Add files using upload-large-folder tool

schaefer1000/schaefer1000_lr3e-4_1/eval_v2/nsd_cococlip__patch__attn/config.yaml

@@ -0,0 +1,96 @@
+output_root: experiments/schaefer1000/output
+name_prefix: eval_probe
+remote_root: null
+notes: schaefer1000 ablation schaefer1000_lr3e-4_1; eval v2 (nsd_cococlip patch attn)
+model_kwargs:
+  ckpt_path: experiments/schaefer1000/output/schaefer1000/schaefer1000_lr3e-4_1/pretrain/checkpoint-last.pth
+dataset_kwargs: {}
+classifier_kwargs:
+  embed_dim: null
+  dropout: 0.0
+  xavier_init: true
+  norm: true
+lr_scale_grid:
+- 0.02
+- 0.023
+- 0.028
+- 0.033
+- 0.038
+- 0.045
+- 0.053
+- 0.062
+- 0.074
+- 0.087
+- 0.1
+- 0.12
+- 0.14
+- 0.17
+- 0.2
+- 0.23
+- 0.27
+- 0.32
+- 0.38
+- 0.44
+- 0.52
+- 0.61
+- 0.72
+- 0.85
+- 1
+- 1.2
+- 1.4
+- 1.6
+- 1.9
+- 2.3
+- 2.7
+- 3.1
+- 3.7
+- 4.3
+- 5.1
+- 6
+- 7.1
+- 8.3
+- 9.8
+- 12
+- 14
+- 16
+- 19
+- 22
+- 26
+- 31
+- 36
+- 43
+- 50
+wd_scale_grid:
+- 1.0
+num_workers: 8
+prefetch_factor: null
+balanced_sampling: false
+epochs: 20
+steps_per_epoch: 200
+batch_size: 64
+accum_iter: 2
+lr: 0.0003
+warmup_epochs: 5
+no_decay: false
+weight_decay: 0.05
+clip_grad: 1.0
+metrics:
+- acc
+- f1
+cv_metric: acc
+early_stopping: true
+amp: true
+device: cuda
+seed: 4466
+debug: false
+wandb: false
+wandb_entity: null
+wandb_project: fMRI-fm-eval
+name: schaefer1000/schaefer1000_lr3e-4_1/eval_v2/nsd_cococlip__patch__attn
+model: schaefer1000_mae
+representation: patch
+classifier: attn
+dataset: nsd_cococlip
+distributed: false
+output_dir: experiments/schaefer1000/output/schaefer1000/schaefer1000_lr3e-4_1/eval_v2/nsd_cococlip__patch__attn
+remote_dir: null

schaefer1000/schaefer1000_lr3e-4_1/eval_v2/nsd_cococlip__patch__attn/eval_log.json

@@ -0,0 +1 @@
+{"eval/epoch": 13, "eval/id_best": 30, "eval/lr_best": 0.00081, "eval/wd_best": 0.05, "eval/train/loss": 1.835350751876831, "eval/train/acc": 0.4404560681028919, "eval/train/acc_std": 0.002547147642705453, "eval/train/f1": 0.40620007402855185, "eval/train/f1_std": 0.0027522048935764583, "eval/validation/loss": 2.356290817260742, "eval/validation/acc": 0.2978959025470653, "eval/validation/acc_std": 0.005214840599996058, "eval/validation/f1": 0.24268757190837384, "eval/validation/f1_std": 0.005156419585262203, "eval/test/loss": 2.2967960834503174, "eval/test/acc": 0.31020408163265306, "eval/test/acc_std": 0.005662234319688578, "eval/test/f1": 0.2486825032274845, "eval/test/f1_std": 0.005550877940304615, "eval/testid/loss": 2.3030717372894287, "eval/testid/acc": 0.3067283593599383, "eval/testid/acc_std": 0.005608810532786223, "eval/testid/f1": 0.2708750197577065, "eval/testid/f1_std": 0.0056971651946896105}

schaefer1000/schaefer1000_lr3e-4_1/eval_v2/nsd_cococlip__patch__attn/eval_log_best.json

@@ -0,0 +1 @@
+{"eval/best/epoch": 13, "eval/best/id_best": 30, "eval/best/lr_best": 0.00081, "eval/best/wd_best": 0.05, "eval/best/train/loss": 1.835350751876831, "eval/best/train/acc": 0.4404560681028919, "eval/best/train/acc_std": 0.002547147642705453, "eval/best/train/f1": 0.40620007402855185, "eval/best/train/f1_std": 0.0027522048935764583, "eval/best/validation/loss": 2.356290817260742, "eval/best/validation/acc": 0.2978959025470653, "eval/best/validation/acc_std": 0.005214840599996058, "eval/best/validation/f1": 0.24268757190837384, "eval/best/validation/f1_std": 0.005156419585262203, "eval/best/test/loss": 2.2967960834503174, "eval/best/test/acc": 0.31020408163265306, "eval/best/test/acc_std": 0.005662234319688578, "eval/best/test/f1": 0.2486825032274845, "eval/best/test/f1_std": 0.005550877940304615, "eval/best/testid/loss": 2.3030717372894287, "eval/best/testid/acc": 0.3067283593599383, "eval/best/testid/acc_std": 0.005608810532786223, "eval/best/testid/f1": 0.2708750197577065, "eval/best/testid/f1_std": 0.0056971651946896105}

schaefer1000/schaefer1000_lr3e-4_1/eval_v2/nsd_cococlip__patch__attn/eval_log_last.json

@@ -0,0 +1 @@
+{"eval/last/epoch": 19, "eval/last/id_best": 30, "eval/last/lr_best": 0.00081, "eval/last/wd_best": 0.05, "eval/last/train/loss": 1.7259238958358765, "eval/last/train/acc": 0.47490703463536066, "eval/last/train/acc_std": 0.0025508454788737926, "eval/last/train/f1": 0.4378632616293782, "eval/last/train/f1_std": 0.0028041043029554554, "eval/last/validation/loss": 2.384547472000122, "eval/last/validation/acc": 0.29457364341085274, "eval/last/validation/acc_std": 0.005485248113508287, "eval/last/validation/f1": 0.23590938378982018, "eval/last/validation/f1_std": 0.005390696811721903, "eval/last/test/loss": 2.3253486156463623, "eval/last/test/acc": 0.31131725417439704, "eval/last/test/acc_std": 0.00572904451427706, "eval/last/test/f1": 0.25199704175118215, "eval/last/test/f1_std": 0.005796866290642955, "eval/last/testid/loss": 2.2896580696105957, "eval/last/testid/acc": 0.30884904569115096, "eval/last/testid/acc_std": 0.0056364674667000285, "eval/last/testid/f1": 0.2704378012287932, "eval/last/testid/f1_std": 0.005768751533755878}

schaefer1000/schaefer1000_lr3e-4_1/eval_v2/nsd_cococlip__patch__attn/eval_table.csv

@@ -0,0 +1,5 @@
+model,repr,clf,dataset,ckpt,epoch,lr,wd,hparam_id,hparam,split,loss,acc,acc_std,f1,f1_std
+schaefer1000_mae,patch,attn,nsd_cococlip,best,13,0.00081,0.05,30,"[2.7, 1.0]",train,1.835350751876831,0.4404560681028919,0.002547147642705453,0.40620007402855185,0.0027522048935764583
+schaefer1000_mae,patch,attn,nsd_cococlip,best,13,0.00081,0.05,30,"[2.7, 1.0]",validation,2.356290817260742,0.2978959025470653,0.005214840599996058,0.24268757190837384,0.005156419585262203
+schaefer1000_mae,patch,attn,nsd_cococlip,best,13,0.00081,0.05,30,"[2.7, 1.0]",test,2.2967960834503174,0.31020408163265306,0.005662234319688578,0.2486825032274845,0.005550877940304615
+schaefer1000_mae,patch,attn,nsd_cococlip,best,13,0.00081,0.05,30,"[2.7, 1.0]",testid,2.3030717372894287,0.3067283593599383,0.005608810532786223,0.2708750197577065,0.0056971651946896105

schaefer1000/schaefer1000_lr3e-4_1/eval_v2/nsd_cococlip__patch__attn/eval_table_best.csv

@@ -0,0 +1,5 @@
+model,repr,clf,dataset,ckpt,epoch,lr,wd,hparam_id,hparam,split,loss,acc,acc_std,f1,f1_std
+schaefer1000_mae,patch,attn,nsd_cococlip,best,13,0.00081,0.05,30,"[2.7, 1.0]",train,1.835350751876831,0.4404560681028919,0.002547147642705453,0.40620007402855185,0.0027522048935764583
+schaefer1000_mae,patch,attn,nsd_cococlip,best,13,0.00081,0.05,30,"[2.7, 1.0]",validation,2.356290817260742,0.2978959025470653,0.005214840599996058,0.24268757190837384,0.005156419585262203
+schaefer1000_mae,patch,attn,nsd_cococlip,best,13,0.00081,0.05,30,"[2.7, 1.0]",test,2.2967960834503174,0.31020408163265306,0.005662234319688578,0.2486825032274845,0.005550877940304615
+schaefer1000_mae,patch,attn,nsd_cococlip,best,13,0.00081,0.05,30,"[2.7, 1.0]",testid,2.3030717372894287,0.3067283593599383,0.005608810532786223,0.2708750197577065,0.0056971651946896105

schaefer1000/schaefer1000_lr3e-4_1/eval_v2/nsd_cococlip__patch__attn/eval_table_last.csv

@@ -0,0 +1,5 @@
+model,repr,clf,dataset,ckpt,epoch,lr,wd,hparam_id,hparam,split,loss,acc,acc_std,f1,f1_std
+schaefer1000_mae,patch,attn,nsd_cococlip,last,19,0.00081,0.05,30,"[2.7, 1.0]",train,1.7259238958358765,0.47490703463536066,0.0025508454788737926,0.4378632616293782,0.0028041043029554554
+schaefer1000_mae,patch,attn,nsd_cococlip,last,19,0.00081,0.05,30,"[2.7, 1.0]",validation,2.384547472000122,0.29457364341085274,0.005485248113508287,0.23590938378982018,0.005390696811721903
+schaefer1000_mae,patch,attn,nsd_cococlip,last,19,0.00081,0.05,30,"[2.7, 1.0]",test,2.3253486156463623,0.31131725417439704,0.00572904451427706,0.25199704175118215,0.005796866290642955
+schaefer1000_mae,patch,attn,nsd_cococlip,last,19,0.00081,0.05,30,"[2.7, 1.0]",testid,2.2896580696105957,0.30884904569115096,0.0056364674667000285,0.2704378012287932,0.005768751533755878

