, + "label_smoothing": 0.1, + "ESAMbeta": 0.5, + "ESAMgamma": 0.7, + "lookSAMk": 5, + "lookSAMalpha": 0.7, + "lookSAMLayerwise": false, + "optimizer": "ESAM", + "weightDecay": 0.001, + "momentum": 0.9, + "rho": 0.075, + "samRho": 0.075, + "asamRho": 1.0, + "esamRho": 0.075, + "looksamRho": 0.05, + "friendlysamRho": 0.075, + "fishersamRho": 0.5, + "mongesamRho": 0.075, + "gsamRho": 0.075, + "bayesiansamRho": 0.075, + "msamRho": 0.075, + "adamwSamRho": 0.075, + "adamwMsamRho": 0.075, + "adamwFriendlysamRho": 0.075, + "sigma": 1.0, + "lmbda": 0.95, + "fisherEta": 0.1, + "gsamAlpha": 0.1, + "gsamAdaptive": false, + "bayesMSharpness": 8, + "bayesDelta": 10.0, + "bayesBeta1": 0.9, + "bayesBeta2": 0.999, + "bayesSInit": 1.0, + "bayesDamping": 1e-08, + "bayesEps": 1e-12, + "nesterov": false, + "grad_clip_norm": 0.0, + "dataThreads": 4, + "dataDir": "/home/hungpv/.datasets", + "batchSize": 64, + "dataset": "TinyImageNet", + "imageSize": 64, + "flip": true, + "crop": true, + "cut": false, + "cutoutProp": 0.5, + "randAugment": true, + "randAugment_magnitude": 10, + "mixup": false, + "mixupProp": 0.2, + "normalize": "std", + "truncate": false, + "verbose": false, + "metrics": [ + "learningRate", + "loss", + "accuracy" + ], + "logEach": 1, + "rndSeed": false, + "deterministic": false, + "learningRate": 0.0001, + "lrScheduler": [ + "startRamp", + "cos" + ], + "LRScheduler_step_steps": 3.0, + "LRScheduler_step_gamma": 0.2, + "LRScheduler_exp_maxDecay": 0.01, + "LRScheduler_cos_periods": 0.5, + "LRScheduler_StartRamp_epochs": 1.0, + "LRScheduler_WRN_T0": 50.0, + "LRScheduler_WRN_Tmult": 1.2, + "rhoScheduler": [ + "const" + ], + "rhoScheduler_StartRamp_epochs": 8.0, + "rhoScheduler_StartJump_epochs": 8.0, + "saveCheckpoint": true, + "keepLastCheckpoint": false, + "checkpointsList": [], + "saveCheckpointInterval": 0, + "logDir": "/mnt/hungpv/projects/SAM_Variant/logs", + "logSubDir": "TinyImageNet_ESAM_ViT_rho0.075_lr0.0001", + "epochs": 20, + "contin": false +} \ No newline at end of file diff --git a/TinyImageNet_ESAM_ViT_rho0.075_lr0.0001/terminal.log b/TinyImageNet_ESAM_ViT_rho0.075_lr0.0001/terminal.log new file mode 100644 index 0000000000000000000000000000000000000000..63616bb04dcac177d1d9466fafa546b611250977 --- /dev/null +++ b/TinyImageNet_ESAM_ViT_rho0.075_lr0.0001/terminal.log @@ -0,0 +1,24 @@ +┏━━━━━━━╸S╺╸T╺╸A╺╸T╺╸S╺━━━━━━━┳━━━━━━━━━━━━━━━━━T╺╸R╺╸A╺╸I╺╸N ━━━━━━━━━━━━━━━━┳━━━━━━━━━━T╺╸E╺╸S╺╸T ━━━━━━━━━━┓ +┃ ┃ ┃ ┃ +┃ epoch │ time ┃ learningRate │ loss │ accuracy ┃ loss │ accuracy ┃ +┠──────────────┼──────────────╂───────────────┼───────────────┼───────────────╂───────────────┼───────────────┨ +┃ 1 │ 07:13 min ┃ 4.988e-05 │ 4.6157 │ 0.95% ┃ 4.3750 │ 1.83% ┃ +┃ 2 │ 05:29 min ┃ 9.857e-05 │ 4.1348 │ 5.88% ┃ 3.0892 │ 25.36% ┃ +┃ 3 │ 05:29 min ┃ 9.615e-05 │ 1.7850 │ 56.97% ┃ 0.8597 │ 79.93% ┃ +┃ 4 │ 05:29 min ┃ 9.259e-05 │ 0.9346 │ 77.68% ┃ 0.6752 │ 84.21% ┃ +┃ 5 │ 05:29 min ┃ 8.798e-05 │ 0.7858 │ 81.36% ┃ 0.6099 │ 85.58% ┃ +┃ 6 │ 05:29 min ┃ 8.244e-05 │ 0.7125 │ 83.05% ┃ 0.5693 │ 86.36% ┃ +┃ 7 │ 05:29 min ┃ 7.610e-05 │ 0.6619 │ 84.13% ┃ 0.5423 │ 86.92% ┃ +┃ 8 │ 05:29 min ┃ 6.911e-05 │ 0.6264 │ 84.97% ┃ 0.5236 │ 87.15% ┃ +┃ 9 │ 05:29 min ┃ 6.166e-05 │ 0.5986 │ 85.68% ┃ 0.5094 │ 87.71% ┃ +┃ 10 │ 05:29 min ┃ 5.392e-05 │ 0.5746 │ 86.34% ┃ 0.4975 │ 87.93% ┃ +┃ 11 │ 05:29 min ┃ 4.608e-05 │ 0.5602 │ 86.65% ┃ 0.4906 │ 88.11% ┃ +┃ 12 │ 05:29 min ┃ 3.834e-05 │ 0.5490 │ 87.02% ┃ 0.4838 │ 88.09% ┃ +┃ 13 │ 05:29 min ┃ 3.088e-05 │ 0.5370 │ 87.22% ┃ 0.4793 │ 88.47% ┃ +┃ 14 │ 05:29 min ┃ 2.390e-05 │ 0.5276 │ 87.46% ┃ 0.4751 │ 88.50% ┃ +┃ 15 │ 05:29 min ┃ 1.756e-05 │ 0.5229 │ 87.61% ┃ 0.4721 │ 88.49% ┃ +┃ 16 │ 05:29 min ┃ 1.202e-05 │ 0.5173 │ 87.77% ┃ 0.4703 │ 88.62% ┃ +┃ 17 │ 05:29 min ┃ 7.410e-06 │ 0.5143 │ 87.85% ┃ 0.4691 │ 88.63% ┃ +┃ 18 │ 05:29 min ┃ 3.853e-06 │ 0.5099 │ 87.98% ┃ 0.4684 │ 88.61% ┃ +┃ 19 │ 05:29 min ┃ 1.431e-06 │ 0.5082 │ 87.99% ┃ 0.4680 │ 88.58% ┃ +┃ 20 │ 05:29 min ┃ 2.052e-07 │ 0.5101 │ 87.88% ┃ 0.4680 │ 88.58% ┃ diff --git a/TinyImageNet_ESAM_ViT_rho0.075_lr0.0001_run01/data.hdf5 b/TinyImageNet_ESAM_ViT_rho0.075_lr0.0001_run01/data.hdf5 new file mode 100644 index 0000000000000000000000000000000000000000..e64cc65112e554d12ebd3d4ea4b4872764393485 Binary files /dev/null and b/TinyImageNet_ESAM_ViT_rho0.075_lr0.0001_run01/data.hdf5 differ diff --git a/TinyImageNet_ESAM_ViT_rho0.075_lr0.0001_run01/final.model b/TinyImageNet_ESAM_ViT_rho0.075_lr0.0001_run01/final.model new file mode 100644 index 0000000000000000000000000000000000000000..c2fc5fec683a574c2f765729fc66cf5cc39017b5 --- /dev/null +++ b/TinyImageNet_ESAM_ViT_rho0.075_lr0.0001_run01/final.model @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:b8e598e3248fe2f779bf5fff0505f60dee159dc1433b3a569190e160a47d128b +size 1035057418 diff --git a/TinyImageNet_ESAM_ViT_rho0.075_lr0.0001_run01/params.json b/TinyImageNet_ESAM_ViT_rho0.075_lr0.0001_run01/params.json new file mode 100644 index 0000000000000000000000000000000000000000..4a1626cbea6cf8afcb85cbba0d83af331d782de8 --- /dev/null +++ b/TinyImageNet_ESAM_ViT_rho0.075_lr0.0001_run01/params.json @@ -0,0 +1,97 @@ +{ + "widthFactor": 4, + "ViTPatchsize": 16, + "ViTSize": "B", + "pyramidAlpha": 270, + "pyramidBottleneck": true, + "model": "ViT", + "dropout": 0.0, + "BN": true, + "depth": 16, + "label_smoothing": 0.1, + "ESAMbeta": 0.5, + "ESAMgamma": 0.7, + "lookSAMk": 5, + "lookSAMalpha": 0.7, + "lookSAMLayerwise": false, + "optimizer": "ESAM", + "weightDecay": 0.001, + "momentum": 0.9, + "rho": 0.01, + "samRho": 0.01, + "asamRho": 1.0, + "esamRho": 0.075, + "looksamRho": 0.05, + "friendlysamRho": 0.075, + "fishersamRho": 0.1, + "mongesamRho": 0.075, + "gsamRho": 0.075, + "bayesiansamRho": 0.075, + "msamRho": 0.075, + "adamwSamRho": 0.075, + "adamwMsamRho": 0.075, + "adamwFriendlysamRho": 0.075, + "sigma": 1.0, + "lmbda": 0.95, + "fisherEta": 0.1, + "gsamAlpha": 0.1, + "gsamAdaptive": false, + "bayesMSharpness": 8, + "bayesDelta": 10.0, + "bayesBeta1": 0.9, + "bayesBeta2": 0.999, + "bayesSInit": 1.0, + "bayesDamping": 1e-08, + "bayesEps": 1e-12, + "nesterov": false, + "grad_clip_norm": 0.0, + "dataThreads": 4, + "dataDir": "/home/hungpv/.datasets", + "batchSize": 64, + "dataset": "TinyImageNet", + "imageSize": 64, + "flip": true, + "crop": true, + "cut": false, + "cutoutProp": 0.5, + "randAugment": true, + "randAugment_magnitude": 10, + "mixup": false, + "mixupProp": 0.2, + "normalize": "std", + "truncate": false, + "verbose": false, + "metrics": [ + "learningRate", + "loss", + "accuracy" + ], + "logEach": 1, + "rndSeed": false, + "deterministic": false, + "learningRate": 0.0001, + "lrScheduler": [ + "startRamp", + "cos" + ], + "LRScheduler_step_steps": 3.0, + "LRScheduler_step_gamma": 0.2, + "LRScheduler_exp_maxDecay": 0.01, + "LRScheduler_cos_periods": 0.5, + "LRScheduler_StartRamp_epochs": 1.0, + "LRScheduler_WRN_T0": 50.0, + "LRScheduler_WRN_Tmult": 1.2, + "rhoScheduler": [ + "const" + ], + "rhoScheduler_StartRamp_epochs": 8.0, + "rhoScheduler_StartJump_epochs": 8.0, + "saveCheckpoint": true, + "keepLastCheckpoint": false, + "checkpointsList": [], + "saveCheckpointInterval": 0, + "logDir": "/mnt/hungpv/projects/SAM_Variant/logs", + "logSubDir": "TinyImageNet_ESAM_ViT_rho0.075_lr0.0001_run01", + "epochs": 20, + "contin": false +} \ No newline at end of file diff --git a/TinyImageNet_ESAM_ViT_rho0.075_lr0.0001_run01/terminal.log b/TinyImageNet_ESAM_ViT_rho0.075_lr0.0001_run01/terminal.log new file mode 100644 index 0000000000000000000000000000000000000000..9ef034266b3c21acfb8fc2376781580261f0cb5c --- /dev/null +++ b/TinyImageNet_ESAM_ViT_rho0.075_lr0.0001_run01/terminal.log @@ -0,0 +1,24 @@ +┏━━━━━━━╸S╺╸T╺╸A╺╸T╺╸S╺━━━━━━━┳━━━━━━━━━━━━━━━━━T╺╸R╺╸A╺╸I╺╸N ━━━━━━━━━━━━━━━━┳━━━━━━━━━━T╺╸E╺╸S╺╸T ━━━━━━━━━━┓ +┃ ┃ ┃ ┃ +┃ epoch │ time ┃ learningRate │ loss │ accuracy ┃ loss │ accuracy ┃ +┠──────────────┼──────────────╂───────────────┼───────────────┼───────────────╂───────────────┼───────────────┨ +┃ 1 │ 53:44 min ┃ 4.988e-05 │ 4.5522 │ 1.14% ┃ 4.3090 │ 2.98% ┃ +┃ 2 │ 51:58 min ┃ 9.857e-05 │ 2.9266 │ 28.66% ┃ 0.8639 │ 79.41% ┃ +┃ 3 │ 51:56 min ┃ 9.615e-05 │ 0.6978 │ 83.23% ┃ 0.4965 │ 88.62% ┃ +┃ 4 │ 51:57 min ┃ 9.259e-05 │ 0.5044 │ 87.68% ┃ 0.4274 │ 89.98% ┃ +┃ 5 │ 51:54 min ┃ 8.798e-05 │ 0.4224 │ 89.60% ┃ 0.3921 │ 90.48% ┃ +┃ 6 │ 50:23 min ┃ 8.244e-05 │ 0.3705 │ 90.82% ┃ 0.3697 │ 90.86% ┃ +┃ 7 │ 51:53 min ┃ 7.610e-05 │ 0.3315 │ 91.92% ┃ 0.3562 │ 91.17% ┃ +┃ 8 │ 48:21 min ┃ 6.911e-05 │ 0.3052 │ 92.60% ┃ 0.3452 │ 91.23% ┃ +┃ 9 │ 52:02 min ┃ 6.166e-05 │ 0.2831 │ 93.18% ┃ 0.3395 │ 91.34% ┃ +┃ 10 │ 52:02 min ┃ 5.392e-05 │ 0.2629 │ 93.86% ┃ 0.3328 │ 91.53% ┃ +┃ 11 │ 51:57 min ┃ 4.608e-05 │ 0.2480 │ 94.29% ┃ 0.3309 │ 91.38% ┃ +┃ 12 │ 52:01 min ┃ 3.834e-05 │ 0.2357 │ 94.64% ┃ 0.3245 │ 91.65% ┃ +┃ 13 │ 52:02 min ┃ 3.088e-05 │ 0.2265 │ 94.97% ┃ 0.3227 │ 91.58% ┃ +┃ 14 │ 50:19 min ┃ 2.390e-05 │ 0.2179 │ 95.21% ┃ 0.3219 │ 91.51% ┃ +┃ 15 │ 51:56 min ┃ 1.756e-05 │ 0.2133 │ 95.31% ┃ 0.3197 │ 91.65% ┃ +┃ 16 │ 51:56 min ┃ 1.202e-05 │ 0.2093 │ 95.41% ┃ 0.3181 │ 91.61% ┃ +┃ 17 │ 51:51 min ┃ 7.410e-06 │ 0.2053 │ 