fig7a_augmentations | issl Workspace

Yanndubs's group workspace

Group: fig7a_augmentations

1-3

of 3

Tags

Notes

Goal: effect of coarser augmentations (fig 7a).

Author

yanndubs

State

Finished

Start time

August 21st, 2022 9:59:08 PM

Runtime

5h 34m 52s

Tracked hours

5h 34m 49s

Run path

issl/issl_opensource/0_3930953

Linux-5.4.0-107-generic-x86_64-with-glibc2.31

Python version

3.10.4

Command

-m submitit.core._submit /juice/scr/yanndubs/Invariant-Self-Supervised-Learning/outputs/2022-08-21_14-57-05/.submitit/%j

System Hardware

CPU count	96
GPU count	1
GPU type	NVIDIA GeForce RTX 3090

W&B CLI Version

0.13.1

Group

fig7a_augmentations

Config parameters are your model's inputs. Learn more

▶
Config parameters:{} 271 keys
- callbacks/LearningRateMonitor/is_use:
  true
- callbacks/LearningRateMonitor/kwargs/logging_interval:
  "epoch"
- callbacks/RichModelSummary/is_use:
  true
- callbacks/RichModelSummary/kwargs/max_depth:
  4
- callbacks/RichProgressBar/is_use:
  false
- callbacks/RichProgressBar/kwargs/refresh_rate:
  50
- checkpoint_pred/is_load_last:
  true
- checkpoint_pred/kwargs/dirpath:
  "/juice/scr/yanndubs/Invariant-Self-Supervised-Learning/checkpoints/exp_fig7a_augmentations/datarepr_tinyimagenet/augrepr_data-standard/repr_dissl/dec_dissl/enc_resnet18/optrepr_Adam_lr2.0e-03_w1.0e-06/schedrepr_cosine/zdim_512/bs_512/ep_300/beta_1.0e-01/seed_1/addrepr_None/datapred_data_repr_0.002_test/pred_torch_logisticw1.0e-06/optpred_SGD_lr3.0e-01_w1.0e-06/schedpred_cosine_nowarm/eppred_100/bspred_512/addpred_None/jid_0_3930953"
- checkpoint_pred/kwargs/every_n_epochs:
  1
- checkpoint_pred/kwargs/mode:
  "min"
- checkpoint_pred/kwargs/monitor:
  "None"
- checkpoint_pred/kwargs/save_last:
  true
- checkpoint_pred/kwargs/save_top_k:
  1
- checkpoint_pred/kwargs/save_weights_only:
  false
- checkpoint_pred/kwargs/verbose:
  true
- checkpoint_pred/name:
  "last"
- checkpoint_repr/is_load_last:
  true
- checkpoint_repr/kwargs/dirpath:
  "/juice/scr/yanndubs/Invariant-Self-Supervised-Learning/checkpoints/exp_fig7a_augmentations/datarepr_tinyimagenet/augrepr_data-standard/repr_dissl/dec_dissl/enc_resnet18/optrepr_Adam_lr2.0e-03_w1.0e-06/schedrepr_cosine/zdim_512/bs_512/ep_300/beta_1.0e-01/seed_1/addrepr_None/datapred_data_repr_0.002_test/pred_torch_logisticw1.0e-06/optpred_SGD_lr3.0e-01_w1.0e-06/schedpred_cosine_nowarm/eppred_100/bspred_512/addpred_None/jid_0_3930953"
- checkpoint_repr/kwargs/every_n_epochs:
  1
- checkpoint_repr/kwargs/mode:
  "min"
- checkpoint_repr/kwargs/monitor:
  "None"
- checkpoint_repr/kwargs/save_last:
  true
- checkpoint_repr/kwargs/save_top_k:
  1
- checkpoint_repr/kwargs/save_weights_only:
  false
- checkpoint_repr/kwargs/verbose:
  true
- checkpoint_repr/name:
  "last"
- checkpoint/is_load_last:
  true
- checkpoint/kwargs/dirpath:
  "/juice/scr/yanndubs/Invariant-Self-Supervised-Learning/checkpoints/exp_fig7a_augmentations/datarepr_tinyimagenet/augrepr_data-standard/repr_dissl/dec_dissl/enc_resnet18/optrepr_Adam_lr2.0e-03_w1.0e-06/schedrepr_cosine/zdim_512/bs_512/ep_300/beta_1.0e-01/seed_1/addrepr_None/datapred_data_repr_0.002_test/pred_torch_logisticw1.0e-06/optpred_SGD_lr3.0e-01_w1.0e-06/schedpred_cosine_nowarm/eppred_100/bspred_512/addpred_None/jid_0_3930953"
- checkpoint/kwargs/every_n_epochs:
  1
- checkpoint/kwargs/mode:
  "min"
- checkpoint/kwargs/monitor:
  "None"
- checkpoint/kwargs/save_last:
  true
- checkpoint/kwargs/save_top_k:
  1
- checkpoint/kwargs/save_weights_only:
  false
- checkpoint/kwargs/verbose:
  true
- checkpoint/name:
  "last"
- data_pred/aux_shape:
  "???"
- data_pred/aux_target:
  "???"
- data_pred/dataset:
  "tinyimagenet"
- data_pred/is_copy_repr:
  true
- data_pred/kwargs/batch_size:
  512
- data_pred/kwargs/data_dir:
  "/juice/scr/yanndubs/Invariant-Self-Supervised-Learning/data"
- ▶
  data_pred/kwargs/dataset_kwargs/a_augmentations:[] 1 item
  - 0:
    "data-standard"
- data_pred/kwargs/dataset_kwargs/aux_target:
  "None"
- data_pred/kwargs/dataset_kwargs/is_normalize:
  true
- data_pred/kwargs/dataset_kwargs/seed:
  1
- update_trainer_pred/max_epochs:
  1,000
- update_trainer_repr/max_epochs:
  300
- user:
  "yanndubs"
- wandb_entity:
  "issl"