schaefer1000/schaefer1000_lr3e-4_1/eval_v2/nsd_cococlip__patch__attn/log.txt

@@ -0,0 +1,966 @@
+fMRI foundation model probe eval
+version: 0.1.dev95+g65be98d36
+sha: 87e31aaa465443ed5f0da58176ac8395447cdbd0, status: has uncommitted changes, branch: dev/clane9
+cwd: /data/connor/fmri-fm
+start: 2026-05-12 20:54:50
+config:
+output_root: experiments/schaefer1000/output
+name_prefix: eval_probe
+remote_root: null
+notes: schaefer1000 ablation schaefer1000_lr3e-4_1; eval v2 (nsd_cococlip patch attn)
+model_kwargs:
+  ckpt_path: experiments/schaefer1000/output/schaefer1000/schaefer1000_lr3e-4_1/pretrain/checkpoint-last.pth
+dataset_kwargs: {}
+classifier_kwargs:
+  embed_dim: null
+  dropout: 0.0
+  xavier_init: true
+  norm: true
+lr_scale_grid:
+- 0.02
+- 0.023
+- 0.028
+- 0.033
+- 0.038
+- 0.045
+- 0.053
+- 0.062
+- 0.074
+- 0.087
+- 0.1
+- 0.12
+- 0.14
+- 0.17
+- 0.2
+- 0.23
+- 0.27
+- 0.32
+- 0.38
+- 0.44
+- 0.52
+- 0.61
+- 0.72
+- 0.85
+- 1
+- 1.2
+- 1.4
+- 1.6
+- 1.9
+- 2.3
+- 2.7
+- 3.1
+- 3.7
+- 4.3
+- 5.1
+- 6
+- 7.1
+- 8.3
+- 9.8
+- 12
+- 14
+- 16
+- 19
+- 22
+- 26
+- 31
+- 36
+- 43
+- 50
+wd_scale_grid:
+- 1.0
+num_workers: 8
+prefetch_factor: null
+balanced_sampling: false
+epochs: 20
+steps_per_epoch: 200
+batch_size: 64
+accum_iter: 2
+lr: 0.0003
+warmup_epochs: 5
+no_decay: false
+weight_decay: 0.05
+clip_grad: 1.0
+metrics:
+- acc
+- f1
+cv_metric: acc
+early_stopping: true
+amp: true
+device: cuda
+seed: 4466
+debug: false
+wandb: false
+wandb_entity: null
+wandb_project: fMRI-fm-eval
+name: schaefer1000/schaefer1000_lr3e-4_1/eval_v2/nsd_cococlip__patch__attn
+model: schaefer1000_mae
+representation: patch
+classifier: attn
+dataset: nsd_cococlip
+distributed: false
+output_dir: experiments/schaefer1000/output/schaefer1000/schaefer1000_lr3e-4_1/eval_v2/nsd_cococlip__patch__attn
+remote_dir: null
+
+creating frozen backbone model: schaefer1000_mae
+backbone:
+MaskedEncoderWrapper(
+  (model): MaskedEncoder(
+    class_token=True, reg_tokens=0, no_embed_class=True, mask_drop_scale=False
+    (patchify): Patchify3D((16, 1000, 1), (4, 1, 1), in_chans=1)
+    (patch_embed): Linear(in_features=4, out_features=768, bias=True)
+    (pos_embed): SeparablePosEmbed(768, (4, 1000, 1))
+    (blocks): ModuleList(
+      (0-11): 12 x Block(
+        (norm1): LayerNorm((768,), eps=1e-06, elementwise_affine=True)
+        (attn): Attention(
+          num_heads=12
+          (q): Linear(in_features=768, out_features=768, bias=True)
+          (k): Linear(in_features=768, out_features=768, bias=True)
+          (v): Linear(in_features=768, out_features=768, bias=True)
+          (proj): Linear(in_features=768, out_features=768, bias=True)
+        )
+        (drop_path1): Identity()
+        (norm2): LayerNorm((768,), eps=1e-06, elementwise_affine=True)
+        (mlp): Mlp(
+          (fc1): Linear(in_features=768, out_features=3072, bias=True)
+          (act): GELU(approximate='none')
+          (fc2): Linear(in_features=3072, out_features=768, bias=True)
+        )
+        (drop_path2): Identity()
+      )
+    )
+    (norm): LayerNorm((768,), eps=1e-06, elementwise_affine=True)
+  )
+)
+creating dataset: nsd_cococlip (schaefer1000)
+train (n=32539):
+HFDataset(
+    dataset=Dataset({
+    features: ['sub', 'ses', 'run', 'trial_id', 'nsd_id', 'category_id', 'path', 'start', 'end', 'n_frames', 'tr', 'bold', 'mean', 'std'],
+    num_rows: 32539
+}),
+    labels=[ 3  4  5  6 10 11 17 18 19 20 22 23 25 30 31 33 36 37 38 53 55 59 61 74],
+    counts=[1286 1180 1639 1868  834  824 1026 1042  913 1853 1503 2092 1001 1410
+  794 1241 1904 1872 2267 1428  889  904 1447 1322]
+)
+
+validation (n=5418):
+HFDataset(
+    dataset=Dataset({
+    features: ['sub', 'ses', 'run', 'trial_id', 'nsd_id', 'category_id', 'path', 'start', 'end', 'n_frames', 'tr', 'bold', 'mean', 'std'],
+    num_rows: 5418
+}),
+    labels=[ 3  4  5  6 10 11 17 18 19 20 22 23 25 30 31 33 36 37 38 53 55 59 61 74],
+    counts=[197 161 276 345 126 142 143 185 112 295 285 387 169 250 159 193 316 334
+ 343 215 172 141 226 246]
+)
+
+test (n=5390):
+HFDataset(
+    dataset=Dataset({
+    features: ['sub', 'ses', 'run', 'trial_id', 'nsd_id', 'category_id', 'path', 'start', 'end', 'n_frames', 'tr', 'bold', 'mean', 'std'],
+    num_rows: 5390
+}),
+    labels=[ 3  4  5  6 10 11 17 18 19 20 22 23 25 30 31 33 36 37 38 53 55 59 61 74],
+    counts=[202 172 274 298 144 180 134 182 186 293 218 343 165 185 140 177 346 333
+ 345 271 165 140 251 246]
+)
+
+testid (n=5187):
+HFDataset(
+    dataset=Dataset({
+    features: ['sub', 'ses', 'run', 'trial_id', 'nsd_id', 'category_id', 'path', 'start', 'end', 'n_frames', 'tr', 'bold', 'mean', 'std'],
+    num_rows: 5187
+}),
+    labels=[ 3  4  5  6 10 11 17 18 19 20 22 23 25 30 31 33 36 37 38 53 55 59 61 74],
+    counts=[197 159 267 273 123 153 175 184 139 310 215 386 153 230 118 192 330 306
+ 349 223 143 127 249 186]
+)
+
+running backbone on example batch to get embedding dim
+embedding feature dim (patch): 768
+initializing sweep of classifier heads
+classifiers:
+ModuleList(
+  (0-48): 49 x AttnPoolClassifier(
+    (kv): Linear(in_features=768, out_features=1536, bias=True)
+    (linear): Linear(in_features=768, out_features=24, bias=True)
+  )
+)
+classifier params (train): 58.8M (58.8M)
+setting up optimizer
+total batch size: 128 = 64 bs per gpu x 2 accum
+lr: 3.00e-04
+full schedule: epochs = 20 (steps = 4000) (decay = True)
+warmup: epochs = 5 (steps = 1000)
+start training for 20 epochs
+train: [0]  [  0/400]  eta: 0:10:15  lr: nan  time: 1.5397  data: 0.7088  max mem: 56639
+train: [0]  [ 20/400]  eta: 0:04:24  lr: 0.000003  loss: 3.1865 (3.1828)  grad: 0.1803 (0.1846)  time: 0.6525  data: 0.0023  max mem: 57344
+train: [0]  [ 40/400]  eta: 0:04:01  lr: 0.000006  loss: 3.1650 (3.1727)  grad: 0.1751 (0.1774)  time: 0.6458  data: 0.0034  max mem: 57344
+train: [0]  [ 60/400]  eta: 0:03:45  lr: 0.000009  loss: 3.1623 (3.1710)  grad: 0.1769 (0.1798)  time: 0.6458  data: 0.0036  max mem: 57344
+train: [0]  [ 80/400]  eta: 0:03:30  lr: 0.000012  loss: 3.1532 (3.1660)  grad: 0.1771 (0.1783)  time: 0.6470  data: 0.0036  max mem: 57344
+train: [0]  [100/400]  eta: 0:03:16  lr: 0.000015  loss: 3.1495 (3.1647)  grad: 0.1714 (0.1784)  time: 0.6471  data: 0.0036  max mem: 57344
+train: [0]  [120/400]  eta: 0:03:03  lr: 0.000018  loss: 3.1528 (3.1641)  grad: 0.1714 (0.1771)  time: 0.6469  data: 0.0035  max mem: 57344
+train: [0]  [140/400]  eta: 0:02:49  lr: 0.000021  loss: 3.1452 (3.1615)  grad: 0.1725 (0.1767)  time: 0.6468  data: 0.0035  max mem: 57344
+train: [0]  [160/400]  eta: 0:02:36  lr: 0.000024  loss: 3.1603 (3.1615)  grad: 0.1662 (0.1756)  time: 0.6472  data: 0.0035  max mem: 57344
+train: [0]  [180/400]  eta: 0:02:23  lr: 0.000027  loss: 3.1616 (3.1607)  grad: 0.1601 (0.1741)  time: 0.6474  data: 0.0035  max mem: 57344
+train: [0]  [200/400]  eta: 0:02:10  lr: 0.000030  loss: 3.1402 (3.1580)  grad: 0.1625 (0.1735)  time: 0.6472  data: 0.0035  max mem: 57344
+train: [0]  [220/400]  eta: 0:01:57  lr: 0.000033  loss: 3.1225 (3.1556)  grad: 0.1741 (0.1741)  time: 0.6473  data: 0.0035  max mem: 57344
+train: [0]  [240/400]  eta: 0:01:44  lr: 0.000036  loss: 3.1197 (3.1545)  grad: 0.1728 (0.1737)  time: 0.6475  data: 0.0035  max mem: 57344
+train: [0]  [260/400]  eta: 0:01:31  lr: 0.000039  loss: 3.1372 (3.1540)  grad: 0.1651 (0.1729)  time: 0.6477  data: 0.0036  max mem: 57344
+train: [0]  [280/400]  eta: 0:01:18  lr: 0.000042  loss: 3.1263 (3.1513)  grad: 0.1658 (0.1724)  time: 0.6469  data: 0.0035  max mem: 57344
+train: [0]  [300/400]  eta: 0:01:05  lr: 0.000045  loss: 3.1349 (3.1508)  grad: 0.1687 (0.1721)  time: 0.6474  data: 0.0036  max mem: 57344
+train: [0]  [320/400]  eta: 0:00:52  lr: 0.000048  loss: 3.1294 (3.1492)  grad: 0.1725 (0.1727)  time: 0.6477  data: 0.0036  max mem: 57344
+train: [0]  [340/400]  eta: 0:00:39  lr: 0.000051  loss: 3.1074 (3.1473)  grad: 0.1740 (0.1724)  time: 0.6478  data: 0.0036  max mem: 57344
+train: [0]  [360/400]  eta: 0:00:25  lr: 0.000054  loss: 3.1162 (3.1467)  grad: 0.1706 (0.1724)  time: 0.6472  data: 0.0035  max mem: 57344
+train: [0]  [380/400]  eta: 0:00:12  lr: 0.000057  loss: 3.1162 (3.1444)  grad: 0.1680 (0.1724)  time: 0.6476  data: 0.0036  max mem: 57344
+train: [0]  [399/400]  eta: 0:00:00  lr: 0.000060  loss: 3.0986 (3.1423)  grad: 0.1682 (0.1724)  time: 0.6472  data: 0.0034  max mem: 57344
+train: [0] Total time: 0:04:19 (0.6499 s / it)
+train: [0] Summary: lr: 0.000060  loss: 3.0986 (3.1423)  grad: 0.1682 (0.1724)
+eval (validation): [0]  [ 0/85]  eta: 0:01:15    time: 0.8924  data: 0.5356  max mem: 57344
+eval (validation): [0]  [20/85]  eta: 0:00:25    time: 0.3695  data: 0.0030  max mem: 57344
+eval (validation): [0]  [40/85]  eta: 0:00:17    time: 0.3703  data: 0.0033  max mem: 57344
+eval (validation): [0]  [60/85]  eta: 0:00:09    time: 0.3703  data: 0.0033  max mem: 57344
+eval (validation): [0]  [80/85]  eta: 0:00:01    time: 0.3701  data: 0.0032  max mem: 57344
+eval (validation): [0]  [84/85]  eta: 0:00:00    time: 0.3637  data: 0.0032  max mem: 57344
+eval (validation): [0] Total time: 0:00:31 (0.3759 s / it)
+cv: [0]  best hparam: (43, 1.0) (047) ('047_lr4.3e+01_wd1.0e+00')  loss: 2.688  acc: 0.202  f1: 0.139
+saving checkpoint experiments/schaefer1000/output/schaefer1000/schaefer1000_lr3e-4_1/eval_v2/nsd_cococlip__patch__attn/checkpoint-last.pth
+saving best checkpoint experiments/schaefer1000/output/schaefer1000/schaefer1000_lr3e-4_1/eval_v2/nsd_cococlip__patch__attn/checkpoint-best.pth
+train: [1]  [  0/400]  eta: 0:08:15  lr: nan  time: 1.2380  data: 0.6039  max mem: 57344
+train: [1]  [ 20/400]  eta: 0:04:16  lr: 0.000063  loss: 3.0282 (3.0642)  grad: 0.1675 (0.1750)  time: 0.6461  data: 0.0031  max mem: 57344
+train: [1]  [ 40/400]  eta: 0:03:57  lr: 0.000066  loss: 3.0398 (3.0675)  grad: 0.1675 (0.1711)  time: 0.6472  data: 0.0033  max mem: 57344
+train: [1]  [ 60/400]  eta: 0:03:43  lr: 0.000069  loss: 3.0604 (3.0697)  grad: 0.1714 (0.1777)  time: 0.6473  data: 0.0035  max mem: 57344
+train: [1]  [ 80/400]  eta: 0:03:29  lr: 0.000072  loss: 3.0679 (3.0729)  grad: 0.1854 (0.1808)  time: 0.6463  data: 0.0034  max mem: 57344
+train: [1]  [100/400]  eta: 0:03:15  lr: 0.000075  loss: 3.0614 (3.0694)  grad: 0.1884 (0.1824)  time: 0.6476  data: 0.0034  max mem: 57344
+train: [1]  [120/400]  eta: 0:03:02  lr: 0.000078  loss: 3.0417 (3.0641)  grad: 0.1856 (0.1822)  time: 0.6483  data: 0.0036  max mem: 57344
+train: [1]  [140/400]  eta: 0:02:49  lr: 0.000081  loss: 3.0273 (3.0574)  grad: 0.1831 (0.1834)  time: 0.6467  data: 0.0035  max mem: 57344
+train: [1]  [160/400]  eta: 0:02:36  lr: 0.000084  loss: 3.0268 (3.0544)  grad: 0.1831 (0.1835)  time: 0.6471  data: 0.0035  max mem: 57344
+train: [1]  [180/400]  eta: 0:02:23  lr: 0.000087  loss: 3.0326 (3.0546)  grad: 0.1825 (0.1841)  time: 0.6472  data: 0.0035  max mem: 57344
+train: [1]  [200/400]  eta: 0:02:10  lr: 0.000090  loss: 3.0501 (3.0527)  grad: 0.1954 (0.1857)  time: 0.6480  data: 0.0035  max mem: 57344
+train: [1]  [220/400]  eta: 0:01:56  lr: 0.000093  loss: 3.0199 (3.0507)  grad: 0.1954 (0.1868)  time: 0.6477  data: 0.0036  max mem: 57344
+train: [1]  [240/400]  eta: 0:01:43  lr: 0.000096  loss: 3.0003 (3.0450)  grad: 0.1931 (0.1872)  time: 0.6474  data: 0.0036  max mem: 57344
+train: [1]  [260/400]  eta: 0:01:30  lr: 0.000099  loss: 2.9812 (3.0433)  grad: 0.1992 (0.1886)  time: 0.6474  data: 0.0036  max mem: 57344
+train: [1]  [280/400]  eta: 0:01:17  lr: 0.000102  loss: 3.0183 (3.0415)  grad: 0.2099 (0.1905)  time: 0.6474  data: 0.0036  max mem: 57344
+train: [1]  [300/400]  eta: 0:01:04  lr: 0.000105  loss: 2.9866 (3.0380)  grad: 0.2146 (0.1921)  time: 0.6474  data: 0.0035  max mem: 57344
+train: [1]  [320/400]  eta: 0:00:51  lr: 0.000108  loss: 2.9863 (3.0362)  grad: 0.1989 (0.1925)  time: 0.6475  data: 0.0036  max mem: 57344
+train: [1]  [340/400]  eta: 0:00:38  lr: 0.000111  loss: 2.9863 (3.0334)  grad: 0.2159 (0.1944)  time: 0.6470  data: 0.0035  max mem: 57344
+train: [1]  [360/400]  eta: 0:00:25  lr: 0.000114  loss: 2.9760 (3.0309)  grad: 0.2161 (0.1955)  time: 0.6474  data: 0.0035  max mem: 57344
+train: [1]  [380/400]  eta: 0:00:12  lr: 0.000117  loss: 3.0102 (3.0297)  grad: 0.2102 (0.1966)  time: 0.6479  data: 0.0035  max mem: 57344
+train: [1]  [399/400]  eta: 0:00:00  lr: 0.000120  loss: 3.0051 (3.0281)  grad: 0.2101 (0.1975)  time: 0.6475  data: 0.0036  max mem: 57344
+train: [1] Total time: 0:04:19 (0.6491 s / it)
+train: [1] Summary: lr: 0.000120  loss: 3.0051 (3.0281)  grad: 0.2101 (0.1975)
+eval (validation): [1]  [ 0/85]  eta: 0:01:19    time: 0.9304  data: 0.5718  max mem: 57344
+eval (validation): [1]  [20/85]  eta: 0:00:25    time: 0.3697  data: 0.0031  max mem: 57344
+eval (validation): [1]  [40/85]  eta: 0:00:17    time: 0.3707  data: 0.0035  max mem: 57344
+eval (validation): [1]  [60/85]  eta: 0:00:09    time: 0.3705  data: 0.0034  max mem: 57344
+eval (validation): [1]  [80/85]  eta: 0:00:01    time: 0.3699  data: 0.0036  max mem: 57344
+eval (validation): [1]  [84/85]  eta: 0:00:00    time: 0.3633  data: 0.0036  max mem: 57344
+eval (validation): [1] Total time: 0:00:32 (0.3765 s / it)
+cv: [1]  best hparam: (14, 1.0) (040) ('040_lr1.4e+01_wd1.0e+00')  loss: 2.513  acc: 0.241  f1: 0.176
+saving checkpoint experiments/schaefer1000/output/schaefer1000/schaefer1000_lr3e-4_1/eval_v2/nsd_cococlip__patch__attn/checkpoint-last.pth
+saving best checkpoint experiments/schaefer1000/output/schaefer1000/schaefer1000_lr3e-4_1/eval_v2/nsd_cococlip__patch__attn/checkpoint-best.pth
+train: [2]  [  0/400]  eta: 0:08:11  lr: nan  time: 1.2277  data: 0.5961  max mem: 57344
+train: [2]  [ 20/400]  eta: 0:04:15  lr: 0.000123  loss: 2.8967 (2.9314)  grad: 0.2181 (0.2206)  time: 0.6455  data: 0.0034  max mem: 57344
+train: [2]  [ 40/400]  eta: 0:03:57  lr: 0.000126  loss: 2.9141 (2.9307)  grad: 0.2221 (0.2464)  time: 0.6471  data: 0.0035  max mem: 57344
+train: [2]  [ 60/400]  eta: 0:03:43  lr: 0.000129  loss: 3.0058 (3.0684)  grad: 0.3826 (0.5567)  time: 0.6479  data: 0.0036  max mem: 57344
+WARNING: classifier 48 (50, 1.0) diverged (loss=92.81 > 63.56) at step 434. Freezing.
+train: [2]  [ 80/400]  eta: 0:03:29  lr: 0.000132  loss: 3.4618 (3.3017)  grad: 1.6881 (0.9273)  time: 0.6441  data: 0.0035  max mem: 57344
+WARNING: classifier 47 (43, 1.0) diverged (loss=68.79 > 63.56) at step 446. Freezing.
+train: [2]  [100/400]  eta: 0:03:15  lr: 0.000135  loss: 3.8224 (3.3603)  grad: 1.7160 (1.0273)  time: 0.6394  data: 0.0035  max mem: 57344
+train: [2]  [120/400]  eta: 0:03:01  lr: 0.000138  loss: 2.9724 (3.2937)  grad: 0.2070 (0.8896)  time: 0.6358  data: 0.0036  max mem: 57344
+train: [2]  [140/400]  eta: 0:02:48  lr: 0.000141  loss: 2.9504 (3.2416)  grad: 0.2038 (0.7917)  time: 0.6358  data: 0.0036  max mem: 57344
+train: [2]  [160/400]  eta: 0:02:34  lr: 0.000144  loss: 2.9214 (3.2013)  grad: 0.2038 (0.7181)  time: 0.6364  data: 0.0037  max mem: 57344
+train: [2]  [180/400]  eta: 0:02:21  lr: 0.000147  loss: 2.9109 (3.1722)  grad: 0.2126 (0.6633)  time: 0.6361  data: 0.0037  max mem: 57344
+train: [2]  [200/400]  eta: 0:02:08  lr: 0.000150  loss: 2.9343 (3.1463)  grad: 0.2160 (0.6182)  time: 0.6363  data: 0.0037  max mem: 57344
+train: [2]  [220/400]  eta: 0:01:55  lr: 0.000153  loss: 2.9343 (3.1263)  grad: 0.2101 (0.5815)  time: 0.6357  data: 0.0037  max mem: 57344
+train: [2]  [240/400]  eta: 0:01:42  lr: 0.000156  loss: 2.9382 (3.1109)  grad: 0.2229 (0.5518)  time: 0.6364  data: 0.0036  max mem: 57344
+train: [2]  [260/400]  eta: 0:01:29  lr: 0.000159  loss: 2.9275 (3.0968)  grad: 0.2232 (0.5265)  time: 0.6358  data: 0.0036  max mem: 57344
+train: [2]  [280/400]  eta: 0:01:16  lr: 0.000162  loss: 2.9287 (3.0868)  grad: 0.2273 (0.5057)  time: 0.6356  data: 0.0036  max mem: 57344
+train: [2]  [300/400]  eta: 0:01:04  lr: 0.000165  loss: 2.9424 (3.0775)  grad: 0.2290 (0.4874)  time: 0.6356  data: 0.0036  max mem: 57344
+train: [2]  [320/400]  eta: 0:00:51  lr: 0.000168  loss: 2.9395 (3.0695)  grad: 0.2384 (0.4721)  time: 0.6355  data: 0.0036  max mem: 57344
+train: [2]  [340/400]  eta: 0:00:38  lr: 0.000171  loss: 2.9357 (3.0618)  grad: 0.2432 (0.4588)  time: 0.6359  data: 0.0036  max mem: 57344
+train: [2]  [360/400]  eta: 0:00:25  lr: 0.000174  loss: 2.9455 (3.0560)  grad: 0.2432 (0.4474)  time: 0.6356  data: 0.0037  max mem: 57344
+train: [2]  [380/400]  eta: 0:00:12  lr: 0.000177  loss: 2.9329 (3.0489)  grad: 0.2564 (0.4375)  time: 0.6356  data: 0.0036  max mem: 57344
+train: [2]  [399/400]  eta: 0:00:00  lr: 0.000180  loss: 2.9714 (3.0471)  grad: 0.2738 (0.4405)  time: 0.6358  data: 0.0036  max mem: 57344
+train: [2] Total time: 0:04:15 (0.6398 s / it)
+train: [2] Summary: lr: 0.000180  loss: 2.9714 (3.0471)  grad: 0.2738 (0.4405)
+eval (validation): [2]  [ 0/85]  eta: 0:01:17    time: 0.9066  data: 0.5493  max mem: 57344
+eval (validation): [2]  [20/85]  eta: 0:00:25    time: 0.3693  data: 0.0030  max mem: 57344
+eval (validation): [2]  [40/85]  eta: 0:00:17    time: 0.3690  data: 0.0035  max mem: 57344
+eval (validation): [2]  [60/85]  eta: 0:00:09    time: 0.3690  data: 0.0033  max mem: 57344
+eval (validation): [2]  [80/85]  eta: 0:00:01    time: 0.3692  data: 0.0035  max mem: 57344
+eval (validation): [2]  [84/85]  eta: 0:00:00    time: 0.3629  data: 0.0035  max mem: 57344
+eval (validation): [2] Total time: 0:00:31 (0.3751 s / it)
+cv: [2]  best hparam: (5.1, 1.0) (034) ('034_lr5.1e+00_wd1.0e+00')  loss: 2.470  acc: 0.255  f1: 0.200
+saving checkpoint experiments/schaefer1000/output/schaefer1000/schaefer1000_lr3e-4_1/eval_v2/nsd_cococlip__patch__attn/checkpoint-last.pth
+saving best checkpoint experiments/schaefer1000/output/schaefer1000/schaefer1000_lr3e-4_1/eval_v2/nsd_cococlip__patch__attn/checkpoint-best.pth
+train: [3]  [  0/400]  eta: 0:08:27  lr: nan  time: 1.2695  data: 0.6469  max mem: 57344
+train: [3]  [ 20/400]  eta: 0:04:12  lr: 0.000183  loss: 3.3354 (3.5635)  grad: 1.1193 (1.5355)  time: 0.6333  data: 0.0022  max mem: 57344
+WARNING: classifier 45 (31, 1.0) diverged (loss=64.80 > 63.56) at step 612. Freezing.
+WARNING: classifier 46 (36, 1.0) diverged (loss=69.16 > 63.56) at step 618. Freezing.
+train: [3]  [ 40/400]  eta: 0:03:52  lr: 0.000186  loss: 3.4559 (3.6363)  grad: 1.3858 (1.5194)  time: 0.6298  data: 0.0036  max mem: 57344
+train: [3]  [ 60/400]  eta: 0:03:37  lr: 0.000189  loss: 2.9607 (3.3885)  grad: 0.2209 (1.0812)  time: 0.6240  data: 0.0037  max mem: 57344
+train: [3]  [ 80/400]  eta: 0:03:23  lr: 0.000192  loss: 2.8738 (3.2548)  grad: 0.2107 (0.8643)  time: 0.6247  data: 0.0037  max mem: 57344
+train: [3]  [100/400]  eta: 0:03:10  lr: 0.000195  loss: 2.8796 (3.1856)  grad: 0.2168 (0.7363)  time: 0.6244  data: 0.0038  max mem: 57344
+train: [3]  [120/400]  eta: 0:02:56  lr: 0.000198  loss: 2.9122 (3.1391)  grad: 0.2166 (0.6506)  time: 0.6244  data: 0.0038  max mem: 57344
+train: [3]  [140/400]  eta: 0:02:44  lr: 0.000201  loss: 2.9092 (3.1052)  grad: 0.2159 (0.5875)  time: 0.6247  data: 0.0036  max mem: 57344
+train: [3]  [160/400]  eta: 0:02:31  lr: 0.000204  loss: 2.8477 (3.0699)  grad: 0.2099 (0.5410)  time: 0.6234  data: 0.0035  max mem: 57344