95.61% ┃ 0.3178 │ 91.72% ┃ +┃ 18 │ 51:56 min ┃ 3.853e-06 │ 0.2033 │ 95.64% ┃ 0.3174 │ 91.70% ┃ +┃ 19 │ 51:56 min ┃ 1.431e-06 │ 0.2027 │ 95.65% ┃ 0.3172 │ 91.68% ┃ +┃ 20 │ 50:20 min ┃ 2.052e-07 │ 0.2027 │ 95.69% ┃ 0.3172 │ 91.68% ┃ diff --git a/TinyImageNet_FisherSAM_ViT_rho0.01_lr0.0001/data.hdf5 b/TinyImageNet_FisherSAM_ViT_rho0.01_lr0.0001/data.hdf5 new file mode 100644 index 0000000000000000000000000000000000000000..7fef1c31b06c0a8ed04a2a36327fdc4508d56487 Binary files /dev/null and b/TinyImageNet_FisherSAM_ViT_rho0.01_lr0.0001/data.hdf5 differ diff --git a/TinyImageNet_FisherSAM_ViT_rho0.01_lr0.0001/final.model b/TinyImageNet_FisherSAM_ViT_rho0.01_lr0.0001/final.model new file mode 100644 index 0000000000000000000000000000000000000000..96cd2b7a9f7e0d093634dfc0b7bdba6737e2d896 --- /dev/null +++ b/TinyImageNet_FisherSAM_ViT_rho0.01_lr0.0001/final.model @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:9e6220c14d5cc65441d221b0e5bf9eb8d79d68282db3cdbe1e9cda7bd67eb3db +size 1052309770 diff --git a/TinyImageNet_FisherSAM_ViT_rho0.01_lr0.0001/params.json b/TinyImageNet_FisherSAM_ViT_rho0.01_lr0.0001/params.json new file mode 100644 index 0000000000000000000000000000000000000000..c25581d081776db6131caf1ddbcf0fa50d8fc811 --- /dev/null +++ b/TinyImageNet_FisherSAM_ViT_rho0.01_lr0.0001/params.json @@ -0,0 +1,97 @@ +{ + "widthFactor": 4, + "ViTPatchsize": 32, + "ViTSize": "B", + "pyramidAlpha": 270, + "pyramidBottleneck": true, + "model": "ViT", + "dropout": 0.0, + "BN": true, + "depth": 16, + "label_smoothing": 0.1, + "ESAMbeta": 0.5, + "ESAMgamma": 0.7, + "lookSAMk": 5, + "lookSAMalpha": 0.7, + "lookSAMLayerwise": false, + "optimizer": "FisherSAM", + "weightDecay": 0.001, + "momentum": 0.9, + "rho": 0.01, + "samRho": 0.01, + "asamRho": 1.0, + "esamRho": 0.075, + "looksamRho": 0.05, + "friendlysamRho": 0.01, + "fishersamRho": 0.01, + "mongesamRho": 0.075, + "gsamRho": 0.01, + "bayesiansamRho": 0.075, + "msamRho": 0.075, + "adamwSamRho": 0.075, + "adamwMsamRho": 0.075, + "adamwFriendlysamRho": 0.075, + "sigma": 1.0, + "lmbda": 0.95, + "fisherEta": 0.1, + "gsamAlpha": 0.1, + "gsamAdaptive": false, + "bayesMSharpness": 8, + "bayesDelta": 10.0, + "bayesBeta1": 0.9, + "bayesBeta2": 0.999, + "bayesSInit": 1.0, + "bayesDamping": 1e-08, + "bayesEps": 1e-12, + "nesterov": false, + "grad_clip_norm": 0.0, + "dataThreads": 4, + "dataDir": "/home/hungpv/.datasets", + "batchSize": 64, + "dataset": "TinyImageNet", + "imageSize": 64, + "flip": true, + "crop": true, + "cut": false, + "cutoutProp": 0.5, + "randAugment": true, + "randAugment_magnitude": 10, + "mixup": false, + "mixupProp": 0.2, + "normalize": "std", + "truncate": false, + "verbose": false, + "metrics": [ + "learningRate", + "loss", + "accuracy" + ], + "logEach": 1, + "rndSeed": false, + "deterministic": false, + "learningRate": 0.0001, + "lrScheduler": [ + "startRamp", + "cos" + ], + "LRScheduler_step_steps": 3.0, + "LRScheduler_step_gamma": 0.2, + "LRScheduler_exp_maxDecay": 0.01, + "LRScheduler_cos_periods": 0.5, + "LRScheduler_StartRamp_epochs": 1.0, + "LRScheduler_WRN_T0": 50.0, + "LRScheduler_WRN_Tmult": 1.2, + "rhoScheduler": [ + "const" + ], + "rhoScheduler_StartRamp_epochs": 8.0, + "rhoScheduler_StartJump_epochs": 8.0, + "saveCheckpoint": true, + "keepLastCheckpoint": false, + "checkpointsList": [], + "saveCheckpointInterval": 0, + "logDir": "/mnt/hungpv/projects/SAM_Variant/logs", + "logSubDir": "TinyImageNet_FisherSAM_ViT_rho0.01_lr0.0001", + "epochs": 20, + "contin": false +} \ No newline at end of file diff --git a/TinyImageNet_FisherSAM_ViT_rho0.01_lr0.0001/terminal.log b/TinyImageNet_FisherSAM_ViT_rho0.01_lr0.0001/terminal.log new file mode 100644 index 0000000000000000000000000000000000000000..17a73b83f323593da99da2178bc3ad25f52adeb0 --- /dev/null +++ b/TinyImageNet_FisherSAM_ViT_rho0.01_lr0.0001/terminal.log @@ -0,0 +1,24 @@ +┏━━━━━━━╸S╺╸T╺╸A╺╸T╺╸S╺━━━━━━━┳━━━━━━━━━━━━━━━━━T╺╸R╺╸A╺╸I╺╸N ━━━━━━━━━━━━━━━━┳━━━━━━━━━━T╺╸E╺╸S╺╸T ━━━━━━━━━━┓ +┃ ┃ ┃ ┃ +┃ epoch │ time ┃ learningRate │ loss │ accuracy ┃ loss │ accuracy ┃ +┠──────────────┼──────────────╂───────────────┼───────────────┼───────────────╂───────────────┼───────────────┨ +┃ 1 │ 14:09 min ┃ 4.988e-05 │ 4.4779 │ 2.81% ┃ 3.1837 │ 20.95% ┃ +┃ 2 │ 12:56 min ┃ 9.857e-05 │ 1.3964 │ 64.50% ┃ 0.6817 │ 83.22% ┃ +┃ 3 │ 12:54 min ┃ 9.615e-05 │ 0.7532 │ 80.85% ┃ 0.5447 │ 86.20% ┃ +┃ 4 │ 12:49 min ┃ 9.259e-05 │ 0.6257 │ 83.73% ┃ 0.4867 │ 87.41% ┃ +┃ 5 │ 12:56 min ┃ 8.798e-05 │ 0.5528 │ 85.56% ┃ 0.4560 │ 88.12% ┃ +┃ 6 │ 12:56 min ┃ 8.244e-05 │ 0.5053 │ 86.71% ┃ 0.4357 │ 88.30% ┃ +┃ 7 │ 12:55 min ┃ 7.610e-05 │ 0.4692 │ 87.55% ┃ 0.4223 │ 88.93% ┃ +┃ 8 │ 12:50 min ┃ 6.911e-05 │ 0.4416 │ 88.33% ┃ 0.4103 │ 88.99% ┃ +┃ 9 │ 12:54 min ┃ 6.166e-05 │ 0.4218 │ 88.86% ┃ 0.4009 │ 89.21% ┃ +┃ 10 │ 12:55 min ┃ 5.392e-05 │ 0.3999 │ 89.48% ┃ 0.3937 │ 89.42% ┃ +┃ 11 │ 12:54 min ┃ 4.608e-05 │ 0.3863 │ 89.85% ┃ 0.3891 │ 89.46% ┃ +┃ 12 │ 12:53 min ┃ 3.834e-05 │ 0.3769 │ 90.14% ┃ 0.3843 │ 89.66% ┃ +┃ 13 │ 12:56 min ┃ 3.088e-05 │ 0.3656 │ 90.46% ┃ 0.3819 │ 89.68% ┃ +┃ 14 │ 12:56 min ┃ 2.390e-05 │ 0.3563 │ 90.80% ┃ 0.3795 │ 89.76% ┃ +┃ 15 │ 12:49 min ┃ 1.756e-05 │ 0.3502 │ 90.87% ┃ 0.3772 │ 89.72% ┃ +┃ 16 │ 12:55 min ┃ 1.202e-05 │ 0.3446 │ 91.11% ┃ 0.3765 │ 89.76% ┃ +┃ 17 │ 12:54 min ┃ 7.410e-06 │ 0.3426 │ 91.18% ┃ 0.3755 │ 89.84% ┃ +┃ 18 │ 12:54 min ┃ 3.853e-06 │ 0.3392 │ 91.26% ┃ 0.3753 │ 89.88% ┃ +┃ 19 │ 12:49 min ┃ 1.431e-06 │ 0.3372 │ 91.28% ┃ 0.3752 │ 89.81% ┃ +┃ 20 │ 12:56 min ┃ 2.052e-07 │ 0.3387 │ 91.26% ┃ 0.3752 │ 89.82% ┃ diff --git a/TinyImageNet_FisherSAM_ViT_rho0.1_lr0.0001/data.hdf5 b/TinyImageNet_FisherSAM_ViT_rho0.1_lr0.0001/data.hdf5 new file mode 100644 index 0000000000000000000000000000000000000000..b2d98b5482a1243daf7823b2acc3a864b5adcd3d Binary files /dev/null and b/TinyImageNet_FisherSAM_ViT_rho0.1_lr0.0001/data.hdf5 differ diff --git a/TinyImageNet_FisherSAM_ViT_rho0.1_lr0.0001/final.model b/TinyImageNet_FisherSAM_ViT_rho0.1_lr0.0001/final.model new file mode 100644 index 0000000000000000000000000000000000000000..f60faab2f7ec52ed63e2e0e59d382e4149d02f20 --- /dev/null +++ b/TinyImageNet_FisherSAM_ViT_rho0.1_lr0.0001/final.model @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:cd56d668ce71acd516bb6e51596c8baf4400ad20ecfeabe60485c0a8012de4c3 +size 1035057418 diff --git a/TinyImageNet_FisherSAM_ViT_rho0.1_lr0.0001/params.json b/TinyImageNet_FisherSAM_ViT_rho0.1_lr0.0001/params.json new file mode 100644 index 0000000000000000000000000000000000000000..7f821827a62539ee0eaad013c5f21ffdc2577fdf --- /dev/null +++ b/TinyImageNet_FisherSAM_ViT_rho0.1_lr0.0001/params.json @@ -0,0 +1,97 @@ +{ + "widthFactor": 4, + "ViTPatchsize": 16, + "ViTSize": "B", + "pyramidAlpha": 270, + "pyramidBottleneck": true, + "model": "ViT", + "dropout": 0.0, + "BN": true, + "depth": 16, + "label_smoothing": 0.1, + "ESAMbeta": 0.5, + "ESAMgamma": 0.7, + "lookSAMk": 5, + "lookSAMalpha": 0.7, + "lookSAMLayerwise": false, + "optimizer": "FisherSAM", + "weightDecay": 0.001, + "momentum": 0.9, + "rho": 0.01, + "samRho": 0.01, + "asamRho": 1.0, + "esamRho": 0.075, + "looksamRho": 0.05, + "friendlysamRho": 0.075, + "fishersamRho": 0.1, + "mongesamRho": 0.075, + "gsamRho": 0.075, + "bayesiansamRho": 0.075, + "msamRho": 0.075, + "adamwSamRho": 0.075, + "adamwMsamRho": 0.075, + "adamwFriendlysamRho": 0.075, + "sigma": 1.0, + "lmbda": 0.95, + "fisherEta": 0.1, + "gsamAlpha": 0.1, + "gsamAdaptive": false, + "bayesMSharpness": 8, + "bayesDelta": 10.0, + "bayesBeta1": 0.9, + "bayesBeta2": 0.999, + "bayesSInit": 1.0, + "bayesDamping": 1e-08, + "bayesEps": 1e-12, + "nesterov": false, + "grad_clip_norm": 0.0, + "dataThreads": 4, + "dataDir": "/home/hungpv/.datasets", + "batchSize": 64, + "dataset": "TinyImageNet", + "imageSize": 64, + "flip": true, + "crop": true, + "cut": false, + "cutoutProp": 0.5, + "randAugment": true, + "randAugment_magnitude": 10, + "mixup": false, + "mixupProp": 0.2, + "normalize": "std", + "truncate": false, + "verbose": false, + "metrics": [ + "learningRate", + "loss", + "accuracy" + ], + "logEach": 1, + "rndSeed": false, + "deterministic": false, + "learningRate": 0.0001, + "lrScheduler": [ + "startRamp", + "cos" + ], + "LRScheduler_step_steps": 3.0, + "LRScheduler_step_gamma": 0.2, + "LRScheduler_exp_maxDecay": 0.01, + "LRScheduler_cos_periods": 0.5, + "LRScheduler_StartRamp_epochs": 1.0, + "LRScheduler_WRN_T0": 50.0, + "LRScheduler_WRN_Tmult": 1.2, + "rhoScheduler": [ + "const" + ], + "rhoScheduler_StartRamp_epochs": 8.0, + "rhoScheduler_StartJump_epochs": 8.0, + "saveCheckpoint": true, + "keepLastCheckpoint": false, + "checkpointsList": [], + "saveCheckpointInterval": 0, + "logDir": "/mnt/hungpv/projects/SAM_Variant/logs", + "logSubDir": "TinyImageNet_FisherSAM_ViT_rho0.1_lr0.0001", + "epochs": 20, + "contin": false +} \ No newline at end of file diff --git a/TinyImageNet_FisherSAM_ViT_rho0.1_lr0.0001/terminal.log b/TinyImageNet_FisherSAM_ViT_rho0.1_lr0.0001/terminal.log new file mode 100644 index 0000000000000000000000000000000000000000..aadf1dce7c262753c6eb3ae66543441b9fde3a95 --- /dev/null +++ b/TinyImageNet_FisherSAM_ViT_rho0.1_lr0.0001/terminal.log @@ -0,0 +1,24 @@ +┏━━━━━━━╸S╺╸T╺╸A╺╸T╺╸S╺━━━━━━━┳━━━━━━━━━━━━━━━━━T╺╸R╺╸A╺╸I╺╸N ━━━━━━━━━━━━━━━━┳━━━━━━━━━━T╺╸E╺╸S╺╸T ━━━━━━━━━━┓ +┃ ┃ ┃ ┃ +┃ epoch │ time ┃ learningRate │ loss │ accuracy ┃ loss │ accuracy ┃ +┠──────────────┼──────────────╂───────────────┼───────────────┼───────────────╂───────────────┼───────────────┨ +┃ 1 │ 53:23 min ┃ 4.988e-05 │ 4.5657 │ 1.25% ┃ 4.2563 │ 3.66% ┃ +┃ 2 │ 52:14 min ┃ 9.857e-05 │ 2.6575 │ 34.31% ┃ 0.8503 │ 79.59% ┃ +┃ 3 │ 52:08 min ┃ 9.615e-05 │ 0.7120 │ 82.78% ┃ 0.5081 │ 88.13% ┃ +┃ 4 │ 52:13 min ┃ 9.259e-05 │ 0.5254 │ 87.04% ┃ 0.4358 │ 89.64% ┃ +┃ 5 │ 52:12 min ┃ 8.798e-05 │ 0.4458 │ 88.96% ┃ 0.3995 │ 90.05% ┃ +┃ 6 │ 50:22 min ┃ 8.244e-05 │ 0.3959 │ 90.06% ┃ 0.3753 │ 90.60% ┃ +┃ 7 │ 52:11 min ┃ 7.610e-05 │ 0.3591 │ 91.03% ┃ 0.3616 │ 90.93% ┃ +┃ 8 │ 52:14 min ┃ 6.911e-05 │ 0.3339 │ 91.67% ┃ 0.3500 │ 91.12% ┃ +┃ 9 │ 52:10 min ┃ 6.166e-05 │ 0.3135 │ 92.16% ┃ 0.3431 │ 91.25% ┃ +┃ 10 │ 52:15 min ┃ 5.392e-05 │ 0.2944 │ 92.78% ┃ 0.3361 │ 91.41% ┃ +┃ 11 │ 52:15 min ┃ 4.608e-05 │ 0.2808 │ 93.16% ┃ 0.3339 │ 91.30% ┃ +┃ 12 │ 50:39 min ┃ 3.834e-05 │ 0.2687 │ 93.43% ┃ 0.3284 │ 91.62% ┃ +┃ 13 │ 52:10 min ┃ 3.088e-05 │ 0.2600 │ 93.68% ┃ 0.3259 │ 91.51% ┃ +┃ 14 │ 52:14 min ┃ 2.390e-05 │ 0.2521 │ 94.01% ┃ 0.3243 │ 91.61% ┃ +┃ 15 │ 52:10 min ┃ 1.756e-05 │ 0.2479 │ 94.09% ┃ 0.3221 │ 91.70% ┃ +┃ 16 │ 52:15 min ┃ 1.202e-05 │ 0.2435 │ 94.19% ┃ 0.3207 │ 91.68% ┃ +┃ 17 │ 52:15 min ┃ 7.410e-06 │ 0.2403 │ 94.34% ┃ 0.3201 │ 91.72% ┃ +┃ 18 │ 50:39 min ┃ 3.853e-06 │ 0.2383 │ 94.34% ┃ 0.3198 │ 91.72% ┃ +┃ 19 │ 52:09 min ┃ 1.431e-06 │ 0.2379 │ 94.33% ┃ 0.3197 │ 91.75% ┃ +┃ 20 │ 52:14 min ┃ 2.052e-07 │ 0.2380 │ 94.45% ┃ 0.3197 │ 91.73% ┃ diff --git a/TinyImageNet_FriendlySAM_ViT_rho0.01_lr0.0001/data.hdf5 b/TinyImageNet_FriendlySAM_ViT_rho0.01_lr0.0001/data.hdf5 new file mode 100644 index 0000000000000000000000000000000000000000..28d6bdc6a64fef3215215eeaeef29defa59eda50 Binary files /dev/null and b/TinyImageNet_FriendlySAM_ViT_rho0.01_lr0.0001/data.hdf5 differ diff --git a/TinyImageNet_FriendlySAM_ViT_rho0.01_lr0.0001/final.model b/TinyImageNet_FriendlySAM_ViT_rho0.01_lr0.0001/final.model new file mode 100644 index 0000000000000000000000000000000000000000..c20538da172f92bff3441dd55cae2b2f1e73fa8e --- /dev/null +++ b/TinyImageNet_FriendlySAM_ViT_rho0.01_lr0.0001/final.model @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:3149cf5bd989bcf9c3d204a93be8bde2bd89b1aa65d349ecf8925f860aea6e86 +size 1403071210 diff --git a/TinyImageNet_FriendlySAM_ViT_rho0.01_lr0.0001/params.json b/TinyImageNet_FriendlySAM_ViT_rho0.01_lr0.0001/params.json new file mode 100644 index 0000000000000000000000000000000000000000..911f7d43cc25a3436c66bdcc233fcebc76d7a2af --- /dev/null +++ b/TinyImageNet_FriendlySAM_ViT_rho0.01_lr0.0001/params.json @@ -0,0 +1,97 @@ +{ + "widthFactor": 4, + "ViTPatchsize": 32, + "ViTSize": "B", + "pyramidAlpha": 270, + "pyramidBottleneck": true, + "model": "ViT", + "dropout": 0.0, + "BN": true, + "depth": 16, + "label_smoothing": 0.1, + "ESAMbeta": 0.5, + "ESAMgamma": 0.7, + "lookSAMk": 5, + "lookSAMalpha": 0.7, + "lookSAMLayerwise": false, + "optimizer": "FriendlySAM", + "weightDecay": 0.001, + "momentum": 0.9, + "rho": 0.01, + "samRho": 0.01, + "asamRho": 1.0, + "esamRho": 0.075, + "looksamRho": 0.05, + "friendlysamRho": 0.01, + "fishersamRho": 0.01, + "mongesamRho": 0.075, + "gsamRho": 0.01, + "bayesiansamRho": 0.075, + "msamRho": 0.075, + "adamwSamRho": 0.075, + "adamwMsamRho": 0.075, + "adamwFriendlysamRho": 0.075, + "sigma": 1.0, + "lmbda": 0.95, + "fisherEta": 0.1, + "gsamAlpha": 0.1, + "gsamAdaptive": false, + "bayesMSharpness": 8, + "bayesDelta": 10.0, + "bayesBeta1": 0.9, + "bayesBeta2": 0.999, + "bayesSInit": 1.0, + "bayesDamping": 1e-08, + "bayesEps": 1e-12, + "nesterov": false, + "grad_clip_norm": 0.0, + "dataThreads": 4, + "dataDir": "/home/hungpv/.datasets", + "batchSize": 64, + "dataset": "TinyImageNet", + "imageSize": 64, + "flip": true, + "crop": true, + "cut": false, + "cutoutProp": 0.5, + "randAugment": true, + "randAugment_magnitude": 10, + "mixup": false, + "mixupProp": 0.2, + "normalize": "std", + "truncate": false, + "verbose": false, + "metrics": [ + "learningRate", + "loss", + "accuracy" + ], + "logEach": 1, + "rndSeed": false, + "deterministic": false, + "learningRate": 0.0001, + "lrScheduler": [ + "startRamp", + "cos" + ], + "LRScheduler_step_steps": 3.0, + "LRScheduler_step_gamma": 0.2, + "LRScheduler_exp_maxDecay": 0.01, + "LRScheduler_cos_periods": 0.5, + "LRScheduler_StartRamp_epochs": 1.0, + "LRScheduler_WRN_T0": 50.0, + "LRScheduler_WRN_Tmult": 1.2, + "rhoScheduler": [ + "const" + ], + "rhoScheduler_StartRamp_epochs": 8.0, + "rhoScheduler_StartJump_epochs": 8.0, + "saveCheckpoint": true, + "keepLastCheckpoint": false, + "checkpointsList": [], + "saveCheckpointInterval": 0, + "logDir": "/mnt/hungpv/projects/SAM_Variant/logs", + "logSubDir": "TinyImageNet_FriendlySAM_ViT_rho0.01_lr0.0001", + "epochs": 20, + "contin": false +} \ No newline at end of file diff --git a/TinyImageNet_FriendlySAM_ViT_rho0.01_lr0.0001/terminal.log b/TinyImageNet_FriendlySAM_ViT_rho0.01_lr0.0001/terminal.log new file mode 100644 index 0000000000000000000000000000000000000000..0e6d8140244bdf6b635e267243a568d06ff625d1 --- /dev/null +++ b/TinyImageNet_FriendlySAM_ViT_rho0.01_lr0.0001/terminal.log @@ -0,0 +1,24 @@ +┏━━━━━━━╸S╺╸T╺╸A╺╸T╺╸S╺━━━━━━━┳━━━━━━━━━━━━━━━━━T╺╸R╺╸A╺╸I╺╸N ━━━━━━━━━━━━━━━━┳━━━━━━━━━━T╺╸E╺╸S╺╸T ━━━━━━━━━━┓ +┃ ┃ ┃ ┃ +┃ epoch │ time ┃ learningRate │ loss │ accuracy ┃ loss │ accuracy ┃ +┠──────────────┼──────────────╂───────────────┼───────────────┼───────────────╂───────────────┼───────────────┨ +┃ 1 │ 13:46 min ┃ 4.988e-05 │ 4.4879 │ 2.64% ┃ 3.2766 │ 19.22% ┃ +┃ 2 │ 12:35 min ┃ 9.857e-05 │ 1.4256 │ 63.77% ┃ 0.6835 │ 83.11% ┃ +┃ 3 │ 12:34 min ┃ 9.615e-05 │ 0.7552 │ 80.83% ┃ 0.5452 │ 86.15% ┃ +┃ 4 │ 12:29 min ┃ 9.259e-05 │ 0.6267 │ 83.71% ┃ 0.4870 │ 87.44% ┃ +┃ 5 │ 12:34 min ┃ 8.798e-05 │ 0.5533 │ 85.57% ┃ 0.4559 │ 88.09% ┃ +┃ 6 │ 12:34 min ┃ 8.244e-05 │ 0.5055 │ 86.72% ┃ 0.4356 │ 88.37% ┃ +┃ 7 │ 12:35 min ┃ 7.610e-05 │ 0.4694 │ 87.55% ┃ 0.4219 │ 88.88% ┃ +┃ 8 │ 12:29 min ┃ 6.911e-05 │ 0.4417 │ 88.30% ┃ 0.4099 │ 88.96% ┃ +┃ 9 │ 12:34 min ┃ 6.166e-05 │ 0.4218 │ 88.86% ┃ 0.4008 │ 89.23% ┃ +┃ 10 │ 12:34 min ┃ 5.392e-05 │ 0.3999 │ 89.47% ┃ 0.3935 │ 89.36% ┃ +┃ 11 │ 12:34 min ┃ 4.608e-05 │ 0.3863 │ 89.82% ┃ 0.3889 │ 89.49% ┃ +┃ 12 │ 12:28 min ┃ 3.834e-05 │ 0.3768 │ 90.17% ┃ 0.3841 │ 89.68% ┃ +┃ 13 │ 12:34 min ┃ 3.088e-05 │ 0.3655 │ 90.44% ┃ 0.3817 │ 89.69% ┃ +┃ 14 │ 12:34 min ┃ 2.390e-05 │ 0.3562 │ 90.80% ┃ 0.3792 │ 89.78% ┃ +┃ 15 │ 12:34 min ┃ 1.756e-05 │ 0.3501 │ 90.87% ┃ 0.3769 │ 89.73% ┃ +┃ 16 │ 12:29 min ┃ 1.202e-05 │ 0.3445 │ 91.10% ┃ 0.3762 │ 89.80% ┃ +┃ 17 │ 12:34 min ┃ 7.410e-06 │ 0.3425 │ 91.22% ┃ 0.3753 │ 89.81% ┃ +┃ 18 │ 12:34 min ┃ 3.853e-06 │ 0.3391 │ 91.27% ┃ 0.3750 │ 89.83% ┃ +┃ 19 │ 12:34 min ┃ 1.431e-06 │ 0.3371 │ 91.30% ┃ 0.3749 │ 89.84% ┃ +┃ 20 │ 12:29 min ┃ 2.052e-07 │ 0.3386 │ 91.25% ┃ 0.3749 │ 89.85% ┃ diff --git a/TinyImageNet_GSAM_ViT_rho0.01_lr0.0001/data.hdf5 b/TinyImageNet_GSAM_ViT_rho0.01_lr0.0001/data.hdf5 new file mode 100644 index 0000000000000000000000000000000000000000..57ac1a6a70f1e00efc5ca5cab19f847d1103ce5c Binary files /dev/null and b/TinyImageNet_GSAM_ViT_rho0.01_lr0.0001/data.hdf5 differ diff --git a/TinyImageNet_GSAM_ViT_rho0.01_lr0.0001/final.model b/TinyImageNet_GSAM_ViT_rho0.01_lr0.0001/final.model new file mode 100644 index 0000000000000000000000000000000000000000..60664c3666919b5a96200c586e99354c51d97f49 --- /dev/null +++ b/TinyImageNet_GSAM_ViT_rho0.01_lr0.0001/final.model @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:9e83b3f1029f585242c97fbc10ad47b5034880b3ae1c82f4e8b1d2e52104584a +size 1753832650 diff --git a/TinyImageNet_GSAM_ViT_rho0.01_lr0.0001/params.json b/TinyImageNet_GSAM_ViT_rho0.01_lr0.0001/params.json new file mode 100644 index 0000000000000000000000000000000000000000..00229bf0f730f65cb411329d25766e50957ca5c1 --- /dev/null +++ b/TinyImageNet_GSAM_ViT_rho0.01_lr0.0001/params.json @@ -0,0 +1,97 @@ +{ + "widthFactor": 4, + "ViTPatchsize": 32, + "ViTSize": "B", + "pyramidAlpha": 270, + "pyramidBottleneck": true, + "model": "ViT", + "dropout": 0.0, + "BN": true, + "depth": 16, + "label_smoothing": 0.1, + "ESAMbeta": 0.5, + "ESAMgamma": 0.7, + "lookSAMk": 5, + "lookSAMalpha": 0.7, + "lookSAMLayerwise": false, + "optimizer": "GSAM", + "weightDecay": 0.001, + "momentum": 0.9, + "rho": 0.01, + "samRho": 0.01, + "asamRho": 1.0, + "esamRho": 0.075, + "looksamRho": 0.05, + "friendlysamRho": 0.01, + "fishersamRho": 0.01, + "mongesamRho": 0.075, + "gsamRho": 0.01, + "bayesiansamRho": 0.075, + "msamRho": 0.075, + "adamwSamRho": 0.075, + "adamwMsamRho": 0.075, + "adamwFriendlysamRho": 0.075, + "sigma": 1.0, + "lmbda": 0.95, + "fisherEta": 0.1, + "gsamAlpha": 0.1, + "gsamAdaptive": false, + "bayesMSharpness": 8, + "bayesDelta": 10.0, + "bayesBeta1": 0.9, + "bayesBeta2": 0.999, + "bayesSInit": 1.0, + "bayesDamping": 1e-08, + "bayesEps": 