Summary metrics are your model's outputs. Learn more

▶
Summary metrics:{} 86 keys
- epoch:
  999
- lr_evaluator:
  0.000003
- lr_issl:
  0.00000011735443377381
- lr_predictor:
  0.00000148043944253606
- test/pred/torchlogistic_datarepr/acc:
  0.4422000050544739
- test/pred/torchlogistic_datarepr/err:
  0.5577999949455261
- test/pred/torchlogistic_datarepr/loss:
  2.437535047531128
- test/pred/torchlogistic_datarepr001test/acc:
  0.17669999599456787
- test/pred/torchlogistic_datarepr01test/acc:
  0.33250001072883606
- test/pred/torchlogistic_datarepr001test/err:
  0.8233000040054321
- test/pred/torchlogistic_datarepr01test/err:
  0.6675000190734863
- test/pred/torchlogistic_datarepr001test/loss:
  4.246657848358154
- test/pred/torchlogistic_datarepr01test/loss:
  3.2099509239196777
- test/pred/torchlogistic_datarepr0002test/acc:
  0.0754999965429306
- test/pred/torchlogistic_datarepr0002test/err:
  0.9244999885559082
- test/pred/torchlogistic_datarepr0002test/loss:
  5.7552056312561035
- test/pred/torchlogisticw1e-5_datarepr/acc:
  0.4496000111103058
- test/pred/torchlogisticw1e-5_datarepr/err:
  0.5504000186920166
- test/pred/torchlogisticw1e-5_datarepr/loss:
  2.3546643257141113
- test/repr/tinyimagenet/CE_distill_aug:
  4.1249237060546875
- test/repr/tinyimagenet/cosine_neg:
  0.00773619394749403
- test/repr/tinyimagenet/cosine_pos:
  0.8467249870300293
- test/repr/tinyimagenet/dist_to_etf:
  0.16101562976837158
- test/repr/tinyimagenet/eval_acc:
  0.4205000102519989
- test/repr/tinyimagenet/eval_err:
  0.5795000195503235
- test/repr/tinyimagenet/eval_loss:
  2.514723777770996
- test/repr/tinyimagenet/H_M:
  8.032798767089844
- test/repr/tinyimagenet/H_Mlz:
  3.0000643730163574
- test/repr/tinyimagenet/KL_distillation:
  0.2111537903547287
- test/repr/tinyimagenet/KL_invariance:
  1.2372931241989136
- test/repr/tinyimagenet/loss:
  -10.960625648498535
- train/pred/torchlogistic_datarepr/acc:
  0.625
- train/pred/torchlogistic_datarepr/err:
  0.375
- train/pred/torchlogistic_datarepr/loss:
  1.4515994787216189
- train/pred/torchlogistic_datarepr001test/acc:
  1
- train/pred/torchlogistic_datarepr01test/acc:
  0.966911792755127
- train/pred/torchlogistic_datarepr001test/err:
  0
- train/pred/torchlogistic_datarepr01test/err:
  0.03308820724487305
- train/pred/torchlogistic_datarepr001test/loss:
  0.09278981387615204
- train/pred/torchlogistic_datarepr01test/loss:
  0.4164687395095825
- train/pred/torchlogistic_datarepr0002test/acc:
  1
- train/pred/torchlogistic_datarepr0002test/err:
  0
- train/pred/torchlogistic_datarepr0002test/loss:
  0.0010066747199743986
- train/pred/torchlogisticw1e-5_datarepr/acc:
  0.6187499761581421
- train/pred/torchlogisticw1e-5_datarepr/err:
  0.3812500238418579
- train/pred/torchlogisticw1e-5_datarepr/loss:
  1.3806860446929932
- val/repr/tinyimagenet/H_Mlz:
  2.9980344772338867
- val/repr/tinyimagenet/KL_distillation:
  0.2110548764467239
- val/repr/tinyimagenet/KL_invariance:
  1.2418144941329956
- val/repr/tinyimagenet/loss:
  -10.951671600341797

This run produced these artifacts as outputs. Learn more

code

source-issl_opensource-None:v10

code

source-issl_opensource-None:v11

wandb-history

run-0_3930953-history:v0