+train: [3]  [180/400]  eta: 0:02:18  lr: 0.000207  loss: 2.8748 (3.0553)  grad: 0.2285 (0.5080)  time: 0.6230  data: 0.0035  max mem: 57344
+train: [3]  [200/400]  eta: 0:02:05  lr: 0.000210  loss: 2.9172 (3.0429)  grad: 0.2478 (0.4829)  time: 0.6253  data: 0.0037  max mem: 57344
+train: [3]  [220/400]  eta: 0:01:53  lr: 0.000213  loss: 2.9149 (3.0306)  grad: 0.2732 (0.4690)  time: 0.6248  data: 0.0037  max mem: 57344
+train: [3]  [240/400]  eta: 0:01:40  lr: 0.000216  loss: 2.9978 (3.0435)  grad: 0.3983 (0.5042)  time: 0.6245  data: 0.0037  max mem: 57344
+WARNING: classifier 44 (26, 1.0) diverged (loss=66.56 > 63.56) at step 723. Freezing.
+train: [3]  [260/400]  eta: 0:01:27  lr: 0.000219  loss: 3.0656 (3.0545)  grad: 0.5584 (0.5315)  time: 0.6206  data: 0.0037  max mem: 57344
+train: [3]  [280/400]  eta: 0:01:15  lr: 0.000222  loss: 2.8921 (3.0404)  grad: 0.2125 (0.5091)  time: 0.6191  data: 0.0038  max mem: 57344
+train: [3]  [300/400]  eta: 0:01:02  lr: 0.000225  loss: 2.8627 (3.0291)  grad: 0.2139 (0.4899)  time: 0.6191  data: 0.0037  max mem: 57344
+train: [3]  [320/400]  eta: 0:00:50  lr: 0.000228  loss: 2.8644 (3.0201)  grad: 0.2148 (0.4728)  time: 0.6191  data: 0.0037  max mem: 57344
+train: [3]  [340/400]  eta: 0:00:37  lr: 0.000231  loss: 2.8541 (3.0104)  grad: 0.2131 (0.4574)  time: 0.6188  data: 0.0037  max mem: 57344
+train: [3]  [360/400]  eta: 0:00:25  lr: 0.000234  loss: 2.8624 (3.0019)  grad: 0.2203 (0.4445)  time: 0.6185  data: 0.0036  max mem: 57344
+train: [3]  [380/400]  eta: 0:00:12  lr: 0.000237  loss: 2.8577 (2.9938)  grad: 0.2180 (0.4322)  time: 0.6189  data: 0.0037  max mem: 57344
+train: [3]  [399/400]  eta: 0:00:00  lr: 0.000240  loss: 2.8577 (2.9878)  grad: 0.2140 (0.4220)  time: 0.6190  data: 0.0037  max mem: 57344
+train: [3] Total time: 0:04:09 (0.6249 s / it)
+train: [3] Summary: lr: 0.000240  loss: 2.8577 (2.9878)  grad: 0.2140 (0.4220)
+eval (validation): [3]  [ 0/85]  eta: 0:01:17    time: 0.9123  data: 0.5551  max mem: 57344
+eval (validation): [3]  [20/85]  eta: 0:00:25    time: 0.3689  data: 0.0032  max mem: 57344
+eval (validation): [3]  [40/85]  eta: 0:00:17    time: 0.3690  data: 0.0033  max mem: 57344
+eval (validation): [3]  [60/85]  eta: 0:00:09    time: 0.3692  data: 0.0035  max mem: 57344
+eval (validation): [3]  [80/85]  eta: 0:00:01    time: 0.3696  data: 0.0035  max mem: 57344
+eval (validation): [3]  [84/85]  eta: 0:00:00    time: 0.3633  data: 0.0035  max mem: 57344
+eval (validation): [3] Total time: 0:00:31 (0.3753 s / it)
+cv: [3]  best hparam: (3.7, 1.0) (032) ('032_lr3.7e+00_wd1.0e+00')  loss: 2.489  acc: 0.250  f1: 0.195
+saving checkpoint experiments/schaefer1000/output/schaefer1000/schaefer1000_lr3e-4_1/eval_v2/nsd_cococlip__patch__attn/checkpoint-last.pth
+train: [4]  [  0/400]  eta: 0:08:16  lr: nan  time: 1.2405  data: 0.6369  max mem: 57344
+train: [4]  [ 20/400]  eta: 0:04:05  lr: 0.000243  loss: 2.8694 (2.8643)  grad: 0.2194 (0.2269)  time: 0.6170  data: 0.0032  max mem: 57344
+train: [4]  [ 40/400]  eta: 0:03:47  lr: 0.000246  loss: 2.8368 (2.8430)  grad: 0.2128 (0.2169)  time: 0.6182  data: 0.0036  max mem: 57344
+train: [4]  [ 60/400]  eta: 0:03:33  lr: 0.000249  loss: 2.8425 (2.8438)  grad: 0.2168 (0.2206)  time: 0.6181  data: 0.0036  max mem: 57344
+train: [4]  [ 80/400]  eta: 0:03:20  lr: 0.000252  loss: 2.8425 (2.8432)  grad: 0.2278 (0.2233)  time: 0.6185  data: 0.0036  max mem: 57344
+train: [4]  [100/400]  eta: 0:03:07  lr: 0.000255  loss: 2.8308 (2.8410)  grad: 0.2352 (0.2264)  time: 0.6183  data: 0.0036  max mem: 57344
+train: [4]  [120/400]  eta: 0:02:54  lr: 0.000258  loss: 2.8264 (2.8396)  grad: 0.2301 (0.2254)  time: 0.6183  data: 0.0036  max mem: 57344
+train: [4]  [140/400]  eta: 0:02:41  lr: 0.000261  loss: 2.8245 (2.8380)  grad: 0.2285 (0.2263)  time: 0.6187  data: 0.0036  max mem: 57344
+train: [4]  [160/400]  eta: 0:02:29  lr: 0.000264  loss: 2.8202 (2.8346)  grad: 0.2285 (0.2271)  time: 0.6181  data: 0.0036  max mem: 57344
+train: [4]  [180/400]  eta: 0:02:16  lr: 0.000267  loss: 2.8182 (2.8330)  grad: 0.2266 (0.2282)  time: 0.6186  data: 0.0036  max mem: 57344
+train: [4]  [200/400]  eta: 0:02:04  lr: 0.000270  loss: 2.8305 (2.8343)  grad: 0.2326 (0.2292)  time: 0.6183  data: 0.0037  max mem: 57344
+train: [4]  [220/400]  eta: 0:01:51  lr: 0.000273  loss: 2.8469 (2.8341)  grad: 0.2412 (0.2310)  time: 0.6182  data: 0.0036  max mem: 57344
+train: [4]  [240/400]  eta: 0:01:39  lr: 0.000276  loss: 2.8469 (2.8338)  grad: 0.2505 (0.2332)  time: 0.6179  data: 0.0036  max mem: 57344
+train: [4]  [260/400]  eta: 0:01:26  lr: 0.000279  loss: 2.8562 (2.8379)  grad: 0.2660 (0.2411)  time: 0.6179  data: 0.0036  max mem: 57344
+train: [4]  [280/400]  eta: 0:01:14  lr: 0.000282  loss: 2.9680 (2.8666)  grad: 0.5294 (0.3137)  time: 0.6180  data: 0.0036  max mem: 57344
+WARNING: classifier 43 (22, 1.0) diverged (loss=89.71 > 63.56) at step 942. Freezing.
+train: [4]  [300/400]  eta: 0:01:01  lr: 0.000285  loss: 3.0197 (2.8840)  grad: 0.5729 (0.3493)  time: 0.6130  data: 0.0035  max mem: 57344
+train: [4]  [320/400]  eta: 0:00:49  lr: 0.000288  loss: 2.8402 (2.8799)  grad: 0.2186 (0.3414)  time: 0.6125  data: 0.0036  max mem: 57344
+train: [4]  [340/400]  eta: 0:00:37  lr: 0.000291  loss: 2.8204 (2.8783)  grad: 0.2302 (0.3348)  time: 0.6124  data: 0.0036  max mem: 57344
+train: [4]  [360/400]  eta: 0:00:24  lr: 0.000294  loss: 2.8204 (2.8760)  grad: 0.2203 (0.3282)  time: 0.6126  data: 0.0037  max mem: 57344
+train: [4]  [380/400]  eta: 0:00:12  lr: 0.000297  loss: 2.8135 (2.8724)  grad: 0.2222 (0.3235)  time: 0.6124  data: 0.0036  max mem: 57344
+train: [4]  [399/400]  eta: 0:00:00  lr: 0.000300  loss: 2.7852 (2.8692)  grad: 0.2355 (0.3189)  time: 0.6123  data: 0.0036  max mem: 57344
+train: [4] Total time: 0:04:07 (0.6183 s / it)
+train: [4] Summary: lr: 0.000300  loss: 2.7852 (2.8692)  grad: 0.2355 (0.3189)
+eval (validation): [4]  [ 0/85]  eta: 0:01:23    time: 0.9802  data: 0.6230  max mem: 57344
+eval (validation): [4]  [20/85]  eta: 0:00:25    time: 0.3674  data: 0.0020  max mem: 57344
+eval (validation): [4]  [40/85]  eta: 0:00:17    time: 0.3689  data: 0.0033  max mem: 57344
+eval (validation): [4]  [60/85]  eta: 0:00:09    time: 0.3688  data: 0.0035  max mem: 57344
+eval (validation): [4]  [80/85]  eta: 0:00:01    time: 0.3682  data: 0.0033  max mem: 57344
+eval (validation): [4]  [84/85]  eta: 0:00:00    time: 0.3618  data: 0.0033  max mem: 57344
+eval (validation): [4] Total time: 0:00:31 (0.3752 s / it)
+cv: [4]  best hparam: (3.1, 1.0) (031) ('031_lr3.1e+00_wd1.0e+00')  loss: 2.448  acc: 0.270  f1: 0.208
+saving checkpoint experiments/schaefer1000/output/schaefer1000/schaefer1000_lr3e-4_1/eval_v2/nsd_cococlip__patch__attn/checkpoint-last.pth
+saving best checkpoint experiments/schaefer1000/output/schaefer1000/schaefer1000_lr3e-4_1/eval_v2/nsd_cococlip__patch__attn/checkpoint-best.pth
+train: [5]  [  0/400]  eta: 0:08:08  lr: nan  time: 1.2210  data: 0.6224  max mem: 57344
+train: [5]  [ 20/400]  eta: 0:04:03  lr: 0.000300  loss: 2.7865 (2.8232)  grad: 0.2321 (0.2367)  time: 0.6107  data: 0.0030  max mem: 57344
+train: [5]  [ 40/400]  eta: 0:03:45  lr: 0.000300  loss: 2.7843 (2.8087)  grad: 0.2221 (0.2307)  time: 0.6120  data: 0.0037  max mem: 57344
+train: [5]  [ 60/400]  eta: 0:03:31  lr: 0.000300  loss: 2.8128 (2.8133)  grad: 0.2291 (0.2366)  time: 0.6123  data: 0.0037  max mem: 57344
+train: [5]  [ 80/400]  eta: 0:03:18  lr: 0.000300  loss: 2.8258 (2.8092)  grad: 0.2333 (0.2357)  time: 0.6131  data: 0.0037  max mem: 57344
+train: [5]  [100/400]  eta: 0:03:05  lr: 0.000300  loss: 2.8038 (2.8110)  grad: 0.2401 (0.2380)  time: 0.6129  data: 0.0037  max mem: 57344
+train: [5]  [120/400]  eta: 0:02:52  lr: 0.000300  loss: 2.8038 (2.8110)  grad: 0.2578 (0.2420)  time: 0.6123  data: 0.0037  max mem: 57344
+train: [5]  [140/400]  eta: 0:02:40  lr: 0.000300  loss: 2.8036 (2.8103)  grad: 0.2503 (0.2431)  time: 0.6126  data: 0.0037  max mem: 57344
+train: [5]  [160/400]  eta: 0:02:27  lr: 0.000299  loss: 2.7803 (2.8048)  grad: 0.2440 (0.2428)  time: 0.6126  data: 0.0037  max mem: 57344
+train: [5]  [180/400]  eta: 0:02:15  lr: 0.000299  loss: 2.7772 (2.8046)  grad: 0.2405 (0.2428)  time: 0.6132  data: 0.0037  max mem: 57344
+train: [5]  [200/400]  eta: 0:02:03  lr: 0.000299  loss: 2.7802 (2.8001)  grad: 0.2404 (0.2423)  time: 0.6136  data: 0.0037  max mem: 57344
+train: [5]  [220/400]  eta: 0:01:50  lr: 0.000299  loss: 2.7443 (2.7953)  grad: 0.2382 (0.2418)  time: 0.6133  data: 0.0037  max mem: 57344
+train: [5]  [240/400]  eta: 0:01:38  lr: 0.000299  loss: 2.7443 (2.7902)  grad: 0.2313 (0.2405)  time: 0.6126  data: 0.0037  max mem: 57344
+train: [5]  [260/400]  eta: 0:01:26  lr: 0.000299  loss: 2.7697 (2.7912)  grad: 0.2248 (0.2400)  time: 0.6128  data: 0.0036  max mem: 57344
+train: [5]  [280/400]  eta: 0:01:13  lr: 0.000298  loss: 2.7654 (2.7897)  grad: 0.2331 (0.2397)  time: 0.6134  data: 0.0037  max mem: 57344
+train: [5]  [300/400]  eta: 0:01:01  lr: 0.000298  loss: 2.7592 (2.7892)  grad: 0.2341 (0.2390)  time: 0.6128  data: 0.0037  max mem: 57344
+train: [5]  [320/400]  eta: 0:00:49  lr: 0.000298  loss: 2.8031 (2.7904)  grad: 0.2272 (0.2382)  time: 0.6117  data: 0.0035  max mem: 57344
+train: [5]  [340/400]  eta: 0:00:36  lr: 0.000298  loss: 2.8002 (2.7884)  grad: 0.2304 (0.2383)  time: 0.6122  data: 0.0035  max mem: 57344
+train: [5]  [360/400]  eta: 0:00:24  lr: 0.000297  loss: 2.7689 (2.7878)  grad: 0.2355 (0.2379)  time: 0.6133  data: 0.0037  max mem: 57344
+train: [5]  [380/400]  eta: 0:00:12  lr: 0.000297  loss: 2.7689 (2.7873)  grad: 0.2176 (0.2369)  time: 0.6139  data: 0.0038  max mem: 57344
+train: [5]  [399/400]  eta: 0:00:00  lr: 0.000297  loss: 2.7686 (2.7859)  grad: 0.2047 (0.2351)  time: 0.6133  data: 0.0037  max mem: 57344
+train: [5] Total time: 0:04:05 (0.6145 s / it)
+train: [5] Summary: lr: 0.000297  loss: 2.7686 (2.7859)  grad: 0.2047 (0.2351)
+eval (validation): [5]  [ 0/85]  eta: 0:01:15    time: 0.8857  data: 0.5277  max mem: 57344
+eval (validation): [5]  [20/85]  eta: 0:00:25    time: 0.3695  data: 0.0034  max mem: 57344
+eval (validation): [5]  [40/85]  eta: 0:00:17    time: 0.3694  data: 0.0034  max mem: 57344
+eval (validation): [5]  [60/85]  eta: 0:00:09    time: 0.3689  data: 0.0036  max mem: 57344
+eval (validation): [5]  [80/85]  eta: 0:00:01    time: 0.3685  data: 0.0034  max mem: 57344
+eval (validation): [5]  [84/85]  eta: 0:00:00    time: 0.3623  data: 0.0034  max mem: 57344
+eval (validation): [5] Total time: 0:00:31 (0.3748 s / it)
+cv: [5]  best hparam: (2.7, 1.0) (030) ('030_lr2.7e+00_wd1.0e+00')  loss: 2.385  acc: 0.281  f1: 0.212
+saving checkpoint experiments/schaefer1000/output/schaefer1000/schaefer1000_lr3e-4_1/eval_v2/nsd_cococlip__patch__attn/checkpoint-last.pth
+saving best checkpoint experiments/schaefer1000/output/schaefer1000/schaefer1000_lr3e-4_1/eval_v2/nsd_cococlip__patch__attn/checkpoint-best.pth
+train: [6]  [  0/400]  eta: 0:07:57  lr: nan  time: 1.1947  data: 0.5964  max mem: 57344
+train: [6]  [ 20/400]  eta: 0:04:02  lr: 0.000296  loss: 2.7177 (2.7401)  grad: 0.2140 (0.2121)  time: 0.6112  data: 0.0029  max mem: 57344
+train: [6]  [ 40/400]  eta: 0:03:45  lr: 0.000296  loss: 2.7177 (2.7138)  grad: 0.2183 (0.2205)  time: 0.6126  data: 0.0036  max mem: 57344
+train: [6]  [ 60/400]  eta: 0:03:31  lr: 0.000296  loss: 2.7214 (2.7234)  grad: 0.2279 (0.2241)  time: 0.6132  data: 0.0036  max mem: 57344
+train: [6]  [ 80/400]  eta: 0:03:18  lr: 0.000295  loss: 2.7284 (2.7212)  grad: 0.2253 (0.2247)  time: 0.6128  data: 0.0037  max mem: 57344
+train: [6]  [100/400]  eta: 0:03:05  lr: 0.000295  loss: 2.6922 (2.7198)  grad: 0.2253 (0.2250)  time: 0.6134  data: 0.0037  max mem: 57344
+train: [6]  [120/400]  eta: 0:02:52  lr: 0.000295  loss: 2.7251 (2.7254)  grad: 0.2248 (0.2254)  time: 0.6132  data: 0.0037  max mem: 57344
+train: [6]  [140/400]  eta: 0:02:40  lr: 0.000294  loss: 2.7343 (2.7217)  grad: 0.2202 (0.2240)  time: 0.6133  data: 0.0036  max mem: 57344
+train: [6]  [160/400]  eta: 0:02:27  lr: 0.000294  loss: 2.6890 (2.7161)  grad: 0.2223 (0.2254)  time: 0.6129  data: 0.0036  max mem: 57344
+train: [6]  [180/400]  eta: 0:02:15  lr: 0.000293  loss: 2.6785 (2.7177)  grad: 0.2255 (0.2257)  time: 0.6133  data: 0.0037  max mem: 57344
+train: [6]  [200/400]  eta: 0:02:03  lr: 0.000293  loss: 2.7004 (2.7175)  grad: 0.2319 (0.2274)  time: 0.6128  data: 0.0037  max mem: 57344
+train: [6]  [220/400]  eta: 0:01:50  lr: 0.000292  loss: 2.7037 (2.7183)  grad: 0.2302 (0.2269)  time: 0.6135  data: 0.0037  max mem: 57344
+train: [6]  [240/400]  eta: 0:01:38  lr: 0.000292  loss: 2.7155 (2.7191)  grad: 0.2207 (0.2268)  time: 0.6129  data: 0.0036  max mem: 57344
+train: [6]  [260/400]  eta: 0:01:26  lr: 0.000291  loss: 2.7323 (2.7206)  grad: 0.2191 (0.2266)  time: 0.6126  data: 0.0036  max mem: 57344
+train: [6]  [280/400]  eta: 0:01:13  lr: 0.000291  loss: 2.7323 (2.7210)  grad: 0.2163 (0.2257)  time: 0.6131  data: 0.0036  max mem: 57344
+train: [6]  [300/400]  eta: 0:01:01  lr: 0.000290  loss: 2.7212 (2.7213)  grad: 0.2187 (0.2256)  time: 0.6127  data: 0.0037  max mem: 57344
+train: [6]  [320/400]  eta: 0:00:49  lr: 0.000290  loss: 2.7315 (2.7231)  grad: 0.2254 (0.2258)  time: 0.6126  data: 0.0036  max mem: 57344
+train: [6]  [340/400]  eta: 0:00:36  lr: 0.000289  loss: 2.7315 (2.7221)  grad: 0.2301 (0.2264)  time: 0.6126  data: 0.0036  max mem: 57344
+train: [6]  [360/400]  eta: 0:00:24  lr: 0.000288  loss: 2.6996 (2.7212)  grad: 0.2249 (0.2265)  time: 0.6128  data: 0.0036  max mem: 57344
+train: [6]  [380/400]  eta: 0:00:12  lr: 0.000288  loss: 2.6965 (2.7211)  grad: 0.2233 (0.2267)  time: 0.6131  data: 0.0037  max mem: 57344
+train: [6]  [399/400]  eta: 0:00:00  lr: 0.000287  loss: 2.7053 (2.7199)  grad: 0.2274 (0.2267)  time: 0.6131  data: 0.0037  max mem: 57344
+train: [6] Total time: 0:04:05 (0.6146 s / it)
+train: [6] Summary: lr: 0.000287  loss: 2.7053 (2.7199)  grad: 0.2274 (0.2267)
+eval (validation): [6]  [ 0/85]  eta: 0:01:26    time: 1.0150  data: 0.6588  max mem: 57344
+eval (validation): [6]  [20/85]  eta: 0:00:25    time: 0.3687  data: 0.0022  max mem: 57344
+eval (validation): [6]  [40/85]  eta: 0:00:17    time: 0.3697  data: 0.0035  max mem: 57344
+eval (validation): [6]  [60/85]  eta: 0:00:09    time: 0.3693  data: 0.0035  max mem: 57344
+eval (validation): [6]  [80/85]  eta: 0:00:01    time: 0.3687  data: 0.0035  max mem: 57344
+eval (validation): [6]  [84/85]  eta: 0:00:00    time: 0.3624  data: 0.0034  max mem: 57344
+eval (validation): [6] Total time: 0:00:31 (0.3763 s / it)
+cv: [6]  best hparam: (2.3, 1.0) (029) ('029_lr2.3e+00_wd1.0e+00')  loss: 2.405  acc: 0.279  f1: 0.217
+saving checkpoint experiments/schaefer1000/output/schaefer1000/schaefer1000_lr3e-4_1/eval_v2/nsd_cococlip__patch__attn/checkpoint-last.pth
+train: [7]  [  0/400]  eta: 0:07:49  lr: nan  time: 1.1744  data: 0.5757  max mem: 57344
+train: [7]  [ 20/400]  eta: 0:04:02  lr: 0.000286  loss: 2.6409 (2.6306)  grad: 0.2155 (0.2178)  time: 0.6117  data: 0.0034  max mem: 57344
+train: [7]  [ 40/400]  eta: 0:03:45  lr: 0.000286  loss: 2.6637 (2.6607)  grad: 0.2250 (0.2250)  time: 0.6121  data: 0.0036  max mem: 57344
+train: [7]  [ 60/400]  eta: 0:03:31  lr: 0.000285  loss: 2.6809 (2.6671)  grad: 0.2338 (0.2281)  time: 0.6124  data: 0.0037  max mem: 57344
+train: [7]  [ 80/400]  eta: 0:03:18  lr: 0.000284  loss: 2.6558 (2.6660)  grad: 0.2346 (0.2307)  time: 0.6125  data: 0.0036  max mem: 57344
+train: [7]  [100/400]  eta: 0:03:05  lr: 0.000284  loss: 2.6605 (2.6711)  grad: 0.2278 (0.2312)  time: 0.6126  data: 0.0036  max mem: 57344
+train: [7]  [120/400]  eta: 0:02:52  lr: 0.000283  loss: 2.6832 (2.6671)  grad: 0.2358 (0.2334)  time: 0.6128  data: 0.0037  max mem: 57344
+train: [7]  [140/400]  eta: 0:02:40  lr: 0.000282  loss: 2.6553 (2.6698)  grad: 0.2418 (0.2352)  time: 0.6124  data: 0.0036  max mem: 57344
+train: [7]  [160/400]  eta: 0:02:27  lr: 0.000282  loss: 2.6833 (2.6690)  grad: 0.2260 (0.2334)  time: 0.6124  data: 0.0036  max mem: 57344
+train: [7]  [180/400]  eta: 0:02:15  lr: 0.000281  loss: 2.6465 (2.6662)  grad: 0.2144 (0.2324)  time: 0.6125  data: 0.0037  max mem: 57344
+train: [7]  [200/400]  eta: 0:02:03  lr: 0.000280  loss: 2.6454 (2.6645)  grad: 0.2200 (0.2319)  time: 0.6124  data: 0.0038  max mem: 57344
+train: [7]  [220/400]  eta: 0:01:50  lr: 0.000279  loss: 2.6713 (2.6651)  grad: 0.2260 (0.2324)  time: 0.6120  data: 0.0037  max mem: 57344
+train: [7]  [240/400]  eta: 0:01:38  lr: 0.000278  loss: 2.6553 (2.6630)  grad: 0.2273 (0.2318)  time: 0.6122  data: 0.0036  max mem: 57344
+train: [7]  [260/400]  eta: 0:01:26  lr: 0.000278  loss: 2.6755 (2.6665)  grad: 0.2295 (0.2326)  time: 0.6127  data: 0.0036  max mem: 57344
+train: [7]  [280/400]  eta: 0:01:13  lr: 0.000277  loss: 2.7129 (2.6726)  grad: 0.2414 (0.2333)  time: 0.6124  data: 0.0036  max mem: 57344
+train: [7]  [300/400]  eta: 0:01:01  lr: 0.000276  loss: 2.6928 (2.6729)  grad: 0.2507 (0.2347)  time: 0.6121  data: 0.0036  max mem: 57344
+train: [7]  [320/400]  eta: 0:00:49  lr: 0.000275  loss: 2.6745 (2.6740)  grad: 0.2527 (0.2358)  time: 0.6121  data: 0.0036  max mem: 57344
+train: [7]  [340/400]  eta: 0:00:36  lr: 0.000274  loss: 2.6978 (2.6769)  grad: 0.2483 (0.2364)  time: 0.6126  data: 0.0035  max mem: 57344
+train: [7]  [360/400]  eta: 0:00:24  lr: 0.000273  loss: 2.6840 (2.6767)  grad: 0.2395 (0.2363)  time: 0.6125  data: 0.0036  max mem: 57344
+train: [7]  [380/400]  eta: 0:00:12  lr: 0.000272  loss: 2.7175 (2.6799)  grad: 0.2358 (0.2361)  time: 0.6123  data: 0.0038  max mem: 57344
+train: [7]  [399/400]  eta: 0:00:00  lr: 0.000271  loss: 2.7324 (2.6830)  grad: 0.2291 (0.2359)  time: 0.6122  data: 0.0036  max mem: 57344
+train: [7] Total time: 0:04:05 (0.6140 s / it)
+train: [7] Summary: lr: 0.000271  loss: 2.7324 (2.6830)  grad: 0.2291 (0.2359)
+eval (validation): [7]  [ 0/85]  eta: 0:01:17    time: 0.9099  data: 0.5518  max mem: 57344
+eval (validation): [7]  [20/85]  eta: 0:00:25    time: 0.3688  data: 0.0031  max mem: 57344
+eval (validation): [7]  [40/85]  eta: 0:00:17    time: 0.3690  data: 0.0035  max mem: 57344
+eval (validation): [7]  [60/85]  eta: 0:00:09    time: 0.3685  data: 0.0034  max mem: 57344
+eval (validation): [7]  [80/85]  eta: 0:00:01    time: 0.3684  data: 0.0035  max mem: 57344
+eval (validation): [7]  [84/85]  eta: 0:00:00    time: 0.3627  data: 0.0035  max mem: 57344
+eval (validation): [7] Total time: 0:00:31 (0.3749 s / it)
+cv: [7]  best hparam: (2.7, 1.0) (030) ('030_lr2.7e+00_wd1.0e+00')  loss: 2.405  acc: 0.279  f1: 0.212
+saving checkpoint experiments/schaefer1000/output/schaefer1000/schaefer1000_lr3e-4_1/eval_v2/nsd_cococlip__patch__attn/checkpoint-last.pth
+train: [8]  [  0/400]  eta: 0:07:58  lr: nan  time: 1.1973  data: 0.5981  max mem: 57344
+train: [8]  [ 20/400]  eta: 0:04:03  lr: 0.000270  loss: 2.6380 (2.6436)  grad: 0.2226 (0.2228)  time: 0.6128  data: 0.0032  max mem: 57344
+train: [8]  [ 40/400]  eta: 0:03:45  lr: 0.000270  loss: 2.6380 (2.6301)  grad: 0.2240 (0.2245)  time: 0.6129  data: 0.0037  max mem: 57344
+train: [8]  [ 60/400]  eta: 0:03:31  lr: 0.000269  loss: 2.6398 (2.6448)  grad: 0.2311 (0.2300)  time: 0.6125  data: 0.0036  max mem: 57344