1e-12, + "nesterov": false, + "grad_clip_norm": 0.0, + "dataThreads": 4, + "dataDir": "/home/hungpv/.datasets", + "batchSize": 64, + "dataset": "TinyImageNet", + "imageSize": 64, + "flip": true, + "crop": true, + "cut": false, + "cutoutProp": 0.5, + "randAugment": true, + "randAugment_magnitude": 10, + "mixup": false, + "mixupProp": 0.2, + "normalize": "std", + "truncate": false, + "verbose": false, + "metrics": [ + "learningRate", + "loss", + "accuracy" + ], + "logEach": 1, + "rndSeed": false, + "deterministic": false, + "learningRate": 0.0001, + "lrScheduler": [ + "startRamp", + "cos" + ], + "LRScheduler_step_steps": 3.0, + "LRScheduler_step_gamma": 0.2, + "LRScheduler_exp_maxDecay": 0.01, + "LRScheduler_cos_periods": 0.5, + "LRScheduler_StartRamp_epochs": 1.0, + "LRScheduler_WRN_T0": 50.0, + "LRScheduler_WRN_Tmult": 1.2, + "rhoScheduler": [ + "const" + ], + "rhoScheduler_StartRamp_epochs": 8.0, + "rhoScheduler_StartJump_epochs": 8.0, + "saveCheckpoint": true, + "keepLastCheckpoint": false, + "checkpointsList": [], + "saveCheckpointInterval": 0, + "logDir": "/mnt/hungpv/projects/SAM_Variant/logs", + "logSubDir": "TinyImageNet_GSAM_ViT_rho0.01_lr0.0001", + "epochs": 20, + "contin": false +} \ No newline at end of file diff --git a/TinyImageNet_GSAM_ViT_rho0.01_lr0.0001/terminal.log b/TinyImageNet_GSAM_ViT_rho0.01_lr0.0001/terminal.log new file mode 100644 index 0000000000000000000000000000000000000000..78aa007a0bb6f8286c1115bbdd8984fdb75829a4 --- /dev/null +++ b/TinyImageNet_GSAM_ViT_rho0.01_lr0.0001/terminal.log @@ -0,0 +1,24 @@ +┏━━━━━━━╸S╺╸T╺╸A╺╸T╺╸S╺━━━━━━━┳━━━━━━━━━━━━━━━━━T╺╸R╺╸A╺╸I╺╸N ━━━━━━━━━━━━━━━━┳━━━━━━━━━━T╺╸E╺╸S╺╸T ━━━━━━━━━━┓ +┃ ┃ ┃ ┃ +┃ epoch │ time ┃ learningRate │ loss │ accuracy ┃ loss │ accuracy ┃ +┠──────────────┼──────────────╂───────────────┼───────────────┼───────────────╂───────────────┼───────────────┨ +┃ 1 │ 14:24 min ┃ 4.988e-05 │ 4.4898 │ 2.64% ┃ 3.2798 │ 19.23% ┃ +┃ 2 │ 13:15 min ┃ 9.857e-05 │ 1.4294 │ 63.69% ┃ 0.6861 │ 83.01% ┃ +┃ 3 │ 13:09 min ┃ 9.615e-05 │ 0.7579 │ 80.76% ┃ 0.5472 │ 86.17% ┃ +┃ 4 │ 13:16 min ┃ 9.259e-05 │ 0.6296 │ 83.66% ┃ 0.4887 │ 87.45% ┃ +┃ 5 │ 13:14 min ┃ 8.798e-05 │ 0.5566 │ 85.49% ┃ 0.4577 │ 88.06% ┃ +┃ 6 │ 13:14 min ┃ 8.244e-05 │ 0.5090 │ 86.65% ┃ 0.4369 │ 88.31% ┃ +┃ 7 │ 13:09 min ┃ 7.610e-05 │ 0.4731 │ 87.46% ┃ 0.4236 │ 88.81% ┃ +┃ 8 │ 13:14 min ┃ 6.911e-05 │ 0.4456 │ 88.23% ┃ 0.4116 │ 88.98% ┃ +┃ 9 │ 13:14 min ┃ 6.166e-05 │ 0.4259 │ 88.77% ┃ 0.4026 │ 89.19% ┃ +┃ 10 │ 13:14 min ┃ 5.392e-05 │ 0.4042 │ 89.38% ┃ 0.3950 │ 89.39% ┃ +┃ 11 │ 13:16 min ┃ 4.608e-05 │ 0.3907 │ 89.74% ┃ 0.3903 │ 89.42% ┃ +┃ 12 │ 13:14 min ┃ 3.834e-05 │ 0.3814 │ 90.05% ┃ 0.3856 │ 89.62% ┃ +┃ 13 │ 13:14 min ┃ 3.088e-05 │ 0.3702 │ 90.36% ┃ 0.3832 │ 89.66% ┃ +┃ 14 │ 13:13 min ┃ 2.390e-05 │ 0.3609 │ 90.67% ┃ 0.3807 │ 89.77% ┃ +┃ 15 │ 13:09 min ┃ 1.756e-05 │ 0.3550 │ 90.78% ┃ 0.3783 │ 89.59% ┃ +┃ 16 │ 13:14 min ┃ 1.202e-05 │ 0.3493 │ 91.01% ┃ 0.3776 │ 89.72% ┃ +┃ 17 │ 13:14 min ┃ 7.410e-06 │ 0.3474 │ 91.07% ┃ 0.3767 │ 89.76% ┃ +┃ 18 │ 12:25 min ┃ 3.853e-06 │ 0.3440 │ 91.16% ┃ 0.3765 │ 89.78% ┃ +┃ 19 │ 13:14 min ┃ 1.431e-06 │ 0.3421 │ 91.17% ┃ 0.3763 │ 89.80% ┃ +┃ 20 │ 13:14 min ┃ 2.052e-07 │ 0.3436 │ 91.15% ┃ 0.3763 │ 89.81% ┃ diff --git a/TinyImageNet_MSAM_ViT_rho0.075_lr0.0001/data.hdf5 b/TinyImageNet_MSAM_ViT_rho0.075_lr0.0001/data.hdf5 new file mode 100644 index 0000000000000000000000000000000000000000..09f2e73e45408bcc626628bb45255e25ae9daf7c Binary files /dev/null and b/TinyImageNet_MSAM_ViT_rho0.075_lr0.0001/data.hdf5 differ diff --git a/TinyImageNet_MSAM_ViT_rho0.075_lr0.0001/final.model b/TinyImageNet_MSAM_ViT_rho0.075_lr0.0001/final.model new file mode 100644 index 0000000000000000000000000000000000000000..532b89a79462b9ba10bd4bbf03adfee152342ce0 --- /dev/null +++ b/TinyImageNet_MSAM_ViT_rho0.075_lr0.0001/final.model @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:bb6a44e93cfae55dfb62dd23e14cebb14a7120a63ab64ce8e42ccf684a74cade +size 701548454 diff --git a/TinyImageNet_MSAM_ViT_rho0.075_lr0.0001/params.json b/TinyImageNet_MSAM_ViT_rho0.075_lr0.0001/params.json new file mode 100644 index 0000000000000000000000000000000000000000..e443a5dc4705d4b220722fef7d1619d7744cd7a6 --- /dev/null +++ b/TinyImageNet_MSAM_ViT_rho0.075_lr0.0001/params.json @@ -0,0 +1,97 @@ +{ + "widthFactor": 4, + "ViTPatchsize": 32, + "ViTSize": "B", + "pyramidAlpha": 270, + "pyramidBottleneck": true, + "model": "ViT", + "dropout": 0.0, + "BN": true, + "depth": 16, + "label_smoothing": 0.1, + "ESAMbeta": 0.5, + "ESAMgamma": 0.7, + "lookSAMk": 5, + "lookSAMalpha": 0.7, + "lookSAMLayerwise": false, + "optimizer": "MSAM", + "weightDecay": 0.001, + "momentum": 0.9, + "rho": 0.075, + "samRho": 0.075, + "asamRho": 1.0, + "esamRho": 0.075, + "looksamRho": 0.05, + "friendlysamRho": 0.075, + "fishersamRho": 0.5, + "mongesamRho": 0.075, + "gsamRho": 0.075, + "bayesiansamRho": 0.075, + "msamRho": 0.075, + "adamwSamRho": 0.075, + "adamwMsamRho": 0.075, + "adamwFriendlysamRho": 0.075, + "sigma": 1.0, + "lmbda": 0.95, + "fisherEta": 0.1, + "gsamAlpha": 0.1, + "gsamAdaptive": false, + "bayesMSharpness": 8, + "bayesDelta": 10.0, + "bayesBeta1": 0.9, + "bayesBeta2": 0.999, + "bayesSInit": 1.0, + "bayesDamping": 1e-08, + "bayesEps": 1e-12, + "nesterov": false, + "grad_clip_norm": 0.0, + "dataThreads": 4, + "dataDir": "/home/hungpv/.datasets", + "batchSize": 64, + "dataset": "TinyImageNet", + "imageSize": 64, + "flip": true, + "crop": true, + "cut": false, + "cutoutProp": 0.5, + "randAugment": true, + "randAugment_magnitude": 10, + "mixup": false, + "mixupProp": 0.2, + "normalize": "std", + "truncate": false, + "verbose": false, + "metrics": [ + "learningRate", + "loss", + "accuracy" + ], + "logEach": 1, + "rndSeed": false, + "deterministic": false, + "learningRate": 0.0001, + "lrScheduler": [ + "startRamp", + "cos" + ], + "LRScheduler_step_steps": 3.0, + "LRScheduler_step_gamma": 0.2, + "LRScheduler_exp_maxDecay": 0.01, + "LRScheduler_cos_periods": 0.5, + "LRScheduler_StartRamp_epochs": 1.0, + "LRScheduler_WRN_T0": 50.0, + "LRScheduler_WRN_Tmult": 1.2, + "rhoScheduler": [ + "const" + ], + "rhoScheduler_StartRamp_epochs": 8.0, + "rhoScheduler_StartJump_epochs": 8.0, + "saveCheckpoint": true, + "keepLastCheckpoint": false, + "checkpointsList": [], + "saveCheckpointInterval": 0, + "logDir": "/mnt/hungpv/projects/SAM_Variant/logs", + "logSubDir": "TinyImageNet_MSAM_ViT_rho0.075_lr0.0001", + "epochs": 20, + "contin": false +} \ No newline at end of file diff --git a/TinyImageNet_MSAM_ViT_rho0.075_lr0.0001/terminal.log b/TinyImageNet_MSAM_ViT_rho0.075_lr0.0001/terminal.log new file mode 100644 index 0000000000000000000000000000000000000000..9bd89a801570c1b8d386ce6aabd16d81db36cbfb --- /dev/null +++ b/TinyImageNet_MSAM_ViT_rho0.075_lr0.0001/terminal.log @@ -0,0 +1,24 @@ +┏━━━━━━━╸S╺╸T╺╸A╺╸T╺╸S╺━━━━━━━┳━━━━━━━━━━━━━━━━━T╺╸R╺╸A╺╸I╺╸N ━━━━━━━━━━━━━━━━┳━━━━━━━━━━T╺╸E╺╸S╺╸T ━━━━━━━━━━┓ +┃ ┃ ┃ ┃ +┃ epoch │ time ┃ learningRate │ loss │ accuracy ┃ loss │ accuracy ┃ +┠──────────────┼──────────────╂───────────────┼───────────────┼───────────────╂───────────────┼───────────────┨ +┃ 1 │ 04:16 min ┃ 4.988e-05 │ 4.2781 │ 4.77% ┃ 2.4957 │ 35.53% ┃ +┃ 2 │ 03:00 min ┃ 9.857e-05 │ 1.2220 │ 68.81% ┃ 0.6547 │ 83.88% ┃ +┃ 3 │ 03:04 min ┃ 9.615e-05 │ 0.7458 │ 81.02% ┃ 0.5378 │ 86.17% ┃ +┃ 4 │ 03:01 min ┃ 9.259e-05 │ 0.6224 │ 83.70% ┃ 0.4808 │ 87.50% ┃ +┃ 5 │ 03:01 min ┃ 8.798e-05 │ 0.5484 │ 85.54% ┃ 0.4513 │ 88.13% ┃ +┃ 6 │ 03:01 min ┃ 8.244e-05 │ 0.5001 │ 86.67% ┃ 0.4340 │ 88.58% ┃ +┃ 7 │ 03:01 min ┃ 7.610e-05 │ 0.4615 │ 87.59% ┃ 0.4191 │ 88.78% ┃ +┃ 8 │ 03:10 min ┃ 6.911e-05 │ 0.4321 │ 88.31% ┃ 0.4073 │ 88.97% ┃ +┃ 9 │ 03:02 min ┃ 6.166e-05 │ 0.4116 │ 88.86% ┃ 0.3996 │ 89.16% ┃ +┃ 10 │ 03:01 min ┃ 5.392e-05 │ 0.3880 │ 89.56% ┃ 0.3953 │ 89.44% ┃ +┃ 11 │ 03:01 min ┃ 4.608e-05 │ 0.3740 │ 89.91% ┃ 0.3910 │ 89.38% ┃ +┃ 12 │ 03:01 min ┃ 3.834e-05 │ 0.3619 │ 90.25% ┃ 0.3871 │ 89.51% ┃ +┃ 13 │ 03:03 min ┃ 3.088e-05 │ 0.3492 │ 90.62% ┃ 0.3844 │ 89.57% ┃ +┃ 14 │ 03:01 min ┃ 2.390e-05 │ 0.3399 │ 90.92% ┃ 0.3819 │ 89.65% ┃ +┃ 15 │ 03:01 min ┃ 1.756e-05 │ 0.3326 │ 91.11% ┃ 0.3806 │ 89.60% ┃ +┃ 16 │ 03:01 min ┃ 1.202e-05 │ 0.3265 │ 91.33% ┃ 0.3801 │ 89.65% ┃ +┃ 17 │ 03:04 min ┃ 7.410e-06 │ 0.3239 │ 91.43% ┃ 0.3794 │ 89.73% ┃ +┃ 18 │ 03:01 min ┃ 3.853e-06 │ 0.3210 │ 91.46% ┃ 0.3788 │ 89.71% ┃ +┃ 19 │ 03:01 min ┃ 1.431e-06 │ 0.3178 │ 91.59% ┃ 0.3790 │ 89.68% ┃ +┃ 20 │ 03:05 min ┃ 2.052e-07 │ 0.3200 │ 91.48% ┃ 0.3789 │ 89.70% ┃ diff --git a/TinyImageNet_MSAM_ViT_rho0.075_lr0.0001_run01/data.hdf5 b/TinyImageNet_MSAM_ViT_rho0.075_lr0.0001_run01/data.hdf5 new file mode 100644 index 0000000000000000000000000000000000000000..e1414bc54f3c7f66afeb2a9e665174cbbff5b90c Binary files /dev/null and b/TinyImageNet_MSAM_ViT_rho0.075_lr0.0001_run01/data.hdf5 differ diff --git a/TinyImageNet_MSAM_ViT_rho0.075_lr0.0001_run01/final.model b/TinyImageNet_MSAM_ViT_rho0.075_lr0.0001_run01/final.model new file mode 100644 index 0000000000000000000000000000000000000000..0e6801ceff5df2b42fa942d0ea3c283b94fccdb8 --- /dev/null +++ b/TinyImageNet_MSAM_ViT_rho0.075_lr0.0001_run01/final.model @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:4c0406d8a275e5f666dbad6956740d983faf42badeda1c536108e2db5ae6bc59 +size 690046886 diff --git a/TinyImageNet_MSAM_ViT_rho0.075_lr0.0001_run01/params.json b/TinyImageNet_MSAM_ViT_rho0.075_lr0.0001_run01/params.json new file mode 100644 index 0000000000000000000000000000000000000000..661bf53052a462922c5040460fb737eeaaa391c5 --- /dev/null +++ b/TinyImageNet_MSAM_ViT_rho0.075_lr0.0001_run01/params.json @@ -0,0 +1,97 @@ +{ + "widthFactor": 4, + "ViTPatchsize": 16, + "ViTSize": "B", + "pyramidAlpha": 270, + "pyramidBottleneck": true, + "model": "ViT", + "dropout": 0.0, + "BN": true, + "depth": 16, + "label_smoothing": 0.1, + "ESAMbeta": 0.5, + "ESAMgamma": 0.7, + "lookSAMk": 5, + "lookSAMalpha": 0.7, + "lookSAMLayerwise": false, + "optimizer": "MSAM", + "weightDecay": 0.001, + "momentum": 0.9, + "rho": 0.01, + "samRho": 0.01, + "asamRho": 1.0, + "esamRho": 0.075, + "looksamRho": 0.05, + "friendlysamRho": 0.075, + "fishersamRho": 0.1, + "mongesamRho": 0.075, + "gsamRho": 0.075, + "bayesiansamRho": 0.075, + "msamRho": 0.075, + "adamwSamRho": 0.075, + "adamwMsamRho": 0.075, + "adamwFriendlysamRho": 0.075, + "sigma": 1.0, + "lmbda": 0.95, + "fisherEta": 0.1, + "gsamAlpha": 0.1, + "gsamAdaptive": false, + "bayesMSharpness": 8, + "bayesDelta": 10.0, + "bayesBeta1": 0.9, + "bayesBeta2": 0.999, + "bayesSInit": 1.0, + "bayesDamping": 1e-08, + "bayesEps": 1e-12, + "nesterov": false, + "grad_clip_norm": 0.0, + "dataThreads": 4, + "dataDir": "/home/hungpv/.datasets", + "batchSize": 64, + "dataset": "TinyImageNet", + "imageSize": 64, + "flip": true, + "crop": true, + "cut": false, + "cutoutProp": 0.5, + "randAugment": true, + "randAugment_magnitude": 10, + "mixup": false, + "mixupProp": 0.2, + "normalize": "std", + "truncate": false, + "verbose": false, + "metrics": [ + "learningRate", + "loss", + "accuracy" + ], + "logEach": 1, + "rndSeed": false, + "deterministic": false, + "learningRate": 0.0001, + "lrScheduler": [ + "startRamp", + "cos" + ], + "LRScheduler_step_steps": 3.0, + "LRScheduler_step_gamma": 0.2, + "LRScheduler_exp_maxDecay": 0.01, + "LRScheduler_cos_periods": 0.5, + "LRScheduler_StartRamp_epochs": 1.0, + "LRScheduler_WRN_T0": 50.0, + "LRScheduler_WRN_Tmult": 1.2, + "rhoScheduler": [ + "const" + ], + "rhoScheduler_StartRamp_epochs": 8.0, + "rhoScheduler_StartJump_epochs": 8.0, + "saveCheckpoint": true, + "keepLastCheckpoint": false, + "checkpointsList": [], + "saveCheckpointInterval": 0, + "logDir": "/mnt/hungpv/projects/SAM_Variant/logs", + "logSubDir": "TinyImageNet_MSAM_ViT_rho0.075_lr0.0001_run01", + "epochs": 20, + "contin": false +} \ No newline at end of file diff --git a/TinyImageNet_MSAM_ViT_rho0.075_lr0.0001_run01/terminal.log b/TinyImageNet_MSAM_ViT_rho0.075_lr0.0001_run01/terminal.log new file mode 100644 index 0000000000000000000000000000000000000000..5aac8eefdc737a96d4fb0cfcd44b8033632c67e9 --- /dev/null +++ b/TinyImageNet_MSAM_ViT_rho0.075_lr0.0001_run01/terminal.log @@ -0,0 +1,24 @@ +┏━━━━━━━╸S╺╸T╺╸A╺╸T╺╸S╺━━━━━━━┳━━━━━━━━━━━━━━━━━T╺╸R╺╸A╺╸I╺╸N ━━━━━━━━━━━━━━━━┳━━━━━━━━━━T╺╸E╺╸S╺╸T ━━━━━━━━━━┓ +┃ ┃ ┃ ┃ +┃ epoch │ time ┃ learningRate │ loss │ accuracy ┃ loss │ accuracy ┃ +┠──────────────┼──────────────╂───────────────┼───────────────┼───────────────╂───────────────┼───────────────┨ +┃ 1 │ 37:28 min ┃ 4.988e-05 │ 4.4369 │ 2.38% ┃ 3.7020 │ 9.98% ┃ +┃ 2 │ 36:04 min ┃ 9.857e-05 │ 2.2456 │ 40.90% ┃ 1.1276 │ 71.09% ┃ +┃ 3 │ 36:02 min ┃ 9.615e-05 │ 0.8574 │ 78.75% ┃ 0.5902 │ 86.32% ┃ +┃ 4 │ 36:15 min ┃ 9.259e-05 │ 0.5768 │ 85.98% ┃ 0.4834 │ 88.49% ┃ +┃ 5 │ 34:54 min ┃ 8.798e-05 │ 0.4718 │ 88.47% ┃ 0.4307 │ 89.42% ┃ +┃ 6 │ 34:31 min ┃ 8.244e-05 │ 0.4079 │ 89.73% ┃ 0.3970 │ 90.19% ┃ +┃ 7 │ 36:07 min ┃ 7.610e-05 │ 0.3633 │ 90.81% ┃ 0.3857 │ 90.23% ┃ +┃ 8 │ 36:17 min ┃ 6.911e-05 │ 0.3324 │ 91.58% ┃ 0.3707 │ 90.49% ┃ +┃ 9 │ 35:59 min ┃ 6.166e-05 │ 0.3069 │ 92.24% ┃ 0.3639 │ 90.64% ┃ +┃ 10 │ 36:18 min ┃ 5.392e-05 │ 0.2838 │ 92.97% ┃ 0.3571 │ 90.76% ┃ +┃ 11 │ 36:06 min ┃ 4.608e-05 │ 0.2670 │ 93.37% ┃ 0.3555 │ 90.62% ┃ +┃ 12 │ 25:48 min ┃ 3.834e-05 │ 0.2526 │ 93.77% ┃ 0.3480 │ 91.06% ┃ +┃ 13 │ 12:42 min ┃ 3.088e-05 │ 0.2410 │ 94.12% ┃ 0.3470 │ 90.95% ┃ +┃ 14 │ 12:42 min ┃ 2.390e-05 │ 0.2317 │ 94.41% ┃ 0.3470 │ 91.00% ┃ +┃ 15 │ 12:42 min ┃ 1.756e-05 │ 0.2263 │ 94.54% ┃ 0.3439 │ 90.97% ┃ +┃ 16 │ 12:42 min ┃ 1.202e-05 │ 0.2209 │ 94.70% ┃ 0.3427 │ 91.01% ┃ +┃ 17 │ 12:42 min ┃ 7.410e-06 │ 0.2165 │ 94.85% ┃ 0.3419 │ 91.02% ┃ +┃ 18 │ 12:43 min ┃ 3.853e-06 │ 0.2139 │ 94.89% ┃ 0.3418 │ 91.00% ┃ +┃ 19 │ 12:42 min ┃ 1.431e-06 │ 0.2137 │ 94.91% ┃ 0.3417 │ 91.04% ┃ +┃ 20 │ 12:42 min ┃ 2.052e-07 │ 0.2129 │ 94.95% ┃ 0.3416 │ 91.10% ┃ diff --git a/TinyImageNet_MongeSAM_ViT_rho0.075_lr0.0001/data.hdf5 b/TinyImageNet_MongeSAM_ViT_rho0.075_lr0.0001/data.hdf5 new file mode 100644 index 0000000000000000000000000000000000000000..a87624f0ce48699c7ebe42e33bf550d8b464988a Binary files /dev/null and b/TinyImageNet_MongeSAM_ViT_rho0.075_lr0.0001/data.hdf5 differ diff --git a/TinyImageNet_MongeSAM_ViT_rho0.075_lr0.0001/final.model b/TinyImageNet_MongeSAM_ViT_rho0.075_lr0.0001/final.model new file mode 100644 index 0000000000000000000000000000000000000000..fa40d1bb4b29b9be4bc0bc55c01d17b785aba805 --- /dev/null +++ b/TinyImageNet_MongeSAM_ViT_rho0.075_lr0.0001/final.model @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:2c72611fc23f45bed32a9e92e634dec2e4b7b6010ba0fd64b37e249b4fe97fb5 +size 1052309706 diff --git a/TinyImageNet_MongeSAM_ViT_rho0.075_lr0.0001/params.json b/TinyImageNet_MongeSAM_ViT_rho0.075_lr0.0001/params.json new file mode 100644 index 0000000000000000000000000000000000000000..f482520eac9b8af1f30f3972cf9a5537cdeb3db1 --- /dev/null +++ b/TinyImageNet_MongeSAM_ViT_rho0.075_lr0.0001/params.json @@ -0,0 +1,97 @@ +{ + "widthFactor": 4, + "ViTPatchsize": 32, + "ViTSize": "B", + "pyramidAlpha": 270, + "pyramidBottleneck": true, + "model": "ViT", + "dropout": 0.0, + "BN": true, + "depth": 16, + "label_smoothing": 0.1, + "ESAMbeta": 0.5, + "ESAMgamma": 0.7, + "lookSAMk": 5, + "lookSAMalpha": 0.7, + "lookSAMLayerwise": false, + "optimizer": "MongeSAM", + "weightDecay": 0.001, + "momentum": 0.9, + "rho": 0.075, + "samRho": 0.075, + "asamRho": 1.0, + "esamRho": 0.075, + "looksamRho": 0.05, + "friendlysamRho": 0.075, + "fishersamRho": 0.5, + "mongesamRho": 0.075, + "gsamRho": 0.075, + "bayesiansamRho": 0.075, + "msamRho": 0.075, + "adamwSamRho": 0.075, + "adamwMsamRho": 0.075, + "adamwFriendlysamRho": 0.075, + "sigma": 1.0, + "lmbda": 0.95, + "fisherEta": 0.1, + "gsamAlpha": 0.1, + "gsamAdaptive": false, + "bayesMSharpness": 8, + "bayesDelta": 10.0, + "bayesBeta1": 0.9, + "bayesBeta2": 0.999, + "bayesSInit": 1.0, + "bayesDamping": 1e-08, + "bayesEps": 1e-12, + "nesterov": false, + "grad_clip_norm": 0.0, + "dataThreads": 4, + "dataDir": "/home/hungpv/.datasets", + "batchSize": 64, + "dataset": "TinyImageNet", + "imageSize": 64, + "flip": true, + "crop": true, + "cut": false, + "cutoutProp": 0.5, + "randAugment": true, + "randAugment_magnitude": 10, + "mixup": false, + "mixupProp": 0.2, + "normalize": "std", + "truncate": false, + "verbose": false, + "metrics": [ + "learningRate", + "loss", + "accuracy" + ], + "logEach": 1, + "rndSeed": false, + "deterministic": false, + "learningRate": 0.0001, + "lrScheduler": [ + "startRamp", + "cos" + ], + "LRScheduler_step_steps": 3.0, + "LRScheduler_step_gamma": 0.2, + "LRScheduler_exp_maxDecay": 0.01, + "LRScheduler_cos_periods": 0.5, + "LRScheduler_StartRamp_epochs": 1.0, + "LRScheduler_WRN_T0": 50.0, + "LRScheduler_WRN_Tmult": 1.2, + "rhoScheduler": [ + "const" + ], + "rhoScheduler_StartRamp_epochs": 8.0, + "rhoScheduler_StartJump_epochs": 8.0, + "saveCheckpoint": true, + "keepLastCheckpoint": false, + "checkpointsList": [], + "saveCheckpointInterval": 0, + "logDir": "/mnt/hungpv/projects/SAM_Variant/logs", + "logSubDir": "TinyImageNet_MongeSAM_ViT_rho0.075_lr0.0001", + "epochs": 20, + "contin": false +} \ No newline at end of file diff --git a/TinyImageNet_MongeSAM_ViT_rho0.075_lr0.0001/terminal.log b/TinyImageNet_MongeSAM_ViT_rho0.075_lr0.0001/terminal.log new file mode 100644 index 0000000000000000000000000000000000000000..832f83ee620bcdf2ec5ce16a940dc7f219eeb414 --- /dev/null +++ b/TinyImageNet_MongeSAM_ViT_rho0.075_lr0.0001/terminal.log @@ -0,0 +1,24 @@ +┏━━━━━━━╸S╺╸T╺╸A╺╸T╺╸S╺━━━━━━━┳━━━━━━━━━━━━━━━━━T╺╸R╺╸A╺╸I╺╸N ━━━━━━━━━━━━━━━━┳━━━━━━━━━━T╺╸E╺╸S╺╸T ━━━━━━━━━━┓ +┃ ┃ ┃ ┃ +┃ epoch │ time ┃ learningRate │ loss │ accuracy ┃ loss │ accuracy ┃ +┠──────────────┼──────────────╂───────────────┼───────────────┼───────────────╂───────────────┼───────────────┨ +┃ 1 │ 06:47 min ┃ 4.988e-05 │ 4.4699 │ 2.98% ┃ 3.0921 │ 22.68% ┃ +┃ 2 │ 05:21 min ┃ 9.857e-05 │ 1.3560 │ 65.49% ┃ 0.6739 │ 83.48% ┃ +┃ 3 │ 05:22 min ┃ 9.615e-05 │ 0.7467 │ 81.03% ┃ 0.5441 │ 86.12% ┃ +┃ 4 │ 05:22 min ┃ 9.259e-05 │ 0.6203 │ 83.81% ┃ 0.4857 │ 87.53% ┃ +┃ 5 │ 05:22 min ┃ 8.798e-05 │ 0.5459 │ 85.71% ┃ 0.4536 │ 87.92% ┃ +┃ 6 │ 05:22 min ┃ 8.244e-05 │ 0.4971 │ 86.87% ┃ 0.4348 │ 88.35% ┃ +┃ 7 │ 05:22 min ┃ 7.610e-05 │ 0.4597 │ 87.77% ┃ 