diff options
Diffstat (limited to 'results/vit_test/snapshot_vit_s42.json')
| -rw-r--r-- | results/vit_test/snapshot_vit_s42.json | 244 |
1 files changed, 244 insertions, 0 deletions
diff --git a/results/vit_test/snapshot_vit_s42.json b/results/vit_test/snapshot_vit_s42.json new file mode 100644 index 0000000..f0095db --- /dev/null +++ b/results/vit_test/snapshot_vit_s42.json @@ -0,0 +1,244 @@ +{ + "config": { + "output_dir": "results/vit_test", + "epochs": 2, + "lr": 0.001, + "wd": 0.05, + "seed": 42, + "depth": 4, + "d_model": 128, + "n_heads": 4 + }, + "depth": 4, + "d_model": 128, + "architecture": "ViTMini", + "bp_log": [ + { + "hidden_norms_cls": [ + 0.31052109599113464, + 3.2536301612854004, + 4.7681732177734375, + 6.129599571228027, + 7.124356746673584 + ], + "hidden_norms_avg": [ + 6.008513927459717, + 7.034290313720703, + 7.942849636077881, + 8.915729522705078, + 9.635361671447754 + ], + "bp_grad_per_sample_l2_med": [ + 0.004442029166966677, + 0.0012008437188342214, + 0.0009976272704079747, + 0.0009104337659664452, + 0.0008558662957511842 + ], + "bp_grad_F": [ + 0.18535660207271576, + 0.03968670591711998, + 0.032604776322841644, + 0.029427478089928627, + 0.02746850810945034 + ], + "gamma_dfa": NaN, + "gamma_dfa_per_layer": [], + "acc_eval": 0.1376953125, + "loss_eval": 2.35188627243042, + "epoch": 0 + }, + { + "hidden_norms_cls": [ + 0.5119581818580627, + 12.45346450805664, + 18.367006301879883, + 25.541183471679688, + 31.80698585510254 + ], + "hidden_norms_avg": [ + 4.10694694519043, + 10.166566848754883, + 14.66998291015625, + 19.933429718017578, + 25.32620620727539 + ], + "bp_grad_per_sample_l2_med": [ + 0.001977165462449193, + 0.0002232444821856916, + 0.00016308759222738445, + 0.00014621141599491239, + 0.0001451292773708701 + ], + "bp_grad_F": [ + 0.11066912859678268, + 0.010344989597797394, + 0.006372526753693819, + 0.005224850494414568, + 0.005017552524805069 + ], + "gamma_dfa": NaN, + "gamma_dfa_per_layer": [], + "acc_eval": 0.4501953125, + "loss_eval": 1.5154452323913574, + "epoch": 1 + }, + { + "hidden_norms_cls": [ + 0.5624178647994995, + 10.294456481933594, + 15.314032554626465, + 21.36781883239746, + 25.580961227416992 + ], + "hidden_norms_avg": [ + 3.8611817359924316, + 9.257936477661133, + 13.537341117858887, + 17.418514251708984, + 20.89995002746582 + ], + "bp_grad_per_sample_l2_med": [ + 0.0021491278894245625, + 0.0003653931198641658, + 0.00022418916341848671, + 0.00018154713325202465, + 0.00017043459229171276 + ], + "bp_grad_F": [ + 0.10512559115886688, + 0.014831307344138622, + 0.008541438728570938, + 0.006647223141044378, + 0.005966350436210632 + ], + "gamma_dfa": NaN, + "gamma_dfa_per_layer": [], + "acc_eval": 0.52734375, + "loss_eval": 1.3002638816833496, + "epoch": 2 + } + ], + "dfa_log": [ + { + "hidden_norms_cls": [ + 0.31052109599113464, + 3.2536301612854004, + 4.7681732177734375, + 6.129599571228027, + 7.124356746673584 + ], + "hidden_norms_avg": [ + 6.008513927459717, + 7.034290313720703, + 7.942849636077881, + 8.915729522705078, + 9.635361671447754 + ], + "bp_grad_per_sample_l2_med": [ + 0.004442029166966677, + 0.0012008437188342214, + 0.0009976272704079747, + 0.0009104337659664452, + 0.0008558662957511842 + ], + "bp_grad_F": [ + 0.18535660207271576, + 0.03968670591711998, + 0.032604776322841644, + 0.029427478089928627, + 0.02746850810945034 + ], + "gamma_dfa": 0.003905070887412876, + "gamma_dfa_per_layer": [ + 0.004920602310448885, + 0.005249542184174061, + 0.0038580193649977446, + 0.0015921196900308132 + ], + "acc_eval": 0.1376953125, + "loss_eval": 2.35188627243042, + "epoch": 0 + }, + { + "hidden_norms_cls": [ + 1.9684416055679321, + 2029.23681640625, + 2826.160400390625, + 4672.1005859375, + 8737.1298828125 + ], + "hidden_norms_avg": [ + 45.1302375793457, + 3068.238037109375, + 8890.5302734375, + 11475.16015625, + 14973.22265625 + ], + "bp_grad_per_sample_l2_med": [ + 0.0002814636391121894, + 7.465861813216179e-07, + 6.924062745383708e-07, + 6.833349175394687e-07, + 6.810931267864362e-07 + ], + "bp_grad_F": [ + 0.08884845674037933, + 7.258133700815961e-05, + 4.2221599869662896e-05, + 2.3938106096466072e-05, + 2.3418680939357728e-05 + ], + "gamma_dfa": 0.0073737858911044896, + "gamma_dfa_per_layer": [ + 0.0056831855326890945, + 0.015363219194114208, + 0.009368096478283405, + -0.0009193576406687498 + ], + "acc_eval": 0.23046875, + "loss_eval": 2.054917335510254, + "epoch": 1 + }, + { + "hidden_norms_cls": [ + 3.0175483226776123, + 3912.425048828125, + 6017.75830078125, + 8192.2646484375, + 19695.205078125 + ], + "hidden_norms_avg": [ + 70.26815032958984, + 6088.02001953125, + 20950.01953125, + 25550.255859375, + 33917.15625 + ], + "bp_grad_per_sample_l2_med": [ + 0.000260332744801417, + 3.430939727877558e-07, + 3.090381710535439e-07, + 3.0240653359214775e-07, + 2.9979452165207476e-07 + ], + "bp_grad_F": [ + 0.25490984320640564, + 2.9280510716489516e-05, + 1.6399520973209292e-05, + 9.804005458136089e-06, + 9.64422997640213e-06 + ], + "gamma_dfa": 0.007508119277190417, + "gamma_dfa_per_layer": [ + 0.007269312161952257, + 0.013389883562922478, + 0.01044462900608778, + -0.0010713476222008467 + ], + "acc_eval": 0.2265625, + "loss_eval": 2.0774288177490234, + "epoch": 2 + } + ] +}
\ No newline at end of file |