+train: [8]  [ 80/400]  eta: 0:03:18  lr: 0.000268  loss: 2.6338 (2.6357)  grad: 0.2451 (0.2340)  time: 0.6125  data: 0.0036  max mem: 57344
+train: [8]  [100/400]  eta: 0:03:05  lr: 0.000267  loss: 2.6161 (2.6372)  grad: 0.2443 (0.2352)  time: 0.6132  data: 0.0035  max mem: 57344
+train: [8]  [120/400]  eta: 0:02:52  lr: 0.000266  loss: 2.6397 (2.6408)  grad: 0.2367 (0.2354)  time: 0.6136  data: 0.0036  max mem: 57344
+train: [8]  [140/400]  eta: 0:02:40  lr: 0.000265  loss: 2.6298 (2.6404)  grad: 0.2392 (0.2362)  time: 0.6138  data: 0.0037  max mem: 57344
+train: [8]  [160/400]  eta: 0:02:27  lr: 0.000264  loss: 2.6298 (2.6437)  grad: 0.2422 (0.2378)  time: 0.6123  data: 0.0038  max mem: 57344
+train: [8]  [180/400]  eta: 0:02:15  lr: 0.000263  loss: 2.6419 (2.6439)  grad: 0.2423 (0.2382)  time: 0.6130  data: 0.0037  max mem: 57344
+train: [8]  [200/400]  eta: 0:02:03  lr: 0.000262  loss: 2.6329 (2.6445)  grad: 0.2430 (0.2388)  time: 0.6125  data: 0.0037  max mem: 57344
+train: [8]  [220/400]  eta: 0:01:50  lr: 0.000260  loss: 2.6273 (2.6406)  grad: 0.2417 (0.2384)  time: 0.6121  data: 0.0037  max mem: 57344
+train: [8]  [240/400]  eta: 0:01:38  lr: 0.000259  loss: 2.6258 (2.6427)  grad: 0.2244 (0.2378)  time: 0.6126  data: 0.0037  max mem: 57344
+train: [8]  [260/400]  eta: 0:01:26  lr: 0.000258  loss: 2.6703 (2.6442)  grad: 0.2369 (0.2384)  time: 0.6126  data: 0.0037  max mem: 57344
+train: [8]  [280/400]  eta: 0:01:13  lr: 0.000257  loss: 2.6730 (2.6447)  grad: 0.2369 (0.2380)  time: 0.6127  data: 0.0037  max mem: 57344
+train: [8]  [300/400]  eta: 0:01:01  lr: 0.000256  loss: 2.6123 (2.6431)  grad: 0.2363 (0.2384)  time: 0.6131  data: 0.0037  max mem: 57344
+train: [8]  [320/400]  eta: 0:00:49  lr: 0.000255  loss: 2.6237 (2.6426)  grad: 0.2402 (0.2382)  time: 0.6124  data: 0.0036  max mem: 57344
+train: [8]  [340/400]  eta: 0:00:36  lr: 0.000254  loss: 2.6237 (2.6436)  grad: 0.2344 (0.2381)  time: 0.6133  data: 0.0037  max mem: 57344
+train: [8]  [360/400]  eta: 0:00:24  lr: 0.000253  loss: 2.6185 (2.6430)  grad: 0.2290 (0.2375)  time: 0.6128  data: 0.0036  max mem: 57344
+train: [8]  [380/400]  eta: 0:00:12  lr: 0.000252  loss: 2.6325 (2.6422)  grad: 0.2257 (0.2371)  time: 0.6120  data: 0.0036  max mem: 57344
+train: [8]  [399/400]  eta: 0:00:00  lr: 0.000250  loss: 2.6344 (2.6404)  grad: 0.2294 (0.2372)  time: 0.6125  data: 0.0036  max mem: 57344
+train: [8] Total time: 0:04:05 (0.6145 s / it)
+train: [8] Summary: lr: 0.000250  loss: 2.6344 (2.6404)  grad: 0.2294 (0.2372)
+eval (validation): [8]  [ 0/85]  eta: 0:01:18    time: 0.9287  data: 0.5693  max mem: 57344
+eval (validation): [8]  [20/85]  eta: 0:00:25    time: 0.3680  data: 0.0028  max mem: 57344
+eval (validation): [8]  [40/85]  eta: 0:00:17    time: 0.3682  data: 0.0033  max mem: 57344
+eval (validation): [8]  [60/85]  eta: 0:00:09    time: 0.3683  data: 0.0032  max mem: 57344
+eval (validation): [8]  [80/85]  eta: 0:00:01    time: 0.3684  data: 0.0033  max mem: 57344
+eval (validation): [8]  [84/85]  eta: 0:00:00    time: 0.3619  data: 0.0033  max mem: 57344
+eval (validation): [8] Total time: 0:00:31 (0.3745 s / it)
+cv: [8]  best hparam: (0.85, 1.0) (023) ('023_lr8.5e-01_wd1.0e+00')  loss: 2.403  acc: 0.277  f1: 0.212
+saving checkpoint experiments/schaefer1000/output/schaefer1000/schaefer1000_lr3e-4_1/eval_v2/nsd_cococlip__patch__attn/checkpoint-last.pth
+train: [9]  [  0/400]  eta: 0:07:41  lr: nan  time: 1.1538  data: 0.5528  max mem: 57344
+train: [9]  [ 20/400]  eta: 0:04:02  lr: 0.000249  loss: 2.6095 (2.6020)  grad: 0.2275 (0.2273)  time: 0.6118  data: 0.0033  max mem: 57344
+train: [9]  [ 40/400]  eta: 0:03:45  lr: 0.000248  loss: 2.5759 (2.5796)  grad: 0.2284 (0.2297)  time: 0.6125  data: 0.0036  max mem: 57344
+train: [9]  [ 60/400]  eta: 0:03:31  lr: 0.000247  loss: 2.6015 (2.5958)  grad: 0.2291 (0.2297)  time: 0.6124  data: 0.0036  max mem: 57344
+train: [9]  [ 80/400]  eta: 0:03:18  lr: 0.000246  loss: 2.6412 (2.6055)  grad: 0.2254 (0.2299)  time: 0.6127  data: 0.0037  max mem: 57344
+train: [9]  [100/400]  eta: 0:03:05  lr: 0.000244  loss: 2.5935 (2.6028)  grad: 0.2310 (0.2308)  time: 0.6128  data: 0.0036  max mem: 57344
+train: [9]  [120/400]  eta: 0:02:52  lr: 0.000243  loss: 2.5867 (2.5987)  grad: 0.2310 (0.2301)  time: 0.6127  data: 0.0037  max mem: 57344
+train: [9]  [140/400]  eta: 0:02:40  lr: 0.000242  loss: 2.5337 (2.5881)  grad: 0.2342 (0.2311)  time: 0.6122  data: 0.0037  max mem: 57344
+train: [9]  [160/400]  eta: 0:02:27  lr: 0.000241  loss: 2.5838 (2.5969)  grad: 0.2342 (0.2322)  time: 0.6126  data: 0.0037  max mem: 57344
+train: [9]  [180/400]  eta: 0:02:15  lr: 0.000240  loss: 2.6640 (2.6029)  grad: 0.2351 (0.2328)  time: 0.6120  data: 0.0036  max mem: 57344
+train: [9]  [200/400]  eta: 0:02:03  lr: 0.000238  loss: 2.5967 (2.6018)  grad: 0.2351 (0.2328)  time: 0.6124  data: 0.0038  max mem: 57344
+train: [9]  [220/400]  eta: 0:01:50  lr: 0.000237  loss: 2.5967 (2.6070)  grad: 0.2322 (0.2330)  time: 0.6124  data: 0.0037  max mem: 57344
+train: [9]  [240/400]  eta: 0:01:38  lr: 0.000236  loss: 2.6234 (2.6102)  grad: 0.2321 (0.2331)  time: 0.6127  data: 0.0038  max mem: 57344
+train: [9]  [260/400]  eta: 0:01:26  lr: 0.000234  loss: 2.6234 (2.6113)  grad: 0.2265 (0.2326)  time: 0.6126  data: 0.0037  max mem: 57344
+train: [9]  [280/400]  eta: 0:01:13  lr: 0.000233  loss: 2.5742 (2.6102)  grad: 0.2214 (0.2311)  time: 0.6125  data: 0.0037  max mem: 57344
+train: [9]  [300/400]  eta: 0:01:01  lr: 0.000232  loss: 2.5742 (2.6103)  grad: 0.2259 (0.2318)  time: 0.6122  data: 0.0037  max mem: 57344
+train: [9]  [320/400]  eta: 0:00:49  lr: 0.000230  loss: 2.5970 (2.6096)  grad: 0.2381 (0.2318)  time: 0.6126  data: 0.0037  max mem: 57344
+train: [9]  [340/400]  eta: 0:00:36  lr: 0.000229  loss: 2.6045 (2.6111)  grad: 0.2353 (0.2321)  time: 0.6127  data: 0.0037  max mem: 57344
+train: [9]  [360/400]  eta: 0:00:24  lr: 0.000228  loss: 2.5913 (2.6082)  grad: 0.2332 (0.2320)  time: 0.6126  data: 0.0038  max mem: 57344
+train: [9]  [380/400]  eta: 0:00:12  lr: 0.000226  loss: 2.5842 (2.6059)  grad: 0.2247 (0.2316)  time: 0.6127  data: 0.0037  max mem: 57344
+train: [9]  [399/400]  eta: 0:00:00  lr: 0.000225  loss: 2.5842 (2.6060)  grad: 0.2247 (0.2315)  time: 0.6128  data: 0.0037  max mem: 57344
+train: [9] Total time: 0:04:05 (0.6141 s / it)
+train: [9] Summary: lr: 0.000225  loss: 2.5842 (2.6060)  grad: 0.2247 (0.2315)
+eval (validation): [9]  [ 0/85]  eta: 0:01:19    time: 0.9374  data: 0.5777  max mem: 57344
+eval (validation): [9]  [20/85]  eta: 0:00:25    time: 0.3687  data: 0.0029  max mem: 57344
+eval (validation): [9]  [40/85]  eta: 0:00:17    time: 0.3690  data: 0.0034  max mem: 57344
+eval (validation): [9]  [60/85]  eta: 0:00:09    time: 0.3691  data: 0.0035  max mem: 57344
+eval (validation): [9]  [80/85]  eta: 0:00:01    time: 0.3688  data: 0.0035  max mem: 57344
+eval (validation): [9]  [84/85]  eta: 0:00:00    time: 0.3625  data: 0.0035  max mem: 57344
+eval (validation): [9] Total time: 0:00:31 (0.3753 s / it)
+cv: [9]  best hparam: (5.1, 1.0) (034) ('034_lr5.1e+00_wd1.0e+00')  loss: 2.392  acc: 0.284  f1: 0.228
+saving checkpoint experiments/schaefer1000/output/schaefer1000/schaefer1000_lr3e-4_1/eval_v2/nsd_cococlip__patch__attn/checkpoint-last.pth
+saving best checkpoint experiments/schaefer1000/output/schaefer1000/schaefer1000_lr3e-4_1/eval_v2/nsd_cococlip__patch__attn/checkpoint-best.pth
+train: [10]  [  0/400]  eta: 0:07:59  lr: nan  time: 1.1976  data: 0.5971  max mem: 57344
+train: [10]  [ 20/400]  eta: 0:04:02  lr: 0.000224  loss: 2.5657 (2.5602)  grad: 0.2146 (0.2198)  time: 0.6112  data: 0.0031  max mem: 57344
+train: [10]  [ 40/400]  eta: 0:03:45  lr: 0.000222  loss: 2.4977 (2.5387)  grad: 0.2282 (0.2282)  time: 0.6121  data: 0.0037  max mem: 57344
+train: [10]  [ 60/400]  eta: 0:03:31  lr: 0.000221  loss: 2.5136 (2.5396)  grad: 0.2351 (0.2305)  time: 0.6125  data: 0.0036  max mem: 57344
+train: [10]  [ 80/400]  eta: 0:03:18  lr: 0.000220  loss: 2.5578 (2.5504)  grad: 0.2349 (0.2311)  time: 0.6122  data: 0.0035  max mem: 57344
+train: [10]  [100/400]  eta: 0:03:05  lr: 0.000218  loss: 2.5602 (2.5525)  grad: 0.2349 (0.2315)  time: 0.6124  data: 0.0036  max mem: 57344
+train: [10]  [120/400]  eta: 0:02:52  lr: 0.000217  loss: 2.5719 (2.5625)  grad: 0.2287 (0.2318)  time: 0.6121  data: 0.0036  max mem: 57344
+train: [10]  [140/400]  eta: 0:02:40  lr: 0.000215  loss: 2.5454 (2.5578)  grad: 0.2324 (0.2325)  time: 0.6125  data: 0.0036  max mem: 57344
+train: [10]  [160/400]  eta: 0:02:27  lr: 0.000214  loss: 2.5473 (2.5608)  grad: 0.2379 (0.2345)  time: 0.6122  data: 0.0036  max mem: 57344
+train: [10]  [180/400]  eta: 0:02:15  lr: 0.000213  loss: 2.5563 (2.5608)  grad: 0.2397 (0.2348)  time: 0.6122  data: 0.0035  max mem: 57344
+train: [10]  [200/400]  eta: 0:02:03  lr: 0.000211  loss: 2.5383 (2.5585)  grad: 0.2348 (0.2349)  time: 0.6122  data: 0.0036  max mem: 57344
+train: [10]  [220/400]  eta: 0:01:50  lr: 0.000210  loss: 2.5388 (2.5589)  grad: 0.2356 (0.2347)  time: 0.6116  data: 0.0034  max mem: 57344
+train: [10]  [240/400]  eta: 0:01:38  lr: 0.000208  loss: 2.5328 (2.5570)  grad: 0.2345 (0.2345)  time: 0.6115  data: 0.0033  max mem: 57344
+train: [10]  [260/400]  eta: 0:01:25  lr: 0.000207  loss: 2.5334 (2.5582)  grad: 0.2308 (0.2342)  time: 0.6116  data: 0.0034  max mem: 57344
+train: [10]  [280/400]  eta: 0:01:13  lr: 0.000205  loss: 2.5707 (2.5600)  grad: 0.2227 (0.2335)  time: 0.6116  data: 0.0034  max mem: 57344
+train: [10]  [300/400]  eta: 0:01:01  lr: 0.000204  loss: 2.6222 (2.5633)  grad: 0.2221 (0.2331)  time: 0.6129  data: 0.0036  max mem: 57344
+train: [10]  [320/400]  eta: 0:00:49  lr: 0.000202  loss: 2.5668 (2.5645)  grad: 0.2259 (0.2326)  time: 0.6127  data: 0.0036  max mem: 57344
+train: [10]  [340/400]  eta: 0:00:36  lr: 0.000201  loss: 2.5651 (2.5647)  grad: 0.2251 (0.2321)  time: 0.6127  data: 0.0035  max mem: 57344
+train: [10]  [360/400]  eta: 0:00:24  lr: 0.000199  loss: 2.5588 (2.5629)  grad: 0.2218 (0.2313)  time: 0.6126  data: 0.0036  max mem: 57344
+train: [10]  [380/400]  eta: 0:00:12  lr: 0.000198  loss: 2.5228 (2.5629)  grad: 0.2219 (0.2311)  time: 0.6122  data: 0.0035  max mem: 57344
+train: [10]  [399/400]  eta: 0:00:00  lr: 0.000196  loss: 2.5391 (2.5620)  grad: 0.2217 (0.2307)  time: 0.6125  data: 0.0036  max mem: 57344
+train: [10] Total time: 0:04:05 (0.6139 s / it)
+train: [10] Summary: lr: 0.000196  loss: 2.5391 (2.5620)  grad: 0.2217 (0.2307)
+eval (validation): [10]  [ 0/85]  eta: 0:01:16    time: 0.9007  data: 0.5441  max mem: 57344
+eval (validation): [10]  [20/85]  eta: 0:00:25    time: 0.3680  data: 0.0029  max mem: 57344
+eval (validation): [10]  [40/85]  eta: 0:00:17    time: 0.3683  data: 0.0036  max mem: 57344
+eval (validation): [10]  [60/85]  eta: 0:00:09    time: 0.3680  data: 0.0033  max mem: 57344
+eval (validation): [10]  [80/85]  eta: 0:00:01    time: 0.3684  data: 0.0033  max mem: 57344
+eval (validation): [10]  [84/85]  eta: 0:00:00    time: 0.3619  data: 0.0033  max mem: 57344
+eval (validation): [10] Total time: 0:00:31 (0.3741 s / it)
+cv: [10]  best hparam: (1.6, 1.0) (027) ('027_lr1.6e+00_wd1.0e+00')  loss: 2.365  acc: 0.290  f1: 0.227
+saving checkpoint experiments/schaefer1000/output/schaefer1000/schaefer1000_lr3e-4_1/eval_v2/nsd_cococlip__patch__attn/checkpoint-last.pth
+saving best checkpoint experiments/schaefer1000/output/schaefer1000/schaefer1000_lr3e-4_1/eval_v2/nsd_cococlip__patch__attn/checkpoint-best.pth
+train: [11]  [  0/400]  eta: 0:08:18  lr: nan  time: 1.2472  data: 0.6477  max mem: 57344
+train: [11]  [ 20/400]  eta: 0:04:03  lr: 0.000195  loss: 2.5126 (2.5227)  grad: 0.2249 (0.2316)  time: 0.6102  data: 0.0023  max mem: 57344
+train: [11]  [ 40/400]  eta: 0:03:45  lr: 0.000193  loss: 2.5098 (2.5174)  grad: 0.2277 (0.2294)  time: 0.6119  data: 0.0034  max mem: 57344
+train: [11]  [ 60/400]  eta: 0:03:31  lr: 0.000192  loss: 2.5129 (2.5235)  grad: 0.2209 (0.2259)  time: 0.6125  data: 0.0036  max mem: 57344
+train: [11]  [ 80/400]  eta: 0:03:18  lr: 0.000190  loss: 2.5329 (2.5259)  grad: 0.2209 (0.2260)  time: 0.6133  data: 0.0036  max mem: 57344
+train: [11]  [100/400]  eta: 0:03:05  lr: 0.000189  loss: 2.5488 (2.5400)  grad: 0.2244 (0.2281)  time: 0.6123  data: 0.0036  max mem: 57344
+train: [11]  [120/400]  eta: 0:02:52  lr: 0.000187  loss: 2.5692 (2.5457)  grad: 0.2321 (0.2289)  time: 0.6126  data: 0.0039  max mem: 57344
+train: [11]  [140/400]  eta: 0:02:40  lr: 0.000186  loss: 2.5534 (2.5446)  grad: 0.2239 (0.2271)  time: 0.6123  data: 0.0036  max mem: 57344
+train: [11]  [160/400]  eta: 0:02:27  lr: 0.000184  loss: 2.5354 (2.5411)  grad: 0.2133 (0.2252)  time: 0.6126  data: 0.0037  max mem: 57344
+train: [11]  [180/400]  eta: 0:02:15  lr: 0.000183  loss: 2.4981 (2.5368)  grad: 0.2152 (0.2254)  time: 0.6124  data: 0.0036  max mem: 57344
+train: [11]  [200/400]  eta: 0:02:03  lr: 0.000181  loss: 2.5050 (2.5360)  grad: 0.2303 (0.2261)  time: 0.6125  data: 0.0036  max mem: 57344
+train: [11]  [220/400]  eta: 0:01:50  lr: 0.000180  loss: 2.5230 (2.5358)  grad: 0.2303 (0.2263)  time: 0.6127  data: 0.0036  max mem: 57344
+train: [11]  [240/400]  eta: 0:01:38  lr: 0.000178  loss: 2.5638 (2.5395)  grad: 0.2329 (0.2280)  time: 0.6124  data: 0.0036  max mem: 57344
+train: [11]  [260/400]  eta: 0:01:26  lr: 0.000177  loss: 2.5752 (2.5392)  grad: 0.2382 (0.2286)  time: 0.6121  data: 0.0036  max mem: 57344
+train: [11]  [280/400]  eta: 0:01:13  lr: 0.000175  loss: 2.5095 (2.5361)  grad: 0.2350 (0.2289)  time: 0.6122  data: 0.0037  max mem: 57344
+train: [11]  [300/400]  eta: 0:01:01  lr: 0.000174  loss: 2.5095 (2.5354)  grad: 0.2303 (0.2293)  time: 0.6125  data: 0.0036  max mem: 57344
+train: [11]  [320/400]  eta: 0:00:49  lr: 0.000172  loss: 2.5111 (2.5340)  grad: 0.2303 (0.2294)  time: 0.6127  data: 0.0036  max mem: 57344
+train: [11]  [340/400]  eta: 0:00:36  lr: 0.000170  loss: 2.5196 (2.5352)  grad: 0.2299 (0.2294)  time: 0.6120  data: 0.0037  max mem: 57344
+train: [11]  [360/400]  eta: 0:00:24  lr: 0.000169  loss: 2.5003 (2.5323)  grad: 0.2340 (0.2296)  time: 0.6129  data: 0.0037  max mem: 57344
+train: [11]  [380/400]  eta: 0:00:12  lr: 0.000167  loss: 2.5216 (2.5342)  grad: 0.2352 (0.2298)  time: 0.6126  data: 0.0037  max mem: 57344
+train: [11]  [399/400]  eta: 0:00:00  lr: 0.000166  loss: 2.5683 (2.5358)  grad: 0.2314 (0.2299)  time: 0.6126  data: 0.0036  max mem: 57344
+train: [11] Total time: 0:04:05 (0.6142 s / it)
+train: [11] Summary: lr: 0.000166  loss: 2.5683 (2.5358)  grad: 0.2314 (0.2299)
+eval (validation): [11]  [ 0/85]  eta: 0:01:17    time: 0.9062  data: 0.5466  max mem: 57344
+eval (validation): [11]  [20/85]  eta: 0:00:25    time: 0.3682  data: 0.0031  max mem: 57344
+eval (validation): [11]  [40/85]  eta: 0:00:17    time: 0.3684  data: 0.0036  max mem: 57344
+eval (validation): [11]  [60/85]  eta: 0:00:09    time: 0.3679  data: 0.0034  max mem: 57344
+eval (validation): [11]  [80/85]  eta: 0:00:01    time: 0.3684  data: 0.0035  max mem: 57344
+eval (validation): [11]  [84/85]  eta: 0:00:00    time: 0.3620  data: 0.0034  max mem: 57344
+eval (validation): [11] Total time: 0:00:31 (0.3742 s / it)
+cv: [11]  best hparam: (0.72, 1.0) (022) ('022_lr7.2e-01_wd1.0e+00')  loss: 2.399  acc: 0.283  f1: 0.223
+saving checkpoint experiments/schaefer1000/output/schaefer1000/schaefer1000_lr3e-4_1/eval_v2/nsd_cococlip__patch__attn/checkpoint-last.pth
+train: [12]  [  0/400]  eta: 0:07:57  lr: nan  time: 1.1935  data: 0.5947  max mem: 57344
+train: [12]  [ 20/400]  eta: 0:04:02  lr: 0.000164  loss: 2.5259 (2.5431)  grad: 0.2249 (0.2263)  time: 0.6114  data: 0.0030  max mem: 57344
+train: [12]  [ 40/400]  eta: 0:03:45  lr: 0.000163  loss: 2.5397 (2.5334)  grad: 0.2254 (0.2273)  time: 0.6128  data: 0.0037  max mem: 57344
+train: [12]  [ 60/400]  eta: 0:03:31  lr: 0.000161  loss: 2.5376 (2.5276)  grad: 0.2249 (0.2253)  time: 0.6126  data: 0.0036  max mem: 57344
+train: [12]  [ 80/400]  eta: 0:03:18  lr: 0.000160  loss: 2.5102 (2.5110)  grad: 0.2202 (0.2245)  time: 0.6125  data: 0.0036  max mem: 57344
+train: [12]  [100/400]  eta: 0:03:05  lr: 0.000158  loss: 2.4724 (2.5106)  grad: 0.2202 (0.2252)  time: 0.6122  data: 0.0036  max mem: 57344
+train: [12]  [120/400]  eta: 0:02:52  lr: 0.000156  loss: 2.5048 (2.5156)  grad: 0.2307 (0.2273)  time: 0.6123  data: 0.0035  max mem: 57344
+train: [12]  [140/400]  eta: 0:02:40  lr: 0.000155  loss: 2.5260 (2.5152)  grad: 0.2307 (0.2279)  time: 0.6124  data: 0.0036  max mem: 57344
+train: [12]  [160/400]  eta: 0:02:27  lr: 0.000153  loss: 2.5164 (2.5136)  grad: 0.2278 (0.2280)  time: 0.6125  data: 0.0036  max mem: 57344
+train: [12]  [180/400]  eta: 0:02:15  lr: 0.000152  loss: 2.4777 (2.5107)  grad: 0.2278 (0.2285)  time: 0.6121  data: 0.0036  max mem: 57344
+train: [12]  [200/400]  eta: 0:02:03  lr: 0.000150  loss: 2.5071 (2.5137)  grad: 0.2351 (0.2296)  time: 0.6128  data: 0.0036  max mem: 57344
+train: [12]  [220/400]  eta: 0:01:50  lr: 0.000149  loss: 2.5071 (2.5101)  grad: 0.2351 (0.2298)  time: 0.6126  data: 0.0036  max mem: 57344
+train: [12]  [240/400]  eta: 0:01:38  lr: 0.000147  loss: 2.4824 (2.5080)  grad: 0.2190 (0.2288)  time: 0.6123  data: 0.0036  max mem: 57344
+train: [12]  [260/400]  eta: 0:01:26  lr: 0.000145  loss: 2.5058 (2.5072)  grad: 0.2185 (0.2281)  time: 0.6121  data: 0.0036  max mem: 57344
+train: [12]  [280/400]  eta: 0:01:13  lr: 0.000144  loss: 2.5213 (2.5107)  grad: 0.2218 (0.2283)  time: 0.6122  data: 0.0036  max mem: 57344
+train: [12]  [300/400]  eta: 0:01:01  lr: 0.000142  loss: 2.5213 (2.5125)  grad: 0.2271 (0.2284)  time: 0.6124  data: 0.0036  max mem: 57344
+train: [12]  [320/400]  eta: 0:00:49  lr: 0.000141  loss: 2.4987 (2.5114)  grad: 0.2278 (0.2285)  time: 0.6125  data: 0.0036  max mem: 57344
+train: [12]  [340/400]  eta: 0:00:36  lr: 0.000139  loss: 2.4988 (2.5126)  grad: 0.2278 (0.2285)  time: 0.6122  data: 0.0037  max mem: 57344
+train: [12]  [360/400]  eta: 0:00:24  lr: 0.000138  loss: 2.4947 (2.5107)  grad: 0.2285 (0.2286)  time: 0.6123  data: 0.0036  max mem: 57344
+train: [12]  [380/400]  eta: 0:00:12  lr: 0.000136  loss: 2.4760 (2.5124)  grad: 0.2275 (0.2287)  time: 0.6121  data: 0.0036  max mem: 57344
+train: [12]  [399/400]  eta: 0:00:00  lr: 0.000134  loss: 2.4760 (2.5111)  grad: 0.2265 (0.2284)  time: 0.6123  data: 0.0036  max mem: 57344
+train: [12] Total time: 0:04:05 (0.6140 s / it)
+train: [12] Summary: lr: 0.000134  loss: 2.4760 (2.5111)  grad: 0.2265 (0.2284)
+eval (validation): [12]  [ 0/85]  eta: 0:01:14    time: 0.8750  data: 0.5153  max mem: 57344
+eval (validation): [12]  [20/85]  eta: 0:00:25    time: 0.3684  data: 0.0035  max mem: 57344
+eval (validation): [12]  [40/85]  eta: 0:00:17    time: 0.3686  data: 0.0034  max mem: 57344
+eval (validation): [12]  [60/85]  eta: 0:00:09    time: 0.3681  data: 0.0034  max mem: 57344
+eval (validation): [12]  [80/85]  eta: 0:00:01    time: 0.3683  data: 0.0036  max mem: 57344
+eval (validation): [12]  [84/85]  eta: 0:00:00    time: 0.3624  data: 0.0035  max mem: 57344
+eval (validation): [12] Total time: 0:00:31 (0.3740 s / it)