0.4216 │ 88.84% ┃ +┃ 8 │ 05:22 min ┃ 6.911e-05 │ 0.4311 │ 88.51% ┃ 0.4081 │ 88.96% ┃ +┃ 9 │ 05:22 min ┃ 6.166e-05 │ 0.4108 │ 89.09% ┃ 0.3989 │ 89.09% ┃ +┃ 10 │ 05:22 min ┃ 5.392e-05 │ 0.3880 │ 89.73% ┃ 0.3924 │ 89.52% ┃ +┃ 11 │ 05:22 min ┃ 4.608e-05 │ 0.3737 │ 90.09% ┃ 0.3877 │ 89.44% ┃ +┃ 12 │ 05:22 min ┃ 3.834e-05 │ 0.3640 │ 90.38% ┃ 0.3828 │ 89.62% ┃ +┃ 13 │ 05:22 min ┃ 3.088e-05 │ 0.3522 │ 90.69% ┃ 0.3806 │ 89.66% ┃ +┃ 14 │ 05:22 min ┃ 2.390e-05 │ 0.3426 │ 91.07% ┃ 0.3781 │ 89.72% ┃ +┃ 15 │ 05:22 min ┃ 1.756e-05 │ 0.3359 │ 91.21% ┃ 0.3756 │ 89.85% ┃ +┃ 16 │ 05:22 min ┃ 1.202e-05 │ 0.3304 │ 91.47% ┃ 0.3751 │ 89.79% ┃ +┃ 17 │ 05:22 min ┃ 7.410e-06 │ 0.3283 │ 91.54% ┃ 0.3740 │ 89.83% ┃ +┃ 18 │ 05:22 min ┃ 3.853e-06 │ 0.3247 │ 91.53% ┃ 0.3738 │ 89.86% ┃ +┃ 19 │ 05:22 min ┃ 1.431e-06 │ 0.3227 │ 91.61% ┃ 0.3737 │ 89.86% ┃ +┃ 20 │ 05:22 min ┃ 2.052e-07 │ 0.3242 │ 91.57% ┃ 0.3737 │ 89.85% ┃ diff --git a/TinyImageNet_MongeSAM_ViT_rho0.075_lr0.0001_run01/data.hdf5 b/TinyImageNet_MongeSAM_ViT_rho0.075_lr0.0001_run01/data.hdf5 new file mode 100644 index 0000000000000000000000000000000000000000..93687aed0f4c78caeac09f243b55f86b5a7c84dc Binary files /dev/null and b/TinyImageNet_MongeSAM_ViT_rho0.075_lr0.0001_run01/data.hdf5 differ diff --git a/TinyImageNet_MongeSAM_ViT_rho0.075_lr0.0001_run01/final.model b/TinyImageNet_MongeSAM_ViT_rho0.075_lr0.0001_run01/final.model new file mode 100644 index 0000000000000000000000000000000000000000..12272ace1cdd4f85708e76ff437fd4ff9edd234a --- /dev/null +++ b/TinyImageNet_MongeSAM_ViT_rho0.075_lr0.0001_run01/final.model @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:6339dfb07c1f3c690bdf81c411bc85f7b363999ed11a48b6e3de596810a6e75c +size 1035057354 diff --git a/TinyImageNet_MongeSAM_ViT_rho0.075_lr0.0001_run01/params.json b/TinyImageNet_MongeSAM_ViT_rho0.075_lr0.0001_run01/params.json new file mode 100644 index 0000000000000000000000000000000000000000..beadfdf890e7dba90f9491d48d894ec33e453a1f --- /dev/null +++ b/TinyImageNet_MongeSAM_ViT_rho0.075_lr0.0001_run01/params.json @@ -0,0 +1,97 @@ +{ + "widthFactor": 4, + "ViTPatchsize": 16, + "ViTSize": "B", + "pyramidAlpha": 270, + "pyramidBottleneck": true, + "model": "ViT", + "dropout": 0.0, + "BN": true, + "depth": 16, + "label_smoothing": 0.1, + "ESAMbeta": 0.5, + "ESAMgamma": 0.7, + "lookSAMk": 5, + "lookSAMalpha": 0.7, + "lookSAMLayerwise": false, + "optimizer": "MongeSAM", + "weightDecay": 0.001, + "momentum": 0.9, + "rho": 0.01, + "samRho": 0.01, + "asamRho": 1.0, + "esamRho": 0.075, + "looksamRho": 0.05, + "friendlysamRho": 0.075, + "fishersamRho": 0.1, + "mongesamRho": 0.075, + "gsamRho": 0.075, + "bayesiansamRho": 0.075, + "msamRho": 0.075, + "adamwSamRho": 0.075, + "adamwMsamRho": 0.075, + "adamwFriendlysamRho": 0.075, + "sigma": 1.0, + "lmbda": 0.95, + "fisherEta": 0.1, + "gsamAlpha": 0.1, + "gsamAdaptive": false, + "bayesMSharpness": 8, + "bayesDelta": 10.0, + "bayesBeta1": 0.9, + "bayesBeta2": 0.999, + "bayesSInit": 1.0, + "bayesDamping": 1e-08, + "bayesEps": 1e-12, + "nesterov": false, + "grad_clip_norm": 0.0, + "dataThreads": 4, + "dataDir": "/home/hungpv/.datasets", + "batchSize": 64, + "dataset": "TinyImageNet", + "imageSize": 64, + "flip": true, + "crop": true, + "cut": false, + "cutoutProp": 0.5, + "randAugment": true, + "randAugment_magnitude": 10, + "mixup": false, + "mixupProp": 0.2, + "normalize": "std", + "truncate": false, + "verbose": false, + "metrics": [ + "learningRate", + "loss", + "accuracy" + ], + "logEach": 1, + "rndSeed": false, + "deterministic": false, + "learningRate": 0.0001, + "lrScheduler": [ + "startRamp", + "cos" + ], + "LRScheduler_step_steps": 3.0, + "LRScheduler_step_gamma": 0.2, + "LRScheduler_exp_maxDecay": 0.01, + "LRScheduler_cos_periods": 0.5, + "LRScheduler_StartRamp_epochs": 1.0, + "LRScheduler_WRN_T0": 50.0, + "LRScheduler_WRN_Tmult": 1.2, + "rhoScheduler": [ + "const" + ], + "rhoScheduler_StartRamp_epochs": 8.0, + "rhoScheduler_StartJump_epochs": 8.0, + "saveCheckpoint": true, + "keepLastCheckpoint": false, + "checkpointsList": [], + "saveCheckpointInterval": 0, + "logDir": "/mnt/hungpv/projects/SAM_Variant/logs", + "logSubDir": "TinyImageNet_MongeSAM_ViT_rho0.075_lr0.0001_run01", + "epochs": 20, + "contin": false +} \ No newline at end of file diff --git a/TinyImageNet_MongeSAM_ViT_rho0.075_lr0.0001_run01/terminal.log b/TinyImageNet_MongeSAM_ViT_rho0.075_lr0.0001_run01/terminal.log new file mode 100644 index 0000000000000000000000000000000000000000..2f284ec29d7039695a2a028b5ecf56a2e33db21f --- /dev/null +++ b/TinyImageNet_MongeSAM_ViT_rho0.075_lr0.0001_run01/terminal.log @@ -0,0 +1,24 @@ +┏━━━━━━━╸S╺╸T╺╸A╺╸T╺╸S╺━━━━━━━┳━━━━━━━━━━━━━━━━━T╺╸R╺╸A╺╸I╺╸N ━━━━━━━━━━━━━━━━┳━━━━━━━━━━T╺╸E╺╸S╺╸T ━━━━━━━━━━┓ +┃ ┃ ┃ ┃ +┃ epoch │ time ┃ learningRate │ loss │ accuracy ┃ loss │ accuracy ┃ +┠──────────────┼──────────────╂───────────────┼───────────────┼───────────────╂───────────────┼───────────────┨ +┃ 1 │ 44:55 min ┃ 4.988e-05 │ 4.4800 │ 2.10% ┃ 3.8151 │ 8.30% ┃ +┃ 2 │ 50:58 min ┃ 9.857e-05 │ 2.2448 │ 41.36% ┃ 1.0051 │ 75.02% ┃ +┃ 3 │ 51:47 min ┃ 9.615e-05 │ 0.7864 │ 80.85% ┃ 0.5522 │ 87.27% ┃ +┃ 4 │ 51:52 min ┃ 9.259e-05 │ 0.5521 │ 86.51% ┃ 0.4635 │ 88.78% ┃ +┃ 5 │ 51:52 min ┃ 8.798e-05 │ 0.4574 │ 88.73% ┃ 0.4195 │ 89.57% ┃ +┃ 6 │ 51:46 min ┃ 8.244e-05 │ 0.3973 │ 89.93% ┃ 0.3898 │ 90.35% ┃ +┃ 7 │ 51:10 min ┃ 7.610e-05 │ 0.3552 │ 91.00% ┃ 0.3795 │ 90.26% ┃ +┃ 8 │ 50:58 min ┃ 6.911e-05 │ 0.3264 │ 91.71% ┃ 0.3635 │ 90.81% ┃ +┃ 9 │ 51:51 min ┃ 6.166e-05 │ 0.3016 │ 92.39% ┃ 0.3571 │ 90.86% ┃ +┃ 10 │ 51:47 min ┃ 5.392e-05 │ 0.2789 │ 93.05% ┃ 0.3500 │ 90.82% ┃ +┃ 11 │ 51:52 min ┃ 4.608e-05 │ 0.2624 │ 93.48% ┃ 0.3486 │ 90.88% ┃ +┃ 12 │ 51:52 min ┃ 3.834e-05 │ 0.2490 │ 93.85% ┃ 0.3423 │ 91.12% ┃ +┃ 13 │ 51:06 min ┃ 3.088e-05 │ 0.2373 │ 94.22% ┃ 0.3399 │ 91.16% ┃ +┃ 14 │ 50:59 min ┃ 2.390e-05 │ 0.2286 │ 94.49% ┃ 0.3399 │ 91.23% ┃ +┃ 15 │ 51:52 min ┃ 1.756e-05 │ 0.2232 │ 94.59% ┃ 0.3370 │ 91.31% ┃ +┃ 16 │ 51:51 min ┃ 1.202e-05 │ 0.2180 │ 94.80% ┃ 0.3352 │ 91.30% ┃ +┃ 17 │ 51:49 min ┃ 7.410e-06 │ 0.2133 │ 94.96% ┃ 0.3351 │ 91.32% ┃ +┃ 18 │ 51:52 min ┃ 3.853e-06 │ 0.2112 │ 95.01% ┃ 0.3349 │ 91.27% ┃ +┃ 19 │ 51:51 min ┃ 1.431e-06 │ 0.2112 │ 94.95% ┃ 0.3347 │ 91.35% ┃ +┃ 20 │ 50:16 min ┃ 2.052e-07 │ 0.2104 │ 95.02% ┃ 0.3346 │ 91.37% ┃ diff --git a/TinyImageNet_SAM_ViT_rho0.01_lr0.0001/data.hdf5 b/TinyImageNet_SAM_ViT_rho0.01_lr0.0001/data.hdf5 new file mode 100644 index 0000000000000000000000000000000000000000..8ed952fe0138e7bf021cf58cba1727a8e31640cb Binary files /dev/null and b/TinyImageNet_SAM_ViT_rho0.01_lr0.0001/data.hdf5 differ diff --git a/TinyImageNet_SAM_ViT_rho0.01_lr0.0001/final.model b/TinyImageNet_SAM_ViT_rho0.01_lr0.0001/final.model new file mode 100644 index 0000000000000000000000000000000000000000..2abba0c150a252178a9acea086e6f5bb7788810f --- /dev/null +++ b/TinyImageNet_SAM_ViT_rho0.01_lr0.0001/final.model @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:c6a6c848957188694ee3935910378532f4dbe2d5d624a103115d73fdcce1f6fd +size 1035057418 diff --git a/TinyImageNet_SAM_ViT_rho0.01_lr0.0001/params.json b/TinyImageNet_SAM_ViT_rho0.01_lr0.0001/params.json new file mode 100644 index 0000000000000000000000000000000000000000..5e75d8f25e3cf401e090dcdf982f2f09bf89426c --- /dev/null +++ b/TinyImageNet_SAM_ViT_rho0.01_lr0.0001/params.json @@ -0,0 +1,97 @@ +{ + "widthFactor": 4, + "ViTPatchsize": 16, + "ViTSize": "B", + "pyramidAlpha": 270, + "pyramidBottleneck": true, + "model": "ViT", + "dropout": 0.0, + "BN": true, + "depth": 16, + "label_smoothing": 0.1, + "ESAMbeta": 0.5, + "ESAMgamma": 0.7, + "lookSAMk": 5, + "lookSAMalpha": 0.7, + "lookSAMLayerwise": false, + "optimizer": "SAM", + "weightDecay": 0.001, + "momentum": 0.9, + "rho": 0.01, + "samRho": 0.01, + "asamRho": 1.0, + "esamRho": 0.075, + "looksamRho": 0.05, + "friendlysamRho": 0.075, + "fishersamRho": 0.1, + "mongesamRho": 0.075, + "gsamRho": 0.075, + "bayesiansamRho": 0.075, + "msamRho": 0.075, + "adamwSamRho": 0.075, + "adamwMsamRho": 0.075, + "adamwFriendlysamRho": 0.075, + "sigma": 1.0, + "lmbda": 0.95, + "fisherEta": 0.1, + "gsamAlpha": 0.1, + "gsamAdaptive": false, + "bayesMSharpness": 8, + "bayesDelta": 10.0, + "bayesBeta1": 0.9, + "bayesBeta2": 0.999, + "bayesSInit": 1.0, + "bayesDamping": 1e-08, + "bayesEps": 1e-12, + "nesterov": false, + "grad_clip_norm": 0.0, + "dataThreads": 4, + "dataDir": "/home/hungpv/.datasets", + "batchSize": 64, + "dataset": "TinyImageNet", + "imageSize": 64, + "flip": true, + "crop": true, + "cut": false, + "cutoutProp": 0.5, + "randAugment": true, + "randAugment_magnitude": 10, + "mixup": false, + "mixupProp": 0.2, + "normalize": "std", + "truncate": false, + "verbose": false, + "metrics": [ + "learningRate", + "loss", + "accuracy" + ], + "logEach": 1, + "rndSeed": false, + "deterministic": false, + "learningRate": 0.0001, + "lrScheduler": [ + "startRamp", + "cos" + ], + "LRScheduler_step_steps": 3.0, + "LRScheduler_step_gamma": 0.2, + "LRScheduler_exp_maxDecay": 0.01, + "LRScheduler_cos_periods": 0.5, + "LRScheduler_StartRamp_epochs": 1.0, + "LRScheduler_WRN_T0": 