+cv: [12]  best hparam: (2.7, 1.0) (030) ('030_lr2.7e+00_wd1.0e+00')  loss: 2.373  acc: 0.296  f1: 0.233
+saving checkpoint experiments/schaefer1000/output/schaefer1000/schaefer1000_lr3e-4_1/eval_v2/nsd_cococlip__patch__attn/checkpoint-last.pth
+saving best checkpoint experiments/schaefer1000/output/schaefer1000/schaefer1000_lr3e-4_1/eval_v2/nsd_cococlip__patch__attn/checkpoint-best.pth
+train: [13]  [  0/400]  eta: 0:07:59  lr: nan  time: 1.1997  data: 0.6000  max mem: 57344
+train: [13]  [ 20/400]  eta: 0:04:02  lr: 0.000133  loss: 2.4349 (2.4554)  grad: 0.2322 (0.2314)  time: 0.6101  data: 0.0026  max mem: 57344
+train: [13]  [ 40/400]  eta: 0:03:45  lr: 0.000131  loss: 2.4349 (2.4568)  grad: 0.2292 (0.2289)  time: 0.6128  data: 0.0037  max mem: 57344
+train: [13]  [ 60/400]  eta: 0:03:31  lr: 0.000130  loss: 2.4246 (2.4550)  grad: 0.2218 (0.2278)  time: 0.6137  data: 0.0037  max mem: 57344
+train: [13]  [ 80/400]  eta: 0:03:18  lr: 0.000128  loss: 2.4042 (2.4600)  grad: 0.2218 (0.2263)  time: 0.6131  data: 0.0037  max mem: 57344
+train: [13]  [100/400]  eta: 0:03:05  lr: 0.000127  loss: 2.4799 (2.4648)  grad: 0.2215 (0.2255)  time: 0.6133  data: 0.0036  max mem: 57344
+train: [13]  [120/400]  eta: 0:02:52  lr: 0.000125  loss: 2.5227 (2.4799)  grad: 0.2272 (0.2272)  time: 0.6128  data: 0.0037  max mem: 57344
+train: [13]  [140/400]  eta: 0:02:40  lr: 0.000124  loss: 2.5227 (2.4832)  grad: 0.2258 (0.2269)  time: 0.6130  data: 0.0037  max mem: 57344
+train: [13]  [160/400]  eta: 0:02:27  lr: 0.000122  loss: 2.5064 (2.4907)  grad: 0.2258 (0.2275)  time: 0.6133  data: 0.0037  max mem: 57344
+train: [13]  [180/400]  eta: 0:02:15  lr: 0.000120  loss: 2.5064 (2.4916)  grad: 0.2312 (0.2278)  time: 0.6131  data: 0.0036  max mem: 57344
+train: [13]  [200/400]  eta: 0:02:03  lr: 0.000119  loss: 2.4594 (2.4873)  grad: 0.2212 (0.2276)  time: 0.6133  data: 0.0036  max mem: 57344
+train: [13]  [220/400]  eta: 0:01:50  lr: 0.000117  loss: 2.4594 (2.4873)  grad: 0.2316 (0.2280)  time: 0.6128  data: 0.0036  max mem: 57344
+train: [13]  [240/400]  eta: 0:01:38  lr: 0.000116  loss: 2.4541 (2.4839)  grad: 0.2285 (0.2274)  time: 0.6135  data: 0.0036  max mem: 57344
+train: [13]  [260/400]  eta: 0:01:26  lr: 0.000114  loss: 2.4439 (2.4806)  grad: 0.2217 (0.2268)  time: 0.6132  data: 0.0036  max mem: 57344
+train: [13]  [280/400]  eta: 0:01:13  lr: 0.000113  loss: 2.4447 (2.4788)  grad: 0.2279 (0.2274)  time: 0.6131  data: 0.0036  max mem: 57344
+train: [13]  [300/400]  eta: 0:01:01  lr: 0.000111  loss: 2.4476 (2.4783)  grad: 0.2306 (0.2279)  time: 0.6127  data: 0.0036  max mem: 57344
+train: [13]  [320/400]  eta: 0:00:49  lr: 0.000110  loss: 2.4741 (2.4777)  grad: 0.2295 (0.2280)  time: 0.6126  data: 0.0036  max mem: 57344
+train: [13]  [340/400]  eta: 0:00:36  lr: 0.000108  loss: 2.4930 (2.4807)  grad: 0.2295 (0.2284)  time: 0.6131  data: 0.0036  max mem: 57344
+train: [13]  [360/400]  eta: 0:00:24  lr: 0.000107  loss: 2.5318 (2.4814)  grad: 0.2333 (0.2289)  time: 0.6131  data: 0.0037  max mem: 57344
+train: [13]  [380/400]  eta: 0:00:12  lr: 0.000105  loss: 2.4900 (2.4820)  grad: 0.2291 (0.2291)  time: 0.6132  data: 0.0036  max mem: 57344
+train: [13]  [399/400]  eta: 0:00:00  lr: 0.000104  loss: 2.4952 (2.4833)  grad: 0.2331 (0.2294)  time: 0.6130  data: 0.0036  max mem: 57344
+train: [13] Total time: 0:04:05 (0.6147 s / it)
+train: [13] Summary: lr: 0.000104  loss: 2.4952 (2.4833)  grad: 0.2331 (0.2294)
+eval (validation): [13]  [ 0/85]  eta: 0:01:22    time: 0.9752  data: 0.6137  max mem: 57344
+eval (validation): [13]  [20/85]  eta: 0:00:25    time: 0.3679  data: 0.0023  max mem: 57344
+eval (validation): [13]  [40/85]  eta: 0:00:17    time: 0.3679  data: 0.0034  max mem: 57344
+eval (validation): [13]  [60/85]  eta: 0:00:09    time: 0.3686  data: 0.0036  max mem: 57344
+eval (validation): [13]  [80/85]  eta: 0:00:01    time: 0.3685  data: 0.0036  max mem: 57344
+eval (validation): [13]  [84/85]  eta: 0:00:00    time: 0.3622  data: 0.0035  max mem: 57344
+eval (validation): [13] Total time: 0:00:31 (0.3751 s / it)
+cv: [13]  best hparam: (2.7, 1.0) (030) ('030_lr2.7e+00_wd1.0e+00')  loss: 2.356  acc: 0.298  f1: 0.243
+saving checkpoint experiments/schaefer1000/output/schaefer1000/schaefer1000_lr3e-4_1/eval_v2/nsd_cococlip__patch__attn/checkpoint-last.pth
+saving best checkpoint experiments/schaefer1000/output/schaefer1000/schaefer1000_lr3e-4_1/eval_v2/nsd_cococlip__patch__attn/checkpoint-best.pth
+train: [14]  [  0/400]  eta: 0:07:42  lr: nan  time: 1.1571  data: 0.5587  max mem: 57344
+train: [14]  [ 20/400]  eta: 0:04:02  lr: 0.000102  loss: 2.4120 (2.4233)  grad: 0.2299 (0.2303)  time: 0.6117  data: 0.0031  max mem: 57344
+train: [14]  [ 40/400]  eta: 0:03:45  lr: 0.000101  loss: 2.4800 (2.4603)  grad: 0.2299 (0.2290)  time: 0.6119  data: 0.0035  max mem: 57344
+train: [14]  [ 60/400]  eta: 0:03:31  lr: 0.000099  loss: 2.4709 (2.4494)  grad: 0.2224 (0.2236)  time: 0.6123  data: 0.0036  max mem: 57344
+train: [14]  [ 80/400]  eta: 0:03:18  lr: 0.000098  loss: 2.4387 (2.4519)  grad: 0.2133 (0.2241)  time: 0.6126  data: 0.0036  max mem: 57344
+train: [14]  [100/400]  eta: 0:03:05  lr: 0.000096  loss: 2.4311 (2.4444)  grad: 0.2243 (0.2251)  time: 0.6126  data: 0.0036  max mem: 57344
+train: [14]  [120/400]  eta: 0:02:52  lr: 0.000095  loss: 2.4295 (2.4400)  grad: 0.2236 (0.2241)  time: 0.6121  data: 0.0036  max mem: 57344
+train: [14]  [140/400]  eta: 0:02:40  lr: 0.000093  loss: 2.4286 (2.4389)  grad: 0.2224 (0.2255)  time: 0.6127  data: 0.0036  max mem: 57344
+train: [14]  [160/400]  eta: 0:02:27  lr: 0.000092  loss: 2.4631 (2.4435)  grad: 0.2224 (0.2255)  time: 0.6122  data: 0.0036  max mem: 57344
+train: [14]  [180/400]  eta: 0:02:15  lr: 0.000090  loss: 2.4582 (2.4418)  grad: 0.2258 (0.2257)  time: 0.6126  data: 0.0037  max mem: 57344
+train: [14]  [200/400]  eta: 0:02:03  lr: 0.000089  loss: 2.4276 (2.4449)  grad: 0.2247 (0.2252)  time: 0.6124  data: 0.0036  max mem: 57344
+train: [14]  [220/400]  eta: 0:01:50  lr: 0.000088  loss: 2.4317 (2.4467)  grad: 0.2182 (0.2246)  time: 0.6121  data: 0.0037  max mem: 57344
+train: [14]  [240/400]  eta: 0:01:38  lr: 0.000086  loss: 2.4522 (2.4499)  grad: 0.2190 (0.2248)  time: 0.6126  data: 0.0037  max mem: 57344
+train: [14]  [260/400]  eta: 0:01:26  lr: 0.000085  loss: 2.4522 (2.4499)  grad: 0.2207 (0.2249)  time: 0.6130  data: 0.0037  max mem: 57344
+train: [14]  [280/400]  eta: 0:01:13  lr: 0.000083  loss: 2.4636 (2.4522)  grad: 0.2250 (0.2251)  time: 0.6123  data: 0.0036  max mem: 57344
+train: [14]  [300/400]  eta: 0:01:01  lr: 0.000082  loss: 2.4810 (2.4537)  grad: 0.2245 (0.2252)  time: 0.6124  data: 0.0036  max mem: 57344
+train: [14]  [320/400]  eta: 0:00:49  lr: 0.000081  loss: 2.4727 (2.4535)  grad: 0.2216 (0.2251)  time: 0.6120  data: 0.0036  max mem: 57344
+train: [14]  [340/400]  eta: 0:00:36  lr: 0.000079  loss: 2.4727 (2.4540)  grad: 0.2183 (0.2248)  time: 0.6126  data: 0.0037  max mem: 57344
+train: [14]  [360/400]  eta: 0:00:24  lr: 0.000078  loss: 2.4456 (2.4536)  grad: 0.2187 (0.2247)  time: 0.6129  data: 0.0038  max mem: 57344
+train: [14]  [380/400]  eta: 0:00:12  lr: 0.000076  loss: 2.4440 (2.4528)  grad: 0.2239 (0.2248)  time: 0.6127  data: 0.0037  max mem: 57344
+train: [14]  [399/400]  eta: 0:00:00  lr: 0.000075  loss: 2.4283 (2.4516)  grad: 0.2254 (0.2249)  time: 0.6125  data: 0.0036  max mem: 57344
+train: [14] Total time: 0:04:05 (0.6140 s / it)
+train: [14] Summary: lr: 0.000075  loss: 2.4283 (2.4516)  grad: 0.2254 (0.2249)
+eval (validation): [14]  [ 0/85]  eta: 0:01:18    time: 0.9212  data: 0.5617  max mem: 57344
+eval (validation): [14]  [20/85]  eta: 0:00:25    time: 0.3686  data: 0.0032  max mem: 57344
+eval (validation): [14]  [40/85]  eta: 0:00:17    time: 0.3686  data: 0.0035  max mem: 57344
+eval (validation): [14]  [60/85]  eta: 0:00:09    time: 0.3686  data: 0.0037  max mem: 57344
+eval (validation): [14]  [80/85]  eta: 0:00:01    time: 0.3697  data: 0.0036  max mem: 57344
+eval (validation): [14]  [84/85]  eta: 0:00:00    time: 0.3632  data: 0.0036  max mem: 57344
+eval (validation): [14] Total time: 0:00:31 (0.3751 s / it)
+cv: [14]  best hparam: (2.7, 1.0) (030) ('030_lr2.7e+00_wd1.0e+00')  loss: 2.375  acc: 0.294  f1: 0.240
+saving checkpoint experiments/schaefer1000/output/schaefer1000/schaefer1000_lr3e-4_1/eval_v2/nsd_cococlip__patch__attn/checkpoint-last.pth
+train: [15]  [  0/400]  eta: 0:07:55  lr: nan  time: 1.1889  data: 0.5892  max mem: 57344
+train: [15]  [ 20/400]  eta: 0:04:03  lr: 0.000074  loss: 2.3708 (2.3955)  grad: 0.2254 (0.2297)  time: 0.6121  data: 0.0031  max mem: 57344
+train: [15]  [ 40/400]  eta: 0:03:45  lr: 0.000072  loss: 2.3600 (2.3877)  grad: 0.2323 (0.2295)  time: 0.6128  data: 0.0037  max mem: 57344
+train: [15]  [ 60/400]  eta: 0:03:31  lr: 0.000071  loss: 2.4186 (2.4092)  grad: 0.2334 (0.2316)  time: 0.6127  data: 0.0037  max mem: 57344
+train: [15]  [ 80/400]  eta: 0:03:18  lr: 0.000070  loss: 2.4390 (2.4135)  grad: 0.2321 (0.2319)  time: 0.6126  data: 0.0037  max mem: 57344
+train: [15]  [100/400]  eta: 0:03:05  lr: 0.000068  loss: 2.4206 (2.4155)  grad: 0.2291 (0.2313)  time: 0.6130  data: 0.0038  max mem: 57344
+train: [15]  [120/400]  eta: 0:02:52  lr: 0.000067  loss: 2.4347 (2.4230)  grad: 0.2285 (0.2308)  time: 0.6127  data: 0.0037  max mem: 57344
+train: [15]  [140/400]  eta: 0:02:40  lr: 0.000066  loss: 2.4460 (2.4253)  grad: 0.2281 (0.2311)  time: 0.6129  data: 0.0037  max mem: 57344
+train: [15]  [160/400]  eta: 0:02:27  lr: 0.000064  loss: 2.4460 (2.4238)  grad: 0.2280 (0.2304)  time: 0.6121  data: 0.0035  max mem: 57344
+train: [15]  [180/400]  eta: 0:02:15  lr: 0.000063  loss: 2.4061 (2.4234)  grad: 0.2194 (0.2300)  time: 0.6120  data: 0.0035  max mem: 57344
+train: [15]  [200/400]  eta: 0:02:03  lr: 0.000062  loss: 2.4219 (2.4294)  grad: 0.2313 (0.2307)  time: 0.6124  data: 0.0037  max mem: 57344
+train: [15]  [220/400]  eta: 0:01:50  lr: 0.000061  loss: 2.4690 (2.4324)  grad: 0.2330 (0.2305)  time: 0.6125  data: 0.0036  max mem: 57344
+train: [15]  [240/400]  eta: 0:01:38  lr: 0.000059  loss: 2.4279 (2.4286)  grad: 0.2278 (0.2300)  time: 0.6126  data: 0.0036  max mem: 57344
+train: [15]  [260/400]  eta: 0:01:26  lr: 0.000058  loss: 2.4151 (2.4302)  grad: 0.2196 (0.2294)  time: 0.6125  data: 0.0037  max mem: 57344
+train: [15]  [280/400]  eta: 0:01:13  lr: 0.000057  loss: 2.4411 (2.4298)  grad: 0.2174 (0.2289)  time: 0.6126  data: 0.0037  max mem: 57344
+train: [15]  [300/400]  eta: 0:01:01  lr: 0.000056  loss: 2.4083 (2.4274)  grad: 0.2154 (0.2281)  time: 0.6125  data: 0.0036  max mem: 57344
+train: [15]  [320/400]  eta: 0:00:49  lr: 0.000054  loss: 2.4408 (2.4303)  grad: 0.2154 (0.2280)  time: 0.6128  data: 0.0036  max mem: 57344
+train: [15]  [340/400]  eta: 0:00:36  lr: 0.000053  loss: 2.3974 (2.4259)  grad: 0.2200 (0.2275)  time: 0.6121  data: 0.0037  max mem: 57344
+train: [15]  [360/400]  eta: 0:00:24  lr: 0.000052  loss: 2.3779 (2.4252)  grad: 0.2197 (0.2272)  time: 0.6127  data: 0.0037  max mem: 57344
+train: [15]  [380/400]  eta: 0:00:12  lr: 0.000051  loss: 2.3852 (2.4259)  grad: 0.2194 (0.2268)  time: 0.6126  data: 0.0036  max mem: 57344
+train: [15]  [399/400]  eta: 0:00:00  lr: 0.000050  loss: 2.4197 (2.4273)  grad: 0.2285 (0.2272)  time: 0.6122  data: 0.0036  max mem: 57344
+train: [15] Total time: 0:04:05 (0.6142 s / it)
+train: [15] Summary: lr: 0.000050  loss: 2.4197 (2.4273)  grad: 0.2285 (0.2272)
+eval (validation): [15]  [ 0/85]  eta: 0:01:17    time: 0.9136  data: 0.5558  max mem: 57344
+eval (validation): [15]  [20/85]  eta: 0:00:25    time: 0.3677  data: 0.0030  max mem: 57344
+eval (validation): [15]  [40/85]  eta: 0:00:17    time: 0.3681  data: 0.0035  max mem: 57344
+eval (validation): [15]  [60/85]  eta: 0:00:09    time: 0.3691  data: 0.0035  max mem: 57344
+eval (validation): [15]  [80/85]  eta: 0:00:01    time: 0.3690  data: 0.0036  max mem: 57344
+eval (validation): [15]  [84/85]  eta: 0:00:00    time: 0.3630  data: 0.0036  max mem: 57344
+eval (validation): [15] Total time: 0:00:31 (0.3747 s / it)
+cv: [15]  best hparam: (3.1, 1.0) (031) ('031_lr3.1e+00_wd1.0e+00')  loss: 2.380  acc: 0.297  f1: 0.245
+saving checkpoint experiments/schaefer1000/output/schaefer1000/schaefer1000_lr3e-4_1/eval_v2/nsd_cococlip__patch__attn/checkpoint-last.pth
+train: [16]  [  0/400]  eta: 0:07:46  lr: nan  time: 1.1653  data: 0.5648  max mem: 57344
+train: [16]  [ 20/400]  eta: 0:04:02  lr: 0.000048  loss: 2.3811 (2.4100)  grad: 0.2196 (0.2249)  time: 0.6117  data: 0.0031  max mem: 57344
+train: [16]  [ 40/400]  eta: 0:03:45  lr: 0.000047  loss: 2.3719 (2.4089)  grad: 0.2172 (0.2216)  time: 0.6121  data: 0.0036  max mem: 57344
+train: [16]  [ 60/400]  eta: 0:03:31  lr: 0.000046  loss: 2.3555 (2.3887)  grad: 0.2126 (0.2195)  time: 0.6128  data: 0.0036  max mem: 57344
+train: [16]  [ 80/400]  eta: 0:03:18  lr: 0.000045  loss: 2.3748 (2.3868)  grad: 0.2172 (0.2193)  time: 0.6125  data: 0.0036  max mem: 57344
+train: [16]  [100/400]  eta: 0:03:05  lr: 0.000044  loss: 2.3896 (2.3845)  grad: 0.2173 (0.2194)  time: 0.6125  data: 0.0036  max mem: 57344
+train: [16]  [120/400]  eta: 0:02:52  lr: 0.000043  loss: 2.4427 (2.3974)  grad: 0.2180 (0.2199)  time: 0.6124  data: 0.0035  max mem: 57344
+train: [16]  [140/400]  eta: 0:02:40  lr: 0.000042  loss: 2.4427 (2.3992)  grad: 0.2161 (0.2191)  time: 0.6124  data: 0.0036  max mem: 57344
+train: [16]  [160/400]  eta: 0:02:27  lr: 0.000041  loss: 2.4133 (2.3997)  grad: 0.2128 (0.2190)  time: 0.6124  data: 0.0036  max mem: 57344
+train: [16]  [180/400]  eta: 0:02:15  lr: 0.000040  loss: 2.4193 (2.4035)  grad: 0.2175 (0.2192)  time: 0.6130  data: 0.0037  max mem: 57344
+train: [16]  [200/400]  eta: 0:02:03  lr: 0.000039  loss: 2.4207 (2.4069)  grad: 0.2192 (0.2193)  time: 0.6124  data: 0.0036  max mem: 57344
+train: [16]  [220/400]  eta: 0:01:50  lr: 0.000038  loss: 2.4054 (2.4062)  grad: 0.2225 (0.2198)  time: 0.6128  data: 0.0036  max mem: 57344
+train: [16]  [240/400]  eta: 0:01:38  lr: 0.000036  loss: 2.4064 (2.4097)  grad: 0.2225 (0.2197)  time: 0.6126  data: 0.0036  max mem: 57344
+train: [16]  [260/400]  eta: 0:01:26  lr: 0.000035  loss: 2.4211 (2.4108)  grad: 0.2219 (0.2201)  time: 0.6125  data: 0.0037  max mem: 57344
+train: [16]  [280/400]  eta: 0:01:13  lr: 0.000034  loss: 2.4044 (2.4095)  grad: 0.2254 (0.2209)  time: 0.6127  data: 0.0037  max mem: 57344
+train: [16]  [300/400]  eta: 0:01:01  lr: 0.000033  loss: 2.3775 (2.4074)  grad: 0.2267 (0.2214)  time: 0.6122  data: 0.0036  max mem: 57344
+train: [16]  [320/400]  eta: 0:00:49  lr: 0.000032  loss: 2.4263 (2.4113)  grad: 0.2224 (0.2212)  time: 0.6126  data: 0.0037  max mem: 57344
+train: [16]  [340/400]  eta: 0:00:36  lr: 0.000031  loss: 2.4298 (2.4115)  grad: 0.2224 (0.2217)  time: 0.6128  data: 0.0036  max mem: 57344
+train: [16]  [360/400]  eta: 0:00:24  lr: 0.000031  loss: 2.4165 (2.4124)  grad: 0.2227 (0.2219)  time: 0.6127  data: 0.0036  max mem: 57344
+train: [16]  [380/400]  eta: 0:00:12  lr: 0.000030  loss: 2.4351 (2.4128)  grad: 0.2169 (0.2218)  time: 0.6129  data: 0.0035  max mem: 57344
+train: [16]  [399/400]  eta: 0:00:00  lr: 0.000029  loss: 2.4085 (2.4137)  grad: 0.2165 (0.2217)  time: 0.6124  data: 0.0036  max mem: 57344
+train: [16] Total time: 0:04:05 (0.6142 s / it)
+train: [16] Summary: lr: 0.000029  loss: 2.4085 (2.4137)  grad: 0.2165 (0.2217)
+eval (validation): [16]  [ 0/85]  eta: 0:01:17    time: 0.9060  data: 0.5469  max mem: 57344
+eval (validation): [16]  [20/85]  eta: 0:00:25    time: 0.3674  data: 0.0025  max mem: 57344
+eval (validation): [16]  [40/85]  eta: 0:00:17    time: 0.3681  data: 0.0036  max mem: 57344
+eval (validation): [16]  [60/85]  eta: 0:00:09    time: 0.3689  data: 0.0034  max mem: 57344
+eval (validation): [16]  [80/85]  eta: 0:00:01    time: 0.3691  data: 0.0034  max mem: 57344
+eval (validation): [16]  [84/85]  eta: 0:00:00    time: 0.3631  data: 0.0034  max mem: 57344
+eval (validation): [16] Total time: 0:00:31 (0.3744 s / it)
+cv: [16]  best hparam: (2.7, 1.0) (030) ('030_lr2.7e+00_wd1.0e+00')  loss: 2.398  acc: 0.291  f1: 0.234
+saving checkpoint experiments/schaefer1000/output/schaefer1000/schaefer1000_lr3e-4_1/eval_v2/nsd_cococlip__patch__attn/checkpoint-last.pth
+train: [17]  [  0/400]  eta: 0:07:57  lr: nan  time: 1.1938  data: 0.5918  max mem: 57344
+train: [17]  [ 20/400]  eta: 0:04:03  lr: 0.000028  loss: 2.3969 (2.4198)  grad: 0.2116 (0.2136)  time: 0.6121  data: 0.0031  max mem: 57344
+train: [17]  [ 40/400]  eta: 0:03:45  lr: 0.000027  loss: 2.3875 (2.3834)  grad: 0.2208 (0.2204)  time: 0.6128  data: 0.0037  max mem: 57344
+train: [17]  [ 60/400]  eta: 0:03:31  lr: 0.000026  loss: 2.3649 (2.3768)  grad: 0.2227 (0.2204)  time: 0.6126  data: 0.0037  max mem: 57344
+train: [17]  [ 80/400]  eta: 0:03:18  lr: 0.000025  loss: 2.4160 (2.3922)  grad: 0.2217 (0.2205)  time: 0.6125  data: 0.0037  max mem: 57344
+train: [17]  [100/400]  eta: 0:03:05  lr: 0.000024  loss: 2.4510 (2.4010)  grad: 0.2214 (0.2212)  time: 0.6128  data: 0.0037  max mem: 57344
+train: [17]  [120/400]  eta: 0:02:52  lr: 0.000023  loss: 2.4353 (2.3984)  grad: 0.2214 (0.2218)  time: 0.6122  data: 0.0036  max mem: 57344
+train: [17]  [140/400]  eta: 0:02:40  lr: 0.000023  loss: 2.3842 (2.3940)  grad: 0.2184 (0.2209)  time: 0.6126  data: 0.0037  max mem: 57344
+train: [17]  [160/400]  eta: 0:02:27  lr: 0.000022  loss: 2.4057 (2.3978)  grad: 0.2176 (0.2208)  time: 0.6123  data: 0.0037  max mem: 57344
+train: [17]  [180/400]  eta: 0:02:15  lr: 0.000021  loss: 2.4123 (2.3963)  grad: 0.2198 (0.2209)  time: 0.6122  data: 0.0036  max mem: 57344
+train: [17]  [200/400]  eta: 0:02:03  lr: 0.000020  loss: 2.4074 (2.3966)  grad: 0.2188 (0.2206)  time: 0.6124  data: 0.0036  max mem: 57344
+train: [17]  [220/400]  eta: 0:01:50  lr: 0.000019  loss: 2.4031 (2.3939)  grad: 0.2157 (0.2205)  time: 0.6126  data: 0.0036  max mem: 57344
+train: [17]  [240/400]  eta: 0:01:38  lr: 0.000019  loss: 2.4031 (2.3928)  grad: 0.2153 (0.2198)  time: 0.6121  data: 0.0036  max mem: 57344
+train: [17]  [260/400]  eta: 0:01:26  lr: 0.000018  loss: 2.4079 (2.3942)  grad: 0.2150 (0.2202)  time: 0.6119  data: 0.0036  max mem: 57344
+train: [17]  [280/400]  eta: 0:01:13  lr: 0.000017  loss: 2.3857 (2.3931)  grad: 0.2206 (0.2206)  time: 0.6115  data: 0.0035  max mem: 57344
+train: [17]  [300/400]  eta: 0:01:01  lr: 0.000016  loss: 2.3792 (2.3910)  grad: 0.2216 (0.2206)  time: 0.6115  data: 0.0035  max mem: 57344
+train: [17]  [320/400]  eta: 0:00:49  lr: 0.000016  loss: 2.3916 (2.3923)  grad: 0.2201 (0.2202)  time: 0.6124  data: 0.0036  max mem: 57344
+train: [17]  [340/400]  eta: 0:00:36  lr: 0.000015  loss: 2.3996 (2.3924)  grad: 0.2129 (0.2199)  time: 0.6121  data: 0.0036  max mem: 57344
+train: [17]  [360/400]  eta: 0:00:24  lr: 0.000014  loss: 2.3750 (2.3905)  grad: 0.2165 (0.2201)  time: 0.6126  data: 0.0036  max mem: 57344
+train: [17]  [380/400]  eta: 0:00:12  lr: 0.000014  loss: 2.3882 (2.3934)  grad: 0.2165 (0.2201)  time: 0.6128  data: 0.0036  max mem: 57344
+train: [17]  [399/400]  eta: 0:00:00  lr: 0.000013  loss: 2.3999 (2.3929)  grad: 0.2157 (0.2204)  time: 0.6118  data: 0.0036  max mem: 57344
+train: [17] Total time: 0:04:05 (0.6140 s / it)
+train: [17] Summary: lr: 0.000013  loss: 2.3999 (2.3929)  grad: 0.2157 (0.2204)
+eval (validation): [17]  [ 0/85]  eta: 0:01:21    time: 0.9581  data: 0.5988  max mem: 57344
+eval (validation): [17]  [20/85]  eta: 0:00:25    time: 0.3673  data: 0.0026  max mem: 57344
+eval (validation): [17]  [40/85]  eta: 0:00:17    time: 0.3687  data: 0.0035  max mem: 57344
+eval (validation): [17]  [60/85]  eta: 0:00:09    time: 0.3690  data: 0.0036  max mem: 57344
+eval (validation): [17]  [80/85]  eta: 0:00:01    time: 0.3692  data: 0.0035  max mem: 57344
+eval (validation): [17]  [84/85]  eta: 0:00:00    time: 0.3628  data: 0.0035  max mem: 57344
+eval (validation): [17] Total time: 0:00:31 (0.3753 s / it)
+cv: [17]  best hparam: (3.1, 1.0) (031) ('031_lr3.1e+00_wd1.0e+00')  loss: 2.388  acc: 0.292  f1: 0.241
+saving checkpoint experiments/schaefer1000/output/schaefer1000/schaefer1000_lr3e-4_1/eval_v2/nsd_cococlip__patch__attn/checkpoint-last.pth
+train: [18]  [  0/400]  eta: 0:08:02  lr: nan  time: 1.2059  data: 0.6040  max mem: 57344
+train: [18]  [ 20/400]  eta: 0:04:03  lr: 0.000012  loss: 2.3533 (2.3449)  grad: 0.2158 (0.2186)  time: 0.6126  data: 0.0032  max mem: 57344
+train: [18]  [ 40/400]  eta: 0:03:45  lr: 0.000012  loss: 2.3515 (2.3399)  grad: 0.2158 (0.2163)  time: 0.6126  data: 0.0033  max mem: 57344
+train: [18]  [ 60/400]  eta: 0:03:31  lr: 0.000011  loss: 2.3202 (2.3354)  grad: 0.2168 (0.2182)  time: 0.6124  data: 0.0036  max mem: 57344
+train: [18]  [ 80/400]  eta: 0:03:18  lr: 0.000011  loss: 2.3227 (2.3393)  grad: 0.2123 (0.2168)  time: 0.6123  data: 0.0036  max mem: 57344
+train: [18]  [100/400]  eta: 0:03:05  lr: 0.000010  loss: 2.3635 (2.3501)  grad: 0.2202 (0.2194)  time: 0.6122  data: 0.0036  max mem: 57344
+train: [18]  [120/400]  eta: 0:02:52  lr: 0.000009  loss: 2.3798 (2.3554)  grad: 0.2284 (0.2198)  time: 0.6125  data: 0.0036  max mem: 57344
+train: [18]  [140/400]  eta: 0:02:40  lr: 0.000009  loss: 2.3676 (2.3519)  grad: 0.2184 (0.2183)  time: 0.6125  data: 0.0036  max mem: 57344
+train: [18]  [160/400]  eta: 0:02:27  lr: 0.000008  loss: 2.3677 (2.3577)  grad: 0.2108 (0.2185)  time: 0.6126  data: 0.0036  max mem: 57344
+train: [18]  [180/400]  eta: 0:02:15  lr: 0.000008  loss: 2.3805 (2.3608)  grad: 0.2148 (0.2191)  time: 0.6128  data: 0.0036  max mem: 57344
+train: [18]  [200/400]  eta: 0:02:03  lr: 0.000007  loss: 2.3694 (2.3575)  grad: 0.2139 (0.2189)  time: 0.6118  data: 0.0036  max mem: 57344
+train: [18]  [220/400]  eta: 0:01:50  lr: 0.000007  loss: 2.3828 (2.3616)  grad: 0.2103 (0.2193)  time: 0.6129  data: 0.0037  max mem: 57344
+train: [18]  [240/400]  eta: 0:01:38  lr: 0.000006  loss: 2.3895 (2.3644)  grad: 0.2201 (0.2196)  time: 0.6132  data: 0.0036  max mem: 57344
+train: [18]  [260/400]  eta: 0:01:26  lr: 0.000006  loss: 2.3523 (2.3621)  grad: 0.2204 (0.2201)  time: 0.6128  data: 0.0037  max mem: 57344
+train: [18]  [280/400]  eta: 0:01:13  lr: 0.000006  loss: 2.3360 (2.3622)  grad: 0.2160 (0.2197)  time: 0.6124  data: 0.0036  max mem: 57344
+train: [18]  [300/400]  eta: 0:01:01  lr: 0.000005  loss: 2.3492 (2.3637)  grad: 0.2163 (0.2202)  time: 0.6126  data: 0.0036  max mem: 57344
+train: [18]  [320/400]  eta: 0:00:49  lr: 0.000005  loss: 2.3563 (2.3627)  grad: 0.2249 (0.2207)  time: 0.6129  data: 0.0036  max mem: 57344
+train: [18]  [340/400]  eta: 0:00:36  lr: 0.000004  loss: 2.3370 (2.3623)  grad: 0.2220 (0.2207)  time: 0.6132  data: 0.0036  max mem: 57344
+train: [18]  [360/400]  eta: 0:00:24  lr: 0.000004  loss: 2.3541 (2.3636)  grad: 0.2197 (0.2204)  time: 0.6132  data: 0.0037  max mem: 57344
+train: [18]  [380/400]  eta: 0:00:12  lr: 0.000004  loss: 2.3805 (2.3659)  grad: 0.2215 (0.2205)  time: 0.6129  data: 0.0037  max mem: 57344
+train: [18]  [399/400]  eta: 0:00:00  lr: 0.000003  loss: 2.3670 (2.3679)  grad: 0.2248 (0.2209)  time: 0.6118  data: 0.0036  max mem: 57344
+train: [18] Total time: 0:04:05 (0.6143 s / it)
+train: [18] Summary: lr: 0.000003  loss: 2.3670 (2.3679)  grad: 0.2248 (0.2209)
+eval (validation): [18]  [ 0/85]  eta: 0:01:21    time: 0.9628  data: 0.6046  max mem: 57344
+eval (validation): [18]  [20/85]  eta: 0:00:25    time: 0.3682  data: 0.0028  max mem: 57344
+eval (validation): [18]  [40/85]  eta: 0:00:17    time: 0.3687  data: 0.0035  max mem: 57344
+eval (validation): [18]  [60/85]  eta: 0:00:09    time: 0.3695  data: 0.0035  max mem: 57344
+eval (validation): [18]  [80/85]  eta: 0:00:01    time: 0.3700  data: 0.0036  max mem: 57344
+eval (validation): [18]  [84/85]  eta: 0:00:00    time: 0.3639  data: 0.0035  max mem: 57344
+eval (validation): [18] Total time: 0:00:31 (0.3757 s / it)
+cv: [18]  best hparam: (2.7, 1.0) (030) ('030_lr2.7e+00_wd1.0e+00')  loss: 2.386  acc: 0.295  f1: 0.236
+saving checkpoint experiments/schaefer1000/output/schaefer1000/schaefer1000_lr3e-4_1/eval_v2/nsd_cococlip__patch__attn/checkpoint-last.pth
+train: [19]  [  0/400]  eta: 0:07:59  lr: nan  time: 1.1977  data: 0.5991  max mem: 57344
+train: [19]  [ 20/400]  eta: 0:04:03  lr: 0.000003  loss: 2.3653 (2.3847)  grad: 0.2181 (0.2188)  time: 0.6117  data: 0.0029  max mem: 57344
+train: [19]  [ 40/400]  eta: 0:03:45  lr: 0.000003  loss: 2.3653 (2.3953)  grad: 0.2187 (0.2200)  time: 0.6126  data: 0.0037  max mem: 57344
+train: [19]  [ 60/400]  eta: 0:03:31  lr: 0.000002  loss: 2.3593 (2.3879)  grad: 0.2166 (0.2193)  time: 0.6126  data: 0.0036  max mem: 57344
+train: [19]  [ 80/400]  eta: 0:03:18  lr: 0.000002  loss: 2.3593 (2.3838)  grad: 0.2192 (0.2216)  time: 0.6122  data: 0.0036  max mem: 57344
+train: [19]  [100/400]  eta: 0:03:05  lr: 0.000002  loss: 2.4004 (2.3905)  grad: 0.2205 (0.2208)  time: 0.6126  data: 0.0035  max mem: 57344
+train: [19]  [120/400]  eta: 0:02:52  lr: 0.000002  loss: 2.3880 (2.3837)  grad: 0.2103 (0.2188)  time: 0.6124  data: 0.0036  max mem: 57344
+train: [19]  [140/400]  eta: 0:02:40  lr: 0.000001  loss: 2.3838 (2.3819)  grad: 0.2107 (0.2187)  time: 0.6131  data: 0.0036  max mem: 57344
+train: [19]  [160/400]  eta: 0:02:27  lr: 0.000001  loss: 2.3562 (2.3826)  grad: 0.2212 (0.2187)  time: 0.6129  data: 0.0035  max mem: 57344
+train: [19]  [180/400]  eta: 0:02:15  lr: 0.000001  loss: 2.3785 (2.3800)  grad: 0.2166 (0.2182)  time: 0.6135  data: 0.0036  max mem: 57344
+train: [19]  [200/400]  eta: 0:02:03  lr: 0.000001  loss: 2.3830 (2.3813)  grad: 0.2197 (0.2185)  time: 0.6128  data: 0.0036  max mem: 57344
+train: [19]  [220/400]  eta: 0:01:50  lr: 0.000001  loss: 2.3798 (2.3791)  grad: 0.2194 (0.2185)  time: 0.6126  data: 0.0035  max mem: 57344
+train: [19]  [240/400]  eta: 0:01:38  lr: 0.000001  loss: 2.3470 (2.3771)  grad: 0.2120 (0.2180)  time: 0.6130  data: 0.0036  max mem: 57344
+train: [19]  [260/400]  eta: 0:01:26  lr: 0.000000  loss: 2.3208 (2.3759)  grad: 0.2116 (0.2176)  time: 0.6126  data: 0.0036  max mem: 57344
+train: [19]  [280/400]  eta: 0:01:13  lr: 0.000000  loss: 2.3960 (2.3795)  grad: 0.2163 (0.2176)  time: 0.6133  data: 0.0037  max mem: 57344
+train: [19]  [300/400]  eta: 0:01:01  lr: 0.000000  loss: 2.3795 (2.3804)  grad: 0.2137 (0.2171)  time: 0.6127  data: 0.0036  max mem: 57344
+train: [19]  [320/400]  eta: 0:00:49  lr: 0.000000  loss: 2.3632 (2.3785)  grad: 0.2139 (0.2174)  time: 0.6126  data: 0.0037  max mem: 57344
+train: [19]  [340/400]  eta: 0:00:36  lr: 0.000000  loss: 2.3420 (2.3778)  grad: 0.2145 (0.2171)  time: 0.6131  data: 0.0036  max mem: 57344
+train: [19]  [360/400]  eta: 0:00:24  lr: 0.000000  loss: 2.3863 (2.3799)  grad: 0.2158 (0.2174)  time: 0.6125  data: 0.0036  max mem: 57344
+train: [19]  [380/400]  eta: 0:00:12  lr: 0.000000  loss: 2.3904 (2.3801)  grad: 0.2178 (0.2173)  time: 0.6125  data: 0.0037  max mem: 57344
+train: [19]  [399/400]  eta: 0:00:00  lr: 0.000000  loss: 2.3811 (2.3811)  grad: 0.2106 (0.2170)  time: 0.6119  data: 0.0035  max mem: 57344
+train: [19] Total time: 0:04:05 (0.6144 s / it)
+train: [19] Summary: lr: 0.000000  loss: 2.3811 (2.3811)  grad: 0.2106 (0.2170)
+eval (validation): [19]  [ 0/85]  eta: 0:01:13    time: 0.8623  data: 0.5023  max mem: 57344
+eval (validation): [19]  [20/85]  eta: 0:00:25    time: 0.3685  data: 0.0032  max mem: 57344
+eval (validation): [19]  [40/85]  eta: 0:00:17    time: 0.3688  data: 0.0033  max mem: 57344
+eval (validation): [19]  [60/85]  eta: 0:00:09    time: 0.3691  data: 0.0034  max mem: 57344
+eval (validation): [19]  [80/85]  eta: 0:00:01    time: 0.3693  data: 0.0035  max mem: 57344
+eval (validation): [19]  [84/85]  eta: 0:00:00    time: 0.3630  data: 0.0035  max mem: 57344
+eval (validation): [19] Total time: 0:00:31 (0.3744 s / it)
+cv: [19]  best hparam: (2.7, 1.0) (030) ('030_lr2.7e+00_wd1.0e+00')  loss: 2.385  acc: 0.295  f1: 0.236
+saving checkpoint experiments/schaefer1000/output/schaefer1000/schaefer1000_lr3e-4_1/eval_v2/nsd_cococlip__patch__attn/checkpoint-last.pth
+evaluating last checkpoint: experiments/schaefer1000/output/schaefer1000/schaefer1000_lr3e-4_1/eval_v2/nsd_cococlip__patch__attn/checkpoint-last.pth
+eval model info:
+{"score": 0.29457364341085274, "hparam": [2.7, 1.0], "hparam_id": 30, "epoch": 19, "is_best": false, "best_score": 0.2978959025470653}
+eval (train): [20]  [  0/509]  eta: 0:08:21    time: 0.9843  data: 0.6270  max mem: 57344
+eval (train): [20]  [ 20/509]  eta: 0:03:14    time: 0.3674  data: 0.0027  max mem: 57344
+eval (train): [20]  [ 40/509]  eta: 0:02:59    time: 0.3681  data: 0.0035  max mem: 57344
+eval (train): [20]  [ 60/509]  eta: 0:02:49    time: 0.3684  data: 0.0035  max mem: 57344
+eval (train): [20]  [ 80/509]  eta: 0:02:41    time: 0.3682  data: 0.0035  max mem: 57344
+eval (train): [20]  [100/509]  eta: 0:02:33    time: 0.3688  data: 0.0035  max mem: 57344
+eval (train): [20]  [120/509]  eta: 0:02:25    time: 0.3689  data: 0.0036  max mem: 57344
+eval (train): [20]  [140/509]  eta: 0:02:17    time: 0.3688  data: 0.0035  max mem: 57344
+eval (train): [20]  [160/509]  eta: 0:02:09    time: 0.3685  data: 0.0033  max mem: 57344
+eval (train): [20]  [180/509]  eta: 0:02:02    time: 0.3682  data: 0.0035  max mem: 57344
+eval (train): [20]  [200/509]  eta: 0:01:54    time: 0.3680  data: 0.0035  max mem: 57344
+eval (train): [20]  [220/509]  eta: 0:01:47    time: 0.3685  data: 0.0035  max mem: 57344
+eval (train): [20]  [240/509]  eta: 0:01:39    time: 0.3679  data: 0.0034  max mem: 57344
+eval (train): [20]  [260/509]  eta: 0:01:32    time: 0.3679  data: 0.0033  max mem: 57344
+eval (train): [20]  [280/509]  eta: 0:01:24    time: 0.3682  data: 0.0035  max mem: 57344
+eval (train): [20]  [300/509]  eta: 0:01:17    time: 0.3682  data: 0.0035  max mem: 57344
+eval (train): [20]  [320/509]  eta: 0:01:09    time: 0.3678  data: 0.0035  max mem: 57344
+eval (train): [20]  [340/509]  eta: 0:01:02    time: 0.3678  data: 0.0034  max mem: 57344
+eval (train): [20]  [360/509]  eta: 0:00:55    time: 0.3678  data: 0.0033  max mem: 57344
+eval (train): [20]  [380/509]  eta: 0:00:47    time: 0.3685  data: 0.0034  max mem: 57344
+eval (train): [20]  [400/509]  eta: 0:00:40    time: 0.3688  data: 0.0034  max mem: 57344
+eval (train): [20]  [420/509]  eta: 0:00:32    time: 0.3682  data: 0.0036  max mem: 57344
+eval (train): [20]  [440/509]  eta: 0:00:25    time: 0.3681  data: 0.0034  max mem: 57344
+eval (train): [20]  [460/509]  eta: 0:00:18    time: 0.3680  data: 0.0033  max mem: 57344
+eval (train): [20]  [480/509]  eta: 0:00:10    time: 0.3691  data: 0.0035  max mem: 57344
+eval (train): [20]  [500/509]  eta: 0:00:03    time: 0.3694  data: 0.0035  max mem: 57344
+eval (train): [20]  [508/509]  eta: 0:00:00    time: 0.3594  data: 0.0035  max mem: 57344
+eval (train): [20] Total time: 0:03:08 (0.3695 s / it)
+eval (validation): [20]  [ 0/85]  eta: 0:01:20    time: 0.9437  data: 0.5859  max mem: 57344
+eval (validation): [20]  [20/85]  eta: 0:00:25    time: 0.3672  data: 0.0031  max mem: 57344
+eval (validation): [20]  [40/85]  eta: 0:00:17    time: 0.3679  data: 0.0036  max mem: 57344
+eval (validation): [20]  [60/85]  eta: 0:00:09    time: 0.3692  data: 0.0035  max mem: 57344
+eval (validation): [20]  [80/85]  eta: 0:00:01    time: 0.3694  data: 0.0033  max mem: 57344
+eval (validation): [20]  [84/85]  eta: 0:00:00    time: 0.3632  data: 0.0034  max mem: 57344
+eval (validation): [20] Total time: 0:00:31 (0.3750 s / it)
+eval (test): [20]  [ 0/85]  eta: 0:01:25    time: 1.0087  data: 0.6536  max mem: 57344
+eval (test): [20]  [20/85]  eta: 0:00:25    time: 0.3670  data: 0.0024  max mem: 57344
+eval (test): [20]  [40/85]  eta: 0:00:17    time: 0.3687  data: 0.0035  max mem: 57344
+eval (test): [20]  [60/85]  eta: 0:00:09    time: 0.3688  data: 0.0034  max mem: 57344
+eval (test): [20]  [80/85]  eta: 0:00:01    time: 0.3690  data: 0.0033  max mem: 57344
+eval (test): [20]  [84/85]  eta: 0:00:00    time: 0.3554  data: 0.0033  max mem: 57344
+eval (test): [20] Total time: 0:00:31 (0.3739 s / it)
+eval (testid): [20]  [ 0/82]  eta: 0:01:18    time: 0.9616  data: 0.6027  max mem: 57344
+eval (testid): [20]  [20/82]  eta: 0:00:24    time: 0.3672  data: 0.0028  max mem: 57344
+eval (testid): [20]  [40/82]  eta: 0:00:16    time: 0.3684  data: 0.0036  max mem: 57344
+eval (testid): [20]  [60/82]  eta: 0:00:08    time: 0.3683  data: 0.0035  max mem: 57344
+eval (testid): [20]  [80/82]  eta: 0:00:00    time: 0.3686  data: 0.0035  max mem: 57344
+eval (testid): [20]  [81/82]  eta: 0:00:00    time: 0.3516  data: 0.0035  max mem: 57344
+eval (testid): [20] Total time: 0:00:30 (0.3724 s / it)
+evaluating best checkpoint: experiments/schaefer1000/output/schaefer1000/schaefer1000_lr3e-4_1/eval_v2/nsd_cococlip__patch__attn/checkpoint-best.pth
+eval model info:
+{"score": 0.2978959025470653, "hparam": [2.7, 1.0], "hparam_id": 30, "epoch": 13, "is_best": true, "best_score": 0.2978959025470653}
+eval (train): [20]  [  0/509]  eta: 0:08:19    time: 0.9820  data: 0.6228  max mem: 57344
+eval (train): [20]  [ 20/509]  eta: 0:03:14    time: 0.3676  data: 0.0030  max mem: 57344
+eval (train): [20]  [ 40/509]  eta: 0:02:59    time: 0.3680  data: 0.0035  max mem: 57344
+eval (train): [20]  [ 60/509]  eta: 0:02:49    time: 0.3686  data: 0.0037  max mem: 57344
+eval (train): [20]  [ 80/509]  eta: 0:02:41    time: 0.3679  data: 0.0034  max mem: 57344
+eval (train): [20]  [100/509]  eta: 0:02:33    time: 0.3686  data: 0.0035  max mem: 57344
+eval (train): [20]  [120/509]  eta: 0:02:25    time: 0.3690  data: 0.0034  max mem: 57344
+eval (train): [20]  [140/509]  eta: 0:02:17    time: 0.3686  data: 0.0033  max mem: 57344
+eval (train): [20]  [160/509]  eta: 0:02:09    time: 0.3682  data: 0.0035  max mem: 57344
+eval (train): [20]  [180/509]  eta: 0:02:02    time: 0.3686  data: 0.0034  max mem: 57344
+eval (train): [20]  [200/509]  eta: 0:01:54    time: 0.3686  data: 0.0035  max mem: 57344
+eval (train): [20]  [220/509]  eta: 0:01:47    time: 0.3686  data: 0.0034  max mem: 57344
+eval (train): [20]  [240/509]  eta: 0:01:39    time: 0.3692  data: 0.0035  max mem: 57344
+eval (train): [20]  [260/509]  eta: 0:01:32    time: 0.3686  data: 0.0034  max mem: 57344
+eval (train): [20]  [280/509]  eta: 0:01:24    time: 0.3690  data: 0.0035  max mem: 57344
+eval (train): [20]  [300/509]  eta: 0:01:17    time: 0.3683  data: 0.0035  max mem: 57344
+eval (train): [20]  [320/509]  eta: 0:01:10    time: 0.3688  data: 0.0033  max mem: 57344
+eval (train): [20]  [340/509]  eta: 0:01:02    time: 0.3681  data: 0.0035  max mem: 57344
+eval (train): [20]  [360/509]  eta: 0:00:55    time: 0.3689  data: 0.0034  max mem: 57344
+eval (train): [20]  [380/509]  eta: 0:00:47    time: 0.3686  data: 0.0035  max mem: 57344
+eval (train): [20]  [400/509]  eta: 0:00:40    time: 0.3686  data: 0.0036  max mem: 57344
+eval (train): [20]  [420/509]  eta: 0:00:32    time: 0.3692  data: 0.0037  max mem: 57344
+eval (train): [20]  [440/509]  eta: 0:00:25    time: 0.3684  data: 0.0036  max mem: 57344
+eval (train): [20]  [460/509]  eta: 0:00:18    time: 0.3684  data: 0.0035  max mem: 57344
+eval (train): [20]  [480/509]  eta: 0:00:10    time: 0.3686  data: 0.0036  max mem: 57344
+eval (train): [20]  [500/509]  eta: 0:00:03    time: 0.3686  data: 0.0034  max mem: 57344
+eval (train): [20]  [508/509]  eta: 0:00:00    time: 0.3585  data: 0.0034  max mem: 57344
+eval (train): [20] Total time: 0:03:08 (0.3697 s / it)
+eval (validation): [20]  [ 0/85]  eta: 0:01:20    time: 0.9504  data: 0.5924  max mem: 57344
+eval (validation): [20]  [20/85]  eta: 0:00:25    time: 0.3680  data: 0.0034  max mem: 57344
+eval (validation): [20]  [40/85]  eta: 0:00:17    time: 0.3685  data: 0.0036  max mem: 57344
+eval (validation): [20]  [60/85]  eta: 0:00:09    time: 0.3691  data: 0.0035  max mem: 57344
+eval (validation): [20]  [80/85]  eta: 0:00:01    time: 0.3689  data: 0.0036  max mem: 57344
+eval (validation): [20]  [84/85]  eta: 0:00:00    time: 0.3627  data: 0.0036  max mem: 57344
+eval (validation): [20] Total time: 0:00:31 (0.3753 s / it)
+eval (test): [20]  [ 0/85]  eta: 0:01:21    time: 0.9566  data: 0.6005  max mem: 57344
+eval (test): [20]  [20/85]  eta: 0:00:25    time: 0.3677  data: 0.0030  max mem: 57344
+eval (test): [20]  [40/85]  eta: 0:00:17    time: 0.3685  data: 0.0035  max mem: 57344
+eval (test): [20]  [60/85]  eta: 0:00:09    time: 0.3687  data: 0.0035  max mem: 57344
+eval (test): [20]  [80/85]  eta: 0:00:01    time: 0.3696  data: 0.0034  max mem: 57344
+eval (test): [20]  [84/85]  eta: 0:00:00    time: 0.3554  data: 0.0034  max mem: 57344
+eval (test): [20] Total time: 0:00:31 (0.3734 s / it)
+eval (testid): [20]  [ 0/82]  eta: 0:01:15    time: 0.9240  data: 0.5643  max mem: 57344
+eval (testid): [20]  [20/82]  eta: 0:00:24    time: 0.3682  data: 0.0030  max mem: 57344
+eval (testid): [20]  [40/82]  eta: 0:00:16    time: 0.3682  data: 0.0033  max mem: 57344
+eval (testid): [20]  [60/82]  eta: 0:00:08    time: 0.3685  data: 0.0032  max mem: 57344
+eval (testid): [20]  [80/82]  eta: 0:00:00    time: 0.3692  data: 0.0034  max mem: 57344
+eval (testid): [20]  [81/82]  eta: 0:00:00    time: 0.3521  data: 0.0034  max mem: 57344
+eval (testid): [20] Total time: 0:00:30 (0.3724 s / it)
+eval results:
+
+| model            | repr   | clf   | dataset      | ckpt   |   epoch |      lr |   wd |   hparam_id | hparam     | split      |   loss |     acc |   acc_std |      f1 |    f1_std |
+|:-----------------|:-------|:------|:-------------|:-------|--------:|--------:|-----:|------------:|:-----------|:-----------|-------:|--------:|----------:|--------:|----------:|
+| schaefer1000_mae | patch  | attn  | nsd_cococlip | best   |      13 | 0.00081 | 0.05 |          30 | [2.7, 1.0] | train      | 1.8354 | 0.44046 | 0.0025471 | 0.4062  | 0.0027522 |
+| schaefer1000_mae | patch  | attn  | nsd_cococlip | best   |      13 | 0.00081 | 0.05 |          30 | [2.7, 1.0] | validation | 2.3563 | 0.2979  | 0.0052148 | 0.24269 | 0.0051564 |
+| schaefer1000_mae | patch  | attn  | nsd_cococlip | best   |      13 | 0.00081 | 0.05 |          30 | [2.7, 1.0] | test       | 2.2968 | 0.3102  | 0.0056622 | 0.24868 | 0.0055509 |
+| schaefer1000_mae | patch  | attn  | nsd_cococlip | best   |      13 | 0.00081 | 0.05 |          30 | [2.7, 1.0] | testid     | 2.3031 | 0.30673 | 0.0056088 | 0.27088 | 0.0056972 |
+
+
+done! total time: 1:44:24