50.0, + "LRScheduler_WRN_Tmult": 1.2, + "rhoScheduler": [ + "const" + ], + "rhoScheduler_StartRamp_epochs": 8.0, + "rhoScheduler_StartJump_epochs": 8.0, + "saveCheckpoint": true, + "keepLastCheckpoint": false, + "checkpointsList": [], + "saveCheckpointInterval": 0, + "logDir": "/mnt/hungpv/projects/SAM_Variant/logs", + "logSubDir": "TinyImageNet_SAM_ViT_rho0.01_lr0.0001", + "epochs": 20, + "contin": false +} \ No newline at end of file diff --git a/TinyImageNet_SAM_ViT_rho0.01_lr0.0001/terminal.log b/TinyImageNet_SAM_ViT_rho0.01_lr0.0001/terminal.log new file mode 100644 index 0000000000000000000000000000000000000000..33e10e2441fd17f3e1e724088f6cafbc47580b7b --- /dev/null +++ b/TinyImageNet_SAM_ViT_rho0.01_lr0.0001/terminal.log @@ -0,0 +1,24 @@ +┏━━━━━━━╸S╺╸T╺╸A╺╸T╺╸S╺━━━━━━━┳━━━━━━━━━━━━━━━━━T╺╸R╺╸A╺╸I╺╸N ━━━━━━━━━━━━━━━━┳━━━━━━━━━━T╺╸E╺╸S╺╸T ━━━━━━━━━━┓ +┃ ┃ ┃ ┃ +┃ epoch │ time ┃ learningRate │ loss │ accuracy ┃ loss │ accuracy ┃ +┠──────────────┼──────────────╂───────────────┼───────────────┼───────────────╂───────────────┼───────────────┨ +┃ 1 │ 41:46 min ┃ 4.988e-05 │ 4.5648 │ 1.27% ┃ 4.2576 │ 3.65% ┃ +┃ 2 │ 51:17 min ┃ 9.857e-05 │ 2.7258 │ 32.51% ┃ 0.9032 │ 78.09% ┃ +┃ 3 │ 51:17 min ┃ 9.615e-05 │ 0.7303 │ 82.30% ┃ 0.5142 │ 88.01% ┃ +┃ 4 │ 51:11 min ┃ 9.259e-05 │ 0.5305 │ 86.94% ┃ 0.4387 │ 89.54% ┃ +┃ 5 │ 51:17 min ┃ 8.798e-05 │ 0.4488 │ 88.88% ┃ 0.4009 │ 90.14% ┃ +┃ 6 │ 49:35 min ┃ 8.244e-05 │ 0.3981 │ 90.01% ┃ 0.3765 │ 90.61% ┃ +┃ 7 │ 51:11 min ┃ 7.610e-05 │ 0.3609 │ 90.97% ┃ 0.3624 │ 90.93% ┃ +┃ 8 │ 51:11 min ┃ 6.911e-05 │ 0.3355 │ 91.64% ┃ 0.3508 │ 91.04% ┃ +┃ 9 │ 51:10 min ┃ 6.166e-05 │ 0.3150 │ 92.13% ┃ 0.3434 │ 91.24% ┃ +┃ 10 │ 51:09 min ┃ 5.392e-05 │ 0.2957 │ 92.73% ┃ 0.3368 │ 91.42% ┃ +┃ 11 │ 49:31 min ┃ 4.608e-05 │ 0.2820 │ 93.11% ┃ 0.3342 │ 91.28% ┃ +┃ 12 │ 51:11 min ┃ 3.834e-05 │ 0.2698 │ 93.41% ┃ 0.3287 │ 91.57% ┃ +┃ 13 │ 51:09 min ┃ 3.088e-05 │ 0.2610 │ 93.65% ┃ 0.3263 │ 91.46% ┃ +┃ 14 │ 51:17 min ┃ 2.390e-05 │ 0.2531 │ 93.97% ┃ 0.3247 │ 91.57% ┃ +┃ 15 │ 51:10 min ┃ 1.756e-05 │ 0.2489 │ 94.06% ┃ 0.3225 │ 91.77% ┃ +┃ 16 │ 49:14 min ┃ 1.202e-05 │ 0.2445 │ 94.20% ┃ 0.3211 │ 91.63% ┃ +┃ 17 │ 50:39 min ┃ 7.410e-06 │ 0.2412 │ 94.31% ┃ 0.3204 │ 91.72% ┃ +┃ 18 │ 50:39 min ┃ 3.853e-06 │ 0.2393 │ 94.31% ┃ 0.3201 │ 91.69% ┃ +┃ 19 │ 50:36 min ┃ 1.431e-06 │ 0.2388 │ 94.30% ┃ 0.3200 │ 91.68% ┃ +┃ 20 │ 50:47 min ┃ 2.052e-07 │ 0.2389 │ 94.43% ┃ 0.3200 │ 91.67% ┃ diff --git a/TinyImageNet_SAM_ViT_rho0.01_lr0.0001_run01/data.hdf5 b/TinyImageNet_SAM_ViT_rho0.01_lr0.0001_run01/data.hdf5 new file mode 100644 index 0000000000000000000000000000000000000000..96ed9517e2e860b8037ff8e8e08af4b92b243fd2 Binary files /dev/null and b/TinyImageNet_SAM_ViT_rho0.01_lr0.0001_run01/data.hdf5 differ diff --git a/TinyImageNet_SAM_ViT_rho0.01_lr0.0001_run01/final.model b/TinyImageNet_SAM_ViT_rho0.01_lr0.0001_run01/final.model new file mode 100644 index 0000000000000000000000000000000000000000..4d5390cb1905d7d21ee51af96876cc17fbca8d35 --- /dev/null +++ b/TinyImageNet_SAM_ViT_rho0.01_lr0.0001_run01/final.model @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:16af97282736bb9a0c0c3ae0a37c84d47da370c497551656b0498b7c079a36c0 +size 1052309770 diff --git a/TinyImageNet_SAM_ViT_rho0.01_lr0.0001_run01/params.json b/TinyImageNet_SAM_ViT_rho0.01_lr0.0001_run01/params.json new file mode 100644 index 0000000000000000000000000000000000000000..4db35fb18ee2031aca1fe9b940e6bc3f9afdda80 --- /dev/null +++ b/TinyImageNet_SAM_ViT_rho0.01_lr0.0001_run01/params.json @@ -0,0 +1,97 @@ +{ + "widthFactor": 4, + "ViTPatchsize": 32, + "ViTSize": "B", + "pyramidAlpha": 270, + "pyramidBottleneck": true, + "model": "ViT", + "dropout": 0.0, + "BN": true, + "depth": 16, + "label_smoothing": 0.1, + "ESAMbeta": 0.5, + "ESAMgamma": 0.7, + "lookSAMk": 5, + "lookSAMalpha": 0.7, + "lookSAMLayerwise": false, + "optimizer": "SAM", + "weightDecay": 0.001, + "momentum": 0.9, + "rho": 0.01, + "samRho": 0.01, + "asamRho": 1.0, + "esamRho": 0.075, + "looksamRho": 0.05, + "friendlysamRho": 0.01, + "fishersamRho": 0.01, + "mongesamRho": 0.075, + "gsamRho": 0.01, + "bayesiansamRho": 0.075, + "msamRho": 0.075, + "adamwSamRho": 0.075, + "adamwMsamRho": 0.075, + "adamwFriendlysamRho": 0.075, + "sigma": 1.0, + "lmbda": 0.95, + "fisherEta": 0.1, + "gsamAlpha": 0.1, + "gsamAdaptive": false, + "bayesMSharpness": 8, + "bayesDelta": 10.0, + "bayesBeta1": 0.9, + "bayesBeta2": 0.999, + "bayesSInit": 1.0, + "bayesDamping": 1e-08, + "bayesEps": 1e-12, + "nesterov": false, + "grad_clip_norm": 0.0, + "dataThreads": 4, + "dataDir": "/home/hungpv/.datasets", + "batchSize": 64, + "dataset": "TinyImageNet", + "imageSize": 64, + "flip": true, + "crop": true, + "cut": false, + "cutoutProp": 0.5, + "randAugment": true, + "randAugment_magnitude": 10, + "mixup": false, + "mixupProp": 0.2, + "normalize": "std", + "truncate": false, + "verbose": false, + "metrics": [ + "learningRate", + "loss", + "accuracy" + ], + "logEach": 1, + "rndSeed": false, + "deterministic": false, + "learningRate": 0.0001, + "lrScheduler": [ + "startRamp", + "cos" + ], + "LRScheduler_step_steps": 3.0, + "LRScheduler_step_gamma": 0.2, + "LRScheduler_exp_maxDecay": 0.01, + "LRScheduler_cos_periods": 0.5, + "LRScheduler_StartRamp_epochs": 1.0, + "LRScheduler_WRN_T0": 50.0, + "LRScheduler_WRN_Tmult": 1.2, + "rhoScheduler": [ + "const" + ], + "rhoScheduler_StartRamp_epochs": 8.0, + "rhoScheduler_StartJump_epochs": 8.0, + "saveCheckpoint": true, + "keepLastCheckpoint": false, + "checkpointsList": [], + "saveCheckpointInterval": 0, + "logDir": "/mnt/hungpv/projects/SAM_Variant/logs", + "logSubDir": "TinyImageNet_SAM_ViT_rho0.01_lr0.0001_run01", + "epochs": 20, + "contin": false +} \ No newline at end of file diff --git a/TinyImageNet_SAM_ViT_rho0.01_lr0.0001_run01/terminal.log b/TinyImageNet_SAM_ViT_rho0.01_lr0.0001_run01/terminal.log new file mode 100644 index 0000000000000000000000000000000000000000..189bce809e58db34f5cf7580081e43885e288c93 --- /dev/null +++ b/TinyImageNet_SAM_ViT_rho0.01_lr0.0001_run01/terminal.log @@ -0,0 +1,24 @@ +┏━━━━━━━╸S╺╸T╺╸A╺╸T╺╸S╺━━━━━━━┳━━━━━━━━━━━━━━━━━T╺╸R╺╸A╺╸I╺╸N ━━━━━━━━━━━━━━━━┳━━━━━━━━━━T╺╸E╺╸S╺╸T ━━━━━━━━━━┓ +┃ ┃ ┃ ┃ +┃ epoch │ time ┃ learningRate │ loss │ accuracy ┃ loss │ accuracy ┃ +┠──────────────┼──────────────╂───────────────┼───────────────┼───────────────╂───────────────┼───────────────┨ +┃ 1 │ 13:31 min ┃ 4.988e-05 │ 4.4772 │ 2.83% ┃ 3.1783 │ 21.05% ┃ +┃ 2 │ 12:20 min ┃ 9.857e-05 │ 1.3946 │ 64.56% ┃ 0.6811 │ 83.22% ┃ +┃ 3 │ 11:10 min ┃ 9.615e-05 │ 0.7530 │ 80.87% ┃ 0.5444 │ 86.26% ┃ +┃ 4 │ 12:21 min ┃ 9.259e-05 │ 0.6255 │ 83.73% ┃ 0.4868 │ 87.43% ┃ +┃ 5 │ 12:21 min ┃ 8.798e-05 │ 0.5526 │ 85.57% ┃ 0.4557 │ 88.02% ┃ +┃ 6 │ 12:21 min ┃ 8.244e-05 │ 0.5051 │ 86.71% ┃ 0.4355 │ 88.25% ┃ +┃ 7 │ 12:15 min ┃ 7.610e-05 │ 0.4705 │ 87.51% ┃ 0.4222 │ 88.87% ┃ +┃ 8 │ 12:21 min ┃ 6.911e-05 │ 0.4418 │ 88.27% ┃ 0.4100 │ 88.92% ┃ +┃ 9 │ 12:21 min ┃ 6.166e-05 │ 0.4218 │ 88.86% ┃ 0.4009 │ 89.19% ┃ +┃ 10 │ 12:21 min ┃ 5.392e-05 │ 0.4000 │ 89.48% ┃ 0.3936 │ 89.32% ┃ +┃ 11 │ 12:15 min ┃ 4.608e-05 │ 0.3863 │ 89.84% ┃ 0.3889 │ 89.52% ┃ +┃ 12 │ 12:21 min ┃ 3.834e-05 │ 0.3768 │ 90.16% ┃ 0.3840 │ 89.67% ┃ +┃ 13 │ 12:21 min ┃ 3.088e-05 │ 0.3656 │ 90.43% ┃ 0.3818 │ 89.65% ┃ +┃ 14 │ 12:21 min ┃ 2.390e-05 │ 0.3563 │ 90.79% ┃ 0.3793 │ 89.79% ┃ +┃ 15 │ 12:21 min ┃ 1.756e-05 │ 0.3502 │ 90.89% ┃ 0.3769 │ 89.78% ┃ +┃ 16 │ 12:15 min ┃ 1.202e-05 │ 0.3445 │ 91.12% ┃ 0.3762 │ 89.76% ┃ +┃ 17 │ 12:21 min ┃ 7.410e-06 │ 0.3426 │ 91.20% ┃ 0.3752 │ 89.88% ┃ +┃ 18 │ 12:20 min ┃ 3.853e-06 │ 0.3392 │ 91.26% ┃ 0.3750 │ 89.83% ┃ +┃ 19 │ 12:21 min ┃ 1.431e-06 │ 0.3372 │ 91.26% ┃ 0.3749 │ 89.84% ┃ +┃ 20 │ 12:15 min ┃ 2.052e-07 │ 0.3388 │ 91.24% ┃ 0.3749 │ 89.85% ┃ diff --git a/TinyImageNet_SGD_ViT_rhona_lr0.0001/data.hdf5 b/TinyImageNet_SGD_ViT_rhona_lr0.0001/data.hdf5 new file mode 100644 index 0000000000000000000000000000000000000000..64015a3fb2f862bf5d3a8225b42685beffeef206 Binary files /dev/null and b/TinyImageNet_SGD_ViT_rhona_lr0.0001/data.hdf5 differ diff --git a/TinyImageNet_SGD_ViT_rhona_lr0.0001/final.model b/TinyImageNet_SGD_ViT_rhona_lr0.0001/final.model new file mode 100644 index 0000000000000000000000000000000000000000..471f2ffc0b6469fcae32738a4681ab29283f1633 --- /dev/null +++ b/TinyImageNet_SGD_ViT_rhona_lr0.0001/final.model @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:210757b786fad52f11a6582271681feebaed139b549fa24e05a9f686a01e5f83 +size 701548138 diff --git a/TinyImageNet_SGD_ViT_rhona_lr0.0001/params.json b/TinyImageNet_SGD_ViT_rhona_lr0.0001/params.json new file mode 100644 index 0000000000000000000000000000000000000000..c69cf67880cccf0bad76a2c317ba2fff5facb64f --- /dev/null +++ b/TinyImageNet_SGD_ViT_rhona_lr0.0001/params.json @@ -0,0 +1,97 @@ +{ + "widthFactor": 4, + "ViTPatchsize": 32, + "ViTSize": "B", + "pyramidAlpha": 270, + "pyramidBottleneck": true, + "model": "ViT", + "dropout": 0.0, + "BN": true, + "depth": 16, + "label_smoothing": 0.1, + "ESAMbeta": 0.5, + "ESAMgamma": 0.7, + "lookSAMk": 5, + "lookSAMalpha": 0.7, + "lookSAMLayerwise": false, + "optimizer": "SGD", + "weightDecay": 