schaefer1000/schaefer1000_lr3e-4_1/pretrain/config.yaml

@@ -0,0 +1,102 @@
+name: schaefer1000/schaefer1000_lr3e-4_1/pretrain
+notes: schaefer1000 ablation schaefer1000_lr3e-4_1 (input_space=schaefer1000 base_lr=3e-4
+  seed=5401)
+output_dir: experiments/schaefer1000/output/schaefer1000/schaefer1000_lr3e-4_1/pretrain
+input_space: schaefer1000
+patch_size: 1
+num_frames: 16
+t_patch_size: 4
+mask_ratio: 0.9
+pred_mask_ratio: null
+masking: tube
+masking_kwargs: {}
+mask_patch_size: null
+model: mae_vit_base
+model_kwargs:
+  decoding: attn
+  pos_embed: sep
+  target_norm: null
+  pca_norm_nc: 2
+  t_pred_stride: 2
+  no_decode_pos: true
+  mask_drop_scale: false
+  pred_edge_pad: 0
+  gauss_sigma: null
+  class_token: true
+  reg_tokens: 0
+  no_embed_class: true
+  head_init_scale: 0.0
+  decoder_depth: 4
+  drop_path_rate: 0.0
+datasets:
+  hcp-train:
+    type: wds
+    url: /data/fmri-datasets/pretrain/hcpya-all.${input_space}.wds/hcpya-all-${input_space}-{00000..01799}.tar
+    clipping: random
+    clipping_kwargs:
+      oversample: 4.0
+    shuffle: true
+    buffer_size: 2000
+    samples_per_epoch: 200000
+  hcp-train-subset:
+    type: arrow
+    root: s3://medarc/fmri-datasets/eval/hcpya-clips.${input_space}.arrow/validation
+    split_range:
+    - 0
+    - 2000
+    shuffle: false
+  hcp-val:
+    type: arrow
+    root: s3://medarc/fmri-datasets/eval/hcpya-clips.${input_space}.arrow/test
+    split_range:
+    - 0
+    - 2000
+    shuffle: false
+train_dataset: hcp-train
+eval_datasets:
+- hcp-train-subset
+- hcp-val
+val_dataset: null
+clip_vmax: 3.0
+normalize: frame
+tr_scale: null
+crop_scale: null
+crop_aspect: null
+gray_jitter: null
+num_workers: 16
+epochs: 100
+batch_size: 32
+accum_iter: 1
+base_lr: 0.0003
+min_lr: 0.0
+warmup_epochs: 5
+weight_decay: 0.05
+betas:
+- 0.9
+- 0.95
+clip_grad: 1.0
+amp: true
+amp_dtype: float16
+ckpt: null
+resume: true
+auto_resume: true
+start_epoch: 0
+max_checkpoints: 0
+checkpoint_period: null
+plot_period: 5
+device: cuda
+presend_cuda: false
+seed: 5401
+debug: false
+wandb: true
+wandb_entity: null
+wandb_project: fMRI-foundation-model
+rank: 0
+world_size: 1
+gpu: 0
+distributed: true
+dist_backend: nccl
+in_chans: 1
+img_size:
+- 1000
+- 1

schaefer1000/schaefer1000_lr3e-4_1/pretrain/log.json

@@ -0,0 +1,100 @@
+{"epoch": 0, "train/lr": 3.7507200230407366e-06, "train/grad": 1.6712108595356663, "train/loss": 0.9859507956981659, "eval/hcp-train-subset/loss": 0.967505342537357, "eval/hcp-val/loss": 0.965046702854095}
+{"epoch": 1, "train/lr": 1.125096003072098e-05, "train/grad": 3.3951989526421964, "train/loss": 0.9344591883277893, "eval/hcp-train-subset/loss": 0.9023149474974601, "eval/hcp-val/loss": 0.900323853377373}
+{"epoch": 2, "train/lr": 1.875120003840123e-05, "train/grad": 2.430334189325776, "train/loss": 0.8663726669883728, "eval/hcp-train-subset/loss": 0.8485412588042598, "eval/hcp-val/loss": 0.8441499875437829}
+{"epoch": 3, "train/lr": 2.625144004608147e-05, "train/grad": 1.5350360517770472, "train/loss": 0.822853887758255, "eval/hcp-train-subset/loss": 0.7901489955763663, "eval/hcp-val/loss": 0.7813271245648784}
+{"epoch": 4, "train/lr": 3.375167986175557e-05, "train/grad": 1.271350692146757, "train/loss": 0.7723959146022796, "eval/hcp-train-subset/loss": 0.7677068748781758, "eval/hcp-val/loss": 0.7587820031950551}
+{"epoch": 5, "train/lr": 3.749658191365373e-05, "train/grad": 0.9016175542485338, "train/loss": 0.754934159116745, "eval/hcp-train-subset/loss": 0.7548967724846255, "eval/hcp-val/loss": 0.7476687315971621}
+{"epoch": 6, "train/lr": 3.74760811613377e-05, "train/grad": 0.6958302143202777, "train/loss": 0.7451256625556946, "eval/hcp-train-subset/loss": 0.7485817891936148, "eval/hcp-val/loss": 0.7380533324133965}
+{"epoch": 7, "train/lr": 3.743510371420547e-05, "train/grad": 0.5815669165976967, "train/loss": 0.7403617179679871, "eval/hcp-train-subset/loss": 0.7434995049430478, "eval/hcp-val/loss": 0.7347037513409892}
+{"epoch": 8, "train/lr": 3.73736943804934e-05, "train/grad": 0.5083776054066316, "train/loss": 0.734418772315979, "eval/hcp-train-subset/loss": 0.7376639458440966, "eval/hcp-val/loss": 0.7291759175638999}
+{"epoch": 9, "train/lr": 3.7291920310406644e-05, "train/grad": 0.4657674653159536, "train/loss": 0.728910969285965, "eval/hcp-train-subset/loss": 0.7343605074190325, "eval/hcp-val/loss": 0.7260435137056536}
+{"epoch": 10, "train/lr": 3.718987092269037e-05, "train/grad": 0.43497297624807385, "train/loss": 0.7260179727077484, "eval/hcp-train-subset/loss": 0.7319134935255973, "eval/hcp-val/loss": 0.7227755092805431}
+{"epoch": 11, "train/lr": 3.706765780685143e-05, "train/grad": 0.4128942857251199, "train/loss": 0.7226078836536407, "eval/hcp-train-subset/loss": 0.7290946831626277, "eval/hcp-val/loss": 0.7207083279086698}
+{"epoch": 12, "train/lr": 3.692541460113792e-05, "train/grad": 0.39792067562463246, "train/loss": 0.7188967194652557, "eval/hcp-train-subset/loss": 0.7264460632877965, "eval/hcp-val/loss": 0.7191831527217742}
+{"epoch": 13, "train/lr": 3.6763296846406675e-05, "train/grad": 0.3831865847902678, "train/loss": 0.7165294539928436, "eval/hcp-train-subset/loss": 0.7260655972265428, "eval/hcp-val/loss": 0.7182057240316945}
+{"epoch": 14, "train/lr": 3.658148181604263e-05, "train/grad": 0.38097352542761936, "train/loss": 0.7144903720378876, "eval/hcp-train-subset/loss": 0.7238107698578988, "eval/hcp-val/loss": 0.7152259340209346}
+{"epoch": 15, "train/lr": 3.6380168322111824e-05, "train/grad": 0.3685756610673352, "train/loss": 0.7134313011550903, "eval/hcp-train-subset/loss": 0.7229695252833828, "eval/hcp-val/loss": 0.7159229026686761}
+{"epoch": 16, "train/lr": 3.615957649796421e-05, "train/grad": 0.3595638900842593, "train/loss": 0.7111813261222839, "eval/hcp-train-subset/loss": 0.7215365507910328, "eval/hcp-val/loss": 0.7120943540526975}
+{"epoch": 17, "train/lr": 3.591994755752113e-05, "train/grad": 0.3557017998603204, "train/loss": 0.7109086339187622, "eval/hcp-train-subset/loss": 0.7180883730611494, "eval/hcp-val/loss": 0.7106600740263539}
+{"epoch": 18, "train/lr": 3.5661543531510486e-05, "train/grad": 0.3506869831542425, "train/loss": 0.707928588848114, "eval/hcp-train-subset/loss": 0.7183591300441373, "eval/hcp-val/loss": 0.7100169754797413}
+{"epoch": 19, "train/lr": 3.538464698094067e-05, "train/grad": 0.34085002410587395, "train/loss": 0.708698856010437, "eval/hcp-train-subset/loss": 0.7173036356126109, "eval/hcp-val/loss": 0.7096435437279363}
+{"epoch": 20, "train/lr": 3.508956068812486e-05, "train/grad": 0.3489449820304577, "train/loss": 0.7011169036960602, "eval/hcp-train-subset/loss": 0.7162846615237575, "eval/hcp-val/loss": 0.7083295343383667}
+{"epoch": 21, "train/lr": 3.4776607325591504e-05, "train/grad": 0.34312217887380664, "train/loss": 0.7021659459590912, "eval/hcp-train-subset/loss": 0.7156490241327593, "eval/hcp-val/loss": 0.7084641379694785}
+{"epoch": 22, "train/lr": 3.4446129103247903e-05, "train/grad": 0.3450461020560499, "train/loss": 0.6991195864391327, "eval/hcp-train-subset/loss": 0.7139196492010548, "eval/hcp-val/loss": 0.7078986427476329}
+{"epoch": 23, "train/lr": 3.4098487394178203e-05, "train/grad": 0.34011013350029534, "train/loss": 0.7005356298351287, "eval/hcp-train-subset/loss": 0.7160567564349021, "eval/hcp-val/loss": 0.7069014849201325}
+{"epoch": 24, "train/lr": 3.37340623394871e-05, "train/grad": 0.34082504409693787, "train/loss": 0.6973995522403718, "eval/hcp-train-subset/loss": 0.7165951036637829, "eval/hcp-val/loss": 0.7085161209106445}
+{"epoch": 25, "train/lr": 3.335325243262167e-05, "train/grad": 0.33862331307075566, "train/loss": 0.6978817233181, "eval/hcp-train-subset/loss": 0.7129770998031862, "eval/hcp-val/loss": 0.7058122177277842}
+{"epoch": 26, "train/lr": 3.295647408362393e-05, "train/grad": 0.3467655432773854, "train/loss": 0.6956600578784943, "eval/hcp-train-subset/loss": 0.7120144646013936, "eval/hcp-val/loss": 0.7037961838706848}
+{"epoch": 27, "train/lr": 3.254416116379294e-05, "train/grad": 0.3470933223407097, "train/loss": 0.6961856775093078, "eval/hcp-train-subset/loss": 0.7115420785642439, "eval/hcp-val/loss": 0.706233222638407}
+{"epoch": 28, "train/lr": 3.211676453125335e-05, "train/grad": 0.3460057961195588, "train/loss": 0.6951398908233642, "eval/hcp-train-subset/loss": 0.711975775418743, "eval/hcp-val/loss": 0.7059361848138994}
+{"epoch": 29, "train/lr": 3.1674751537947714e-05, "train/grad": 0.34924345383963545, "train/loss": 0.6924728297519683, "eval/hcp-train-subset/loss": 0.7106139698336201, "eval/hcp-val/loss": 0.7039365768432617}
+{"epoch": 30, "train/lr": 3.1218605518594614e-05, "train/grad": 0.35291437485780225, "train/loss": 0.6924978620052338, "eval/hcp-train-subset/loss": 0.7119266784960224, "eval/hcp-val/loss": 0.7044628656679585}
+{"epoch": 31, "train/lr": 3.074882526216893e-05, "train/grad": 0.3541540915329246, "train/loss": 0.6916870978164673, "eval/hcp-train-subset/loss": 0.7110956522726244, "eval/hcp-val/loss": 0.7031774030577752}
+{"epoch": 32, "train/lr": 3.0265924466484304e-05, "train/grad": 0.34942363362150497, "train/loss": 0.6905806545639038, "eval/hcp-train-subset/loss": 0.7081248606404951, "eval/hcp-val/loss": 0.703105362192277}
+{"epoch": 33, "train/lr": 2.977043117647054e-05, "train/grad": 0.351709049995926, "train/loss": 0.6881411712741852, "eval/hcp-train-subset/loss": 0.7080096612053532, "eval/hcp-val/loss": 0.7012732932644505}
+{"epoch": 34, "train/lr": 2.9262887206765643e-05, "train/grad": 0.3520911494927424, "train/loss": 0.68959153840065, "eval/hcp-train-subset/loss": 0.7089110064891077, "eval/hcp-val/loss": 0.7034689918641122}
+{"epoch": 35, "train/lr": 2.8743847549248834e-05, "train/grad": 0.34930833157051155, "train/loss": 0.6878345485687256, "eval/hcp-train-subset/loss": 0.7080285876028, "eval/hcp-val/loss": 0.7023586688503143}
+{"epoch": 36, "train/lr": 2.821387976616516e-05, "train/grad": 0.35714524405917036, "train/loss": 0.6867969785785675, "eval/hcp-train-subset/loss": 0.7082647844668357, "eval/hcp-val/loss": 0.700894522090112}
+{"epoch": 37, "train/lr": 2.7673563369504416e-05, "train/grad": 0.3557986267971309, "train/loss": 0.687398661403656, "eval/hcp-train-subset/loss": 0.7071594595909119, "eval/hcp-val/loss": 0.702428650471472}
+{"epoch": 38, "train/lr": 2.7123489187313515e-05, "train/grad": 0.3548295525254031, "train/loss": 0.687827016658783, "eval/hcp-train-subset/loss": 0.7069384282635104, "eval/hcp-val/loss": 0.7011815088410531}
+{"epoch": 39, "train/lr": 2.6564258717634662e-05, "train/grad": 0.3608307055318096, "train/loss": 0.6866573888301849, "eval/hcp-train-subset/loss": 0.7050968283607114, "eval/hcp-val/loss": 0.6991308485308001}
+{"epoch": 40, "train/lr": 2.5996483470775956e-05, "train/grad": 0.3748945764480485, "train/loss": 0.6831279589653015, "eval/hcp-train-subset/loss": 0.7063164239929568, "eval/hcp-val/loss": 0.7011607164336789}
+{"epoch": 41, "train/lr": 2.5420784300633598e-05, "train/grad": 0.37034374556103306, "train/loss": 0.6834621568870545, "eval/hcp-train-subset/loss": 0.706137711963346, "eval/hcp-val/loss": 0.6995143275107106}
+{"epoch": 42, "train/lr": 2.4837790725798037e-05, "train/grad": 0.3657556139260688, "train/loss": 0.6837784346866608, "eval/hcp-train-subset/loss": 0.7039799094200134, "eval/hcp-val/loss": 0.6990455254431693}
+{"epoch": 43, "train/lr": 2.4248140241183872e-05, "train/grad": 0.37313344630982104, "train/loss": 0.680740471315384, "eval/hcp-train-subset/loss": 0.7040019323748927, "eval/hcp-val/loss": 0.6989667175277587}
+{"epoch": 44, "train/lr": 2.36524776209388e-05, "train/grad": 0.3743950690658489, "train/loss": 0.683285284833908, "eval/hcp-train-subset/loss": 0.7021695500419985, "eval/hcp-val/loss": 0.6987485837551856}
+{"epoch": 45, "train/lr": 2.305145421339275e-05, "train/grad": 0.3828235410758547, "train/loss": 0.6799988960266113, "eval/hcp-train-subset/loss": 0.7024719128685613, "eval/hcp-val/loss": 0.698577247319683}
+{"epoch": 46, "train/lr": 2.2445727228817648e-05, "train/grad": 0.3796838558643784, "train/loss": 0.6799856720542907, "eval/hcp-train-subset/loss": 0.7053721402921984, "eval/hcp-val/loss": 0.6994911874494245}
+{"epoch": 47, "train/lr": 2.1835959020777867e-05, "train/grad": 0.3845257077319003, "train/loss": 0.6809644718647003, "eval/hcp-train-subset/loss": 0.7020642094073757, "eval/hcp-val/loss": 0.6988484407624891}
+{"epoch": 48, "train/lr": 2.1222816361856524e-05, "train/grad": 0.3859225776823536, "train/loss": 0.6816038922595978, "eval/hcp-train-subset/loss": 0.7015534120221292, "eval/hcp-val/loss": 0.6970575999829077}
+{"epoch": 49, "train/lr": 2.0606969714549076e-05, "train/grad": 0.385937238626429, "train/loss": 0.6800386638927459, "eval/hcp-train-subset/loss": 0.7021820141423133, "eval/hcp-val/loss": 0.6967212086723696}
+{"epoch": 50, "train/lr": 1.9989092498122458e-05, "train/grad": 0.38696702191290405, "train/loss": 0.6796728707313537, "eval/hcp-train-subset/loss": 0.7023341434617196, "eval/hcp-val/loss": 0.6966614204068338}
+{"epoch": 51, "train/lr": 1.936986035224125e-05, "train/grad": 0.39215091531593854, "train/loss": 0.6805675098514556, "eval/hcp-train-subset/loss": 0.7016156046621261, "eval/hcp-val/loss": 0.6980480359446618}
+{"epoch": 52, "train/lr": 1.8749950398165728e-05, "train/grad": 0.40186738691241425, "train/loss": 0.6756430479335785, "eval/hcp-train-subset/loss": 0.698739504621875, "eval/hcp-val/loss": 0.6954529631522394}
+{"epoch": 53, "train/lr": 1.8130040498328958e-05, "train/grad": 0.3945709976679385, "train/loss": 0.6765738704109192, "eval/hcp-train-subset/loss": 0.6990093175441988, "eval/hcp-val/loss": 0.6967779715215007}
+{"epoch": 54, "train/lr": 1.751080851510517e-05, "train/grad": 0.3938150447596547, "train/loss": 0.6783007579231263, "eval/hcp-train-subset/loss": 0.6985308950947177, "eval/hcp-val/loss": 0.695638116328947}
+{"epoch": 55, "train/lr": 1.689293156957646e-05, "train/grad": 0.40489902639887093, "train/loss": 0.6763680372619629, "eval/hcp-train-subset/loss": 0.6983479165261791, "eval/hcp-val/loss": 0.6957791514935032}
+{"epoch": 56, "train/lr": 1.627708530111125e-05, "train/grad": 0.41571209463898384, "train/loss": 0.6736670857715606, "eval/hcp-train-subset/loss": 0.6971147954463959, "eval/hcp-val/loss": 0.6950231915520083}
+{"epoch": 57, "train/lr": 1.566394312856223e-05, "train/grad": 0.4119336337557178, "train/loss": 0.6744328346252442, "eval/hcp-train-subset/loss": 0.697599438890334, "eval/hcp-val/loss": 0.6955911094142545}
+{"epoch": 58, "train/lr": 1.5054175513893184e-05, "train/grad": 0.41133746763550594, "train/loss": 0.6741365810012817, "eval/hcp-train-subset/loss": 0.6973159101701552, "eval/hcp-val/loss": 0.6953408900768526}
+{"epoch": 59, "train/lr": 1.444844922903809e-05, "train/grad": 0.4213143327820774, "train/loss": 0.6755625086021423, "eval/hcp-train-subset/loss": 0.697580728800066, "eval/hcp-val/loss": 0.6942245008484009}
+{"epoch": 60, "train/lr": 1.3847426626796371e-05, "train/grad": 0.42639006903997057, "train/loss": 0.6741649422264099, "eval/hcp-train-subset/loss": 0.6943357086950733, "eval/hcp-val/loss": 0.695525832714573}
+{"epoch": 61, "train/lr": 1.3251764916559323e-05, "train/grad": 0.4404002724909221, "train/loss": 0.6728203796482086, "eval/hcp-train-subset/loss": 0.6979966163635254, "eval/hcp-val/loss": 0.6954559299253649}
+{"epoch": 62, "train/lr": 1.2662115445661782e-05, "train/grad": 0.4302821857899456, "train/loss": 0.674036775598526, "eval/hcp-train-subset/loss": 0.6954680431273675, "eval/hcp-val/loss": 0.6953855877922427}
+{"epoch": 63, "train/lr": 1.2079122987143025e-05, "train/grad": 0.43268850629865435, "train/loss": 0.672227874917984, "eval/hcp-train-subset/loss": 0.694167229437059, "eval/hcp-val/loss": 0.6939967591916362}
+{"epoch": 64, "train/lr": 1.1503425034696866e-05, "train/grad": 0.43670809213611744, "train/loss": 0.6697749724292755, "eval/hcp-train-subset/loss": 0.6936446187957641, "eval/hcp-val/loss": 0.693389876234916}
+{"epoch": 65, "train/lr": 1.0935651105582284e-05, "train/grad": 0.43763203416663615, "train/loss": 0.6741515974617004, "eval/hcp-train-subset/loss": 0.6928033271143513, "eval/hcp-val/loss": 0.6922241141719203}
+{"epoch": 66, "train/lr": 1.0376422052254546e-05, "train/grad": 0.44050579190216044, "train/loss": 0.6751755339908599, "eval/hcp-train-subset/loss": 0.691379472132652, "eval/hcp-val/loss": 0.6929845665731738}
+{"epoch": 67, "train/lr": 9.82634938347307e-06, "train/grad": 0.4530042156288485, "train/loss": 0.6691429278278351, "eval/hcp-train-subset/loss": 0.6916558156090398, "eval/hcp-val/loss": 0.6919365996314634}
+{"epoch": 68, "train/lr": 9.286034595625e-06, "train/grad": 0.44095532613762745, "train/loss": 0.6726198818302155, "eval/hcp-train-subset/loss": 0.6924581681528399, "eval/hcp-val/loss": 0.6933786974799249}
+{"epoch": 69, "train/lr": 8.756068514998114e-06, "train/grad": 0.46421584733326593, "train/loss": 0.670915218076706, "eval/hcp-train-subset/loss": 0.6914336027637604, "eval/hcp-val/loss": 0.692562710854315}
+{"epoch": 70, "train/lr": 8.237030651720665e-06, "train/grad": 0.4695434965116953, "train/loss": 0.6693005781173706, "eval/hcp-train-subset/loss": 0.6910701278717287, "eval/hcp-val/loss": 0.6915516680286776}
+{"epoch": 71, "train/lr": 7.729488566075681e-06, "train/grad": 0.4761474327674367, "train/loss": 0.6692303531837464, "eval/hcp-train-subset/loss": 0.6914683195852465, "eval/hcp-val/loss": 0.691479048421306}
+{"epoch": 72, "train/lr": 7.233997247882178e-06, "train/grad": 0.4658173286468292, "train/loss": 0.6725237023448944, "eval/hcp-train-subset/loss": 0.6901515933775133, "eval/hcp-val/loss": 0.6912331225410584}
+{"epoch": 73, "train/lr": 6.751098509621876e-06, "train/grad": 0.47341332767783606, "train/loss": 0.6718072845554351, "eval/hcp-train-subset/loss": 0.6885353480615923, "eval/hcp-val/loss": 0.6907766797850209}
+{"epoch": 74, "train/lr": 6.281320393975191e-06, "train/grad": 0.4766692436835639, "train/loss": 0.6708532173538208, "eval/hcp-train-subset/loss": 0.6888128990127195, "eval/hcp-val/loss": 0.6905811332887218}
+{"epoch": 75, "train/lr": 5.825176596414363e-06, "train/grad": 0.49374446946467443, "train/loss": 0.6726134183216095, "eval/hcp-train-subset/loss": 0.687998459223778, "eval/hcp-val/loss": 0.6895569553298335}
+{"epoch": 76, "train/lr": 5.3831659034850535e-06, "train/grad": 0.47400842301270874, "train/loss": 0.6714100589084625, "eval/hcp-train-subset/loss": 0.6871504879766895, "eval/hcp-val/loss": 0.690339712365981}
+{"epoch": 77, "train/lr": 4.955771647390373e-06, "train/grad": 0.482237675431138, "train/loss": 0.6739842251586914, "eval/hcp-train-subset/loss": 0.6871272575470709, "eval/hcp-val/loss": 0.6893601580973594}
+{"epoch": 78, "train/lr": 4.543461177474475e-06, "train/grad": 0.496043531016641, "train/loss": 0.671711159620285, "eval/hcp-train-subset/loss": 0.6880723334127857, "eval/hcp-val/loss": 0.689532229015904}
+{"epoch": 79, "train/lr": 4.146685349182691e-06, "train/grad": 0.4915133484038273, "train/loss": 0.6745123175907135, "eval/hcp-train-subset/loss": 0.6865087699505591, "eval/hcp-val/loss": 0.6892707434392744}
+{"epoch": 80, "train/lr": 3.7658780310578768e-06, "train/grad": 0.5062498424947023, "train/loss": 0.669812826461792, "eval/hcp-train-subset/loss": 0.6869195209395501, "eval/hcp-val/loss": 0.6884125951797732}
+{"epoch": 81, "train/lr": 3.401455630311446e-06, "train/grad": 0.5094274469119374, "train/loss": 0.6700164348506927, "eval/hcp-train-subset/loss": 0.6852873775266832, "eval/hcp-val/loss": 0.6888171905471433}
+{"epoch": 82, "train/lr": 3.0538166374881993e-06, "train/grad": 0.5003155628375784, "train/loss": 0.673251689825058, "eval/hcp-train-subset/loss": 0.6853851406804977, "eval/hcp-val/loss": 0.6882513155860286}
+{"epoch": 83, "train/lr": 2.723341190722833e-06, "train/grad": 0.5353350731926981, "train/loss": 0.6665690827941895, "eval/hcp-train-subset/loss": 0.6846015088019832, "eval/hcp-val/loss": 0.6882900355323669}
+{"epoch": 84, "train/lr": 2.410390660064364e-06, "train/grad": 0.5079555152863985, "train/loss": 0.6729758046627045, "eval/hcp-train-subset/loss": 0.6839890845360295, "eval/hcp-val/loss": 0.6876086015855113}
+{"epoch": 85, "train/lr": 2.115307252323394e-06, "train/grad": 0.5114000629199604, "train/loss": 0.6732473783493041, "eval/hcp-train-subset/loss": 0.682847842093437, "eval/hcp-val/loss": 0.6886363250593985}
+{"epoch": 86, "train/lr": 1.838413636873912e-06, "train/grad": 0.5095351149372928, "train/loss": 0.674699307603836, "eval/hcp-train-subset/loss": 0.6827213283508055, "eval/hcp-val/loss": 0.6873139027626284}
+{"epoch": 87, "train/lr": 1.5800125928190943e-06, "train/grad": 0.5216995950735186, "train/loss": 0.6719457209491729, "eval/hcp-train-subset/loss": 0.6823524976930311, "eval/hcp-val/loss": 0.68680755553707}
+{"epoch": 88, "train/lr": 1.3403866779068222e-06, "train/grad": 0.5157867828849106, "train/loss": 0.6738916006755828, "eval/hcp-train-subset/loss": 0.6819413688875013, "eval/hcp-val/loss": 0.686128419253134}
+{"epoch": 89, "train/lr": 1.1197979195568888e-06, "train/grad": 0.5293626544646154, "train/loss": 0.6723607893371581, "eval/hcp-train-subset/loss": 0.6811894957096346, "eval/hcp-val/loss": 0.6866738603961083}
+{"epoch": 90, "train/lr": 9.184875283379039e-07, "train/grad": 0.5364315564506872, "train/loss": 0.6758320195865631, "eval/hcp-train-subset/loss": 0.6800519006867562, "eval/hcp-val/loss": 0.6874371382497972}
+{"epoch": 91, "train/lr": 7.366756342070463e-07, "train/grad": 0.5408778734563617, "train/loss": 0.6716363162708282, "eval/hcp-train-subset/loss": 0.6806942282184478, "eval/hcp-val/loss": 0.6858199998255698}
+{"epoch": 92, "train/lr": 5.745610458012273e-07, "train/grad": 0.5366432204527755, "train/loss": 0.6754466506195068, "eval/hcp-train-subset/loss": 0.6803765306549687, "eval/hcp-val/loss": 0.6864125795902745}
+{"epoch": 93, "train/lr": 4.323210330427178e-07, "train/grad": 0.5330239566002475, "train/loss": 0.6756944441366196, "eval/hcp-train-subset/loss": 0.6795724774560621, "eval/hcp-val/loss": 0.6865634697098886}
+{"epoch": 94, "train/lr": 3.1011113329712343e-07, "train/grad": 0.5502018680874731, "train/loss": 0.6735032947349548, "eval/hcp-train-subset/loss": 0.6790724502455804, "eval/hcp-val/loss": 0.6853404641151428}
+{"epoch": 95, "train/lr": 2.080649812955481e-07, "train/grad": 0.5306493457534395, "train/loss": 0.6762113899993897, "eval/hcp-train-subset/loss": 0.6782115438292103, "eval/hcp-val/loss": 0.68690893150145}
+{"epoch": 96, "train/lr": 1.2629416300698208e-07, "train/grad": 0.553940239649805, "train/loss": 0.6758231312561035, "eval/hcp-train-subset/loss": 0.6787174305608196, "eval/hcp-val/loss": 0.6860496247968366}
+{"epoch": 97, "train/lr": 6.488809362067338e-08, "train/grad": 0.5536697567005316, "train/loss": 0.6768844241905212, "eval/hcp-train-subset/loss": 0.6778920283240657, "eval/hcp-val/loss": 0.6852716909300897}
+{"epoch": 98, "train/lr": 2.391391977194211e-08, "train/grad": 0.5586897361279488, "train/loss": 0.675275718011856, "eval/hcp-train-subset/loss": 0.678021235812095, "eval/hcp-val/loss": 0.6854026202232607}
+{"epoch": 99, "train/lr": 3.4164461183156008e-09, "train/grad": 0.5736311806530409, "train/loss": 0.6750059153366089, "eval/hcp-train-subset/loss": 0.678764428823225, "eval/hcp-val/loss": 0.6859730895488493}

schaefer1000/schaefer1000_lr3e-4_2/eval_v2/nsd_cococlip__patch__attn/config.yaml

@@ -0,0 +1,96 @@
+output_root: experiments/schaefer1000/output
+name_prefix: eval_probe
+remote_root: null
+notes: schaefer1000 ablation schaefer1000_lr3e-4_2; eval v2 (nsd_cococlip patch attn)
+model_kwargs:
+  ckpt_path: experiments/schaefer1000/output/schaefer1000/schaefer1000_lr3e-4_2/pretrain/checkpoint-last.pth
+dataset_kwargs: {}
+classifier_kwargs:
+  embed_dim: null
+  dropout: 0.0
+  xavier_init: true
+  norm: true
+lr_scale_grid:
+- 0.02
+- 0.023
+- 0.028
+- 0.033
+- 0.038
+- 0.045
+- 0.053
+- 0.062
+- 0.074
+- 0.087
+- 0.1
+- 0.12
+- 0.14
+- 0.17
+- 0.2
+- 0.23
+- 0.27
+- 0.32
+- 0.38
+- 0.44
+- 0.52
+- 0.61
+- 0.72
+- 0.85
+- 1
+- 1.2
+- 1.4
+- 1.6
+- 1.9
+- 2.3
+- 2.7
+- 3.1
+- 3.7
+- 4.3
+- 5.1
+- 6
+- 7.1
+- 8.3
+- 9.8
+- 12
+- 14
+- 16
+- 19
+- 22
+- 26
+- 31
+- 36
+- 43
+- 50
+wd_scale_grid:
+- 1.0
+num_workers: 8
+prefetch_factor: null
+balanced_sampling: false
+epochs: 20
+steps_per_epoch: 200
+batch_size: 64
+accum_iter: 2
+lr: 0.0003
+warmup_epochs: 5
+no_decay: false
+weight_decay: 0.05
+clip_grad: 1.0
+metrics:
+- acc
+- f1
+cv_metric: acc
+early_stopping: true
+amp: true
+device: cuda
+seed: 4466
+debug: false
+wandb: false
+wandb_entity: null
+wandb_project: fMRI-fm-eval
+name: schaefer1000/schaefer1000_lr3e-4_2/eval_v2/nsd_cococlip__patch__attn
+model: schaefer1000_mae
+representation: patch
+classifier: attn
+dataset: nsd_cococlip
+distributed: false
+output_dir: experiments/schaefer1000/output/schaefer1000/schaefer1000_lr3e-4_2/eval_v2/nsd_cococlip__patch__attn