0.001, + "momentum": 0.9, + "rho": 0.075, + "samRho": 0.075, + "asamRho": 1.0, + "esamRho": 0.075, + "looksamRho": 0.05, + "friendlysamRho": 0.075, + "fishersamRho": 0.5, + "mongesamRho": 0.075, + "gsamRho": 0.075, + "bayesiansamRho": 0.075, + "msamRho": 0.075, + "adamwSamRho": 0.075, + "adamwMsamRho": 0.075, + "adamwFriendlysamRho": 0.075, + "sigma": 1.0, + "lmbda": 0.95, + "fisherEta": 0.1, + "gsamAlpha": 0.1, + "gsamAdaptive": false, + "bayesMSharpness": 8, + "bayesDelta": 10.0, + "bayesBeta1": 0.9, + "bayesBeta2": 0.999, + "bayesSInit": 1.0, + "bayesDamping": 1e-08, + "bayesEps": 1e-12, + "nesterov": false, + "grad_clip_norm": 0.0, + "dataThreads": 4, + "dataDir": "/home/hungpv/.datasets", + "batchSize": 64, + "dataset": "TinyImageNet", + "imageSize": 64, + "flip": true, + "crop": true, + "cut": false, + "cutoutProp": 0.5, + "randAugment": true, + "randAugment_magnitude": 10, + "mixup": false, + "mixupProp": 0.2, + "normalize": "std", + "truncate": false, + "verbose": false, + "metrics": [ + "learningRate", + "loss", + "accuracy" + ], + "logEach": 1, + "rndSeed": false, + "deterministic": false, + "learningRate": 0.0001, + "lrScheduler": [ + "startRamp", + "cos" + ], + "LRScheduler_step_steps": 3.0, + "LRScheduler_step_gamma": 0.2, + "LRScheduler_exp_maxDecay": 0.01, + "LRScheduler_cos_periods": 0.5, + "LRScheduler_StartRamp_epochs": 1.0, + "LRScheduler_WRN_T0": 50.0, + "LRScheduler_WRN_Tmult": 1.2, + "rhoScheduler": [ + "startJump" + ], + "rhoScheduler_StartRamp_epochs": 8.0, + "rhoScheduler_StartJump_epochs": 8.0, + "saveCheckpoint": true, + "keepLastCheckpoint": false, + "checkpointsList": [], + "saveCheckpointInterval": 0, + "logDir": "/mnt/hungpv/projects/SAM_Variant/logs", + "logSubDir": "TinyImageNet_SGD_ViT_rhona_lr0.0001", + "epochs": 20, + "contin": false +} \ No newline at end of file diff --git a/TinyImageNet_SGD_ViT_rhona_lr0.0001/terminal.log b/TinyImageNet_SGD_ViT_rhona_lr0.0001/terminal.log new file mode 100644 index 0000000000000000000000000000000000000000..a10f5fbdd25792a3e4073a7fa31718553e0c7c0e --- /dev/null +++ b/TinyImageNet_SGD_ViT_rhona_lr0.0001/terminal.log @@ -0,0 +1,24 @@ +┏━━━━━━━╸S╺╸T╺╸A╺╸T╺╸S╺━━━━━━━┳━━━━━━━━━━━━━━━━━T╺╸R╺╸A╺╸I╺╸N ━━━━━━━━━━━━━━━━┳━━━━━━━━━━T╺╸E╺╸S╺╸T ━━━━━━━━━━┓ +┃ ┃ ┃ ┃ +┃ epoch │ time ┃ learningRate │ loss │ accuracy ┃ loss │ accuracy ┃ +┠──────────────┼──────────────╂───────────────┼───────────────┼───────────────╂───────────────┼───────────────┨ +┃ 1 │ 06:41 min ┃ 4.988e-05 │ 4.3148 │ 5.64% ┃ 2.2299 │ 42.43% ┃ +┃ 2 │ 05:20 min ┃ 9.857e-05 │ 1.1804 │ 69.93% ┃ 0.6607 │ 83.74% ┃ +┃ 3 │ 05:07 min ┃ 9.615e-05 │ 0.7363 │ 81.28% ┃ 0.5418 │ 86.11% ┃ +┃ 4 │ 05:06 min ┃ 9.259e-05 │ 0.6151 │ 83.90% ┃ 0.4864 │ 87.57% ┃ +┃ 5 │ 05:26 min ┃ 8.798e-05 │ 0.5419 │ 85.69% ┃ 0.4547 │ 87.96% ┃ +┃ 6 │ 05:23 min ┃ 8.244e-05 │ 0.4924 │ 86.95% ┃ 0.4387 │ 88.42% ┃ +┃ 7 │ 05:25 min ┃ 7.610e-05 │ 0.4549 │ 87.77% ┃ 0.4239 │ 88.53% ┃ +┃ 8 │ 05:26 min ┃ 6.911e-05 │ 0.4247 │ 88.61% ┃ 0.4113 │ 88.99% ┃ +┃ 9 │ 03:13 min ┃ 6.166e-05 │ 0.4043 │ 89.15% ┃ 0.4015 │ 89.02% ┃ +┃ 10 │ 02:45 min ┃ 5.392e-05 │ 0.3796 │ 89.79% ┃ 0.3970 │ 89.30% ┃ +┃ 11 │ 02:44 min ┃ 4.608e-05 │ 0.3645 │ 90.22% ┃ 0.3934 │ 89.29% ┃ +┃ 12 │ 02:44 min ┃ 3.834e-05 │ 0.3540 │ 90.55% ┃ 0.3882 │ 89.47% ┃ +┃ 13 │ 02:44 min ┃ 3.088e-05 │ 0.3410 │ 90.86% ┃ 0.3853 │ 89.46% ┃ +┃ 14 │ 02:45 min ┃ 2.390e-05 │ 0.3303 │ 91.33% ┃ 0.3832 │ 89.56% ┃ +┃ 15 │ 02:45 min ┃ 1.756e-05 │ 0.3233 │ 91.46% ┃ 0.3818 │ 89.67% ┃ +┃ 16 │ 02:45 min ┃ 1.202e-05 │ 0.3173 │ 91.67% ┃ 0.3811 │ 89.61% ┃ +┃ 17 │ 02:44 min ┃ 7.410e-06 │ 0.3146 │ 91.73% ┃ 0.3804 │ 89.70% ┃ +┃ 18 │ 02:44 min ┃ 3.853e-06 │ 0.3109 │ 91.77% ┃ 0.3798 │ 89.65% ┃ +┃ 19 │ 02:44 min ┃ 1.431e-06 │ 0.3077 │ 91.95% ┃ 0.3798 │ 89.67% ┃ +┃ 20 │ 02:44 min ┃ 2.052e-07 │ 0.3100 │ 91.80% ┃ 0.3798 │ 89.68% ┃ diff --git a/TinyImageNet_SGD_ViT_rhona_lr0.0001_run01/data.hdf5 b/TinyImageNet_SGD_ViT_rhona_lr0.0001_run01/data.hdf5 new file mode 100644 index 0000000000000000000000000000000000000000..b5246a7778f53890c025202efda7a2670208c1c3 Binary files /dev/null and b/TinyImageNet_SGD_ViT_rhona_lr0.0001_run01/data.hdf5 differ diff --git a/TinyImageNet_SGD_ViT_rhona_lr0.0001_run01/final.model b/TinyImageNet_SGD_ViT_rhona_lr0.0001_run01/final.model new file mode 100644 index 0000000000000000000000000000000000000000..6f1fc58d422fc6ad4a90e5de55e031a7613e3bc7 --- /dev/null +++ b/TinyImageNet_SGD_ViT_rhona_lr0.0001_run01/final.model @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:a2a24834184a2efc83b685ccb05f8e88ba25de76ffd0743a106378f689cde706 +size 690046570 diff --git a/TinyImageNet_SGD_ViT_rhona_lr0.0001_run01/params.json b/TinyImageNet_SGD_ViT_rhona_lr0.0001_run01/params.json new file mode 100644 index 0000000000000000000000000000000000000000..d15c12a327fc903276bf67146b359311dc953bf1 --- /dev/null +++ b/TinyImageNet_SGD_ViT_rhona_lr0.0001_run01/params.json @@ -0,0 +1,97 @@ +{ + "widthFactor": 4, + "ViTPatchsize": 16, + "ViTSize": "B", + "pyramidAlpha": 270, + "pyramidBottleneck": true, + "model": "ViT", + "dropout": 0.0, + "BN": true, + "depth": 16, + "label_smoothing": 0.1, + "ESAMbeta": 0.5, + "ESAMgamma": 0.7, + "lookSAMk": 5, + "lookSAMalpha": 0.7, + "lookSAMLayerwise": false, + "optimizer": "SGD", + "weightDecay": 0.001, + "momentum": 0.9, + "rho": 0.01, + "samRho": 0.01, + "asamRho": 1.0, + "esamRho": 0.075, + "looksamRho": 0.05, + "friendlysamRho": 0.075, + "fishersamRho": 0.1, + "mongesamRho": 0.075, + "gsamRho": 0.075, + "bayesiansamRho": 0.075, + "msamRho": 0.075, + "adamwSamRho": 0.075, + "adamwMsamRho": 0.075, + "adamwFriendlysamRho": 0.075, + "sigma": 1.0, + "lmbda": 0.95, + "fisherEta": 0.1, + "gsamAlpha": 0.1, + "gsamAdaptive": false, + "bayesMSharpness": 8, + "bayesDelta": 10.0, + "bayesBeta1": 0.9, + "bayesBeta2": 0.999, + "bayesSInit": 1.0, + "bayesDamping": 1e-08, + "bayesEps": 1e-12, + "nesterov": false, + "grad_clip_norm": 0.0, + "dataThreads": 4, + "dataDir": "/home/hungpv/.datasets", + "batchSize": 64, + "dataset": "TinyImageNet", + "imageSize": 64, + "flip": true, + "crop": true, + "cut": false, + "cutoutProp": 0.5, + "randAugment": true, + "randAugment_magnitude": 10, + "mixup": false, + "mixupProp": 0.2, + "normalize": "std", + "truncate": false, + "verbose": false, + "metrics": [ + "learningRate", + "loss", + "accuracy" + ], + "logEach": 1, + "rndSeed": false, + "deterministic": false, + "learningRate": 0.0001, + "lrScheduler": [ + "startRamp", + "cos" + ], + "LRScheduler_step_steps": 3.0, + "LRScheduler_step_gamma": 0.2, + "LRScheduler_exp_maxDecay": 0.01, + "LRScheduler_cos_periods": 0.5, + "LRScheduler_StartRamp_epochs": 1.0, + "LRScheduler_WRN_T0": 50.0, + "LRScheduler_WRN_Tmult": 1.2, + "rhoScheduler": [ + "const" + ], + "rhoScheduler_StartRamp_epochs": 8.0, + "rhoScheduler_StartJump_epochs": 8.0, + "saveCheckpoint": true, + "keepLastCheckpoint": false, + "checkpointsList": [], + "saveCheckpointInterval": 0, + "logDir": "/mnt/hungpv/projects/SAM_Variant/logs", + "logSubDir": "TinyImageNet_SGD_ViT_rhona_lr0.0001_run01", + "epochs": 20, + "contin": false +} \ No newline at end of file diff --git a/TinyImageNet_SGD_ViT_rhona_lr0.0001_run01/terminal.log b/TinyImageNet_SGD_ViT_rhona_lr0.0001_run01/terminal.log new file mode 100644 index 0000000000000000000000000000000000000000..765f3d2d62f86d964199e4764ce533a95ec9a7c2 --- /dev/null +++ b/TinyImageNet_SGD_ViT_rhona_lr0.0001_run01/terminal.log @@ -0,0 +1,24 @@ +┏━━━━━━━╸S╺╸T╺╸A╺╸T╺╸S╺━━━━━━━┳━━━━━━━━━━━━━━━━━T╺╸R╺╸A╺╸I╺╸N ━━━━━━━━━━━━━━━━┳━━━━━━━━━━T╺╸E╺╸S╺╸T ━━━━━━━━━━┓ +┃ ┃ ┃ ┃ +┃ epoch │ time ┃ learningRate │ loss │ accuracy ┃ loss │ accuracy ┃ +┠──────────────┼──────────────╂───────────────┼───────────────┼───────────────╂───────────────┼───────────────┨ +┃ 1 │ 27:36 min ┃ 4.988e-05 │ 4.4643 │ 2.27% ┃ 3.7650 │ 9.07% ┃ +┃ 2 │ 37:38 min ┃ 9.857e-05 │ 2.2237 │ 41.78% ┃ 1.0258 │ 74.62% ┃ +┃ 3 │ 39:03 min ┃ 9.615e-05 │ 0.7975 │ 80.50% ┃ 0.5586 │ 87.00% ┃ +┃ 4 │ 25:43 min ┃ 9.259e-05 │ 0.5560 │ 86.41% ┃ 0.4662 │ 88.80% ┃ +┃ 5 │ 25:47 min ┃ 8.798e-05 │ 0.4597 │ 88.67% ┃ 0.4208 │ 89.66% ┃ +┃ 6 │ 25:03 min ┃ 8.244e-05 │ 0.3989 │ 89.89% ┃ 0.3920 │ 90.39% ┃ +┃ 7 │ 22:10 min ┃ 7.610e-05 │ 0.3566 │ 90.97% ┃ 0.3794 │ 90.33% ┃ +┃ 8 │ 18:34 min ┃ 6.911e-05 │ 0.3272 │ 91.66% ┃ 0.3655 │ 90.61% ┃ +┃ 9 │ 13:46 min ┃ 6.166e-05 │ 0.3024 │ 92.38% ┃ 0.3593 │ 90.73% ┃ +┃ 10 │ 12:25 min ┃ 5.392e-05 │ 0.2797 │ 93.03% ┃ 0.3524 │ 90.84% ┃ +┃ 11 │ 12:25 min ┃ 4.608e-05 │ 0.2630 │ 93.42% ┃ 0.3519 │ 90.78% ┃ +┃ 12 │ 12:25 min ┃ 3.834e-05 │ 0.2493 │ 93.82% ┃ 0.3448 │ 91.10% ┃ +┃ 13 │ 12:25 min ┃ 3.088e-05 │ 0.2374 │ 94.18% ┃ 0.3426 │ 91.06% ┃ +┃ 14 │ 12:25 min ┃ 2.390e-05 │ 0.2287 │ 94.46% ┃ 0.3439 │ 91.10% ┃ +┃ 15 │ 12:25 min ┃ 1.756e-05 │ 0.2231 │ 94.57% ┃ 0.3406 │ 91.06% ┃ +┃ 16 │ 12:25 min ┃ 1.202e-05 │ 0.2177 │ 94.75% ┃ 0.3387 │ 91.15% ┃ +┃ 17 │ 12:25 min ┃ 7.410e-06 │ 0.2131 │ 94.91% ┃ 0.3386 │ 91.10% ┃ +┃ 18 │ 12:25 min ┃ 3.853e-06 │ 0.2110 │ 94.98% ┃ 0.3386 │ 91.13% ┃ +┃ 19 │ 12:26 min ┃ 1.431e-06 │ 0.2106 │ 94.96% ┃ 0.3383 │ 91.11% ┃ +┃ 20 │ 12:25 min ┃ 2.052e-07 │ 0.2098 │ 95.06% ┃ 0.3383 │ 91.13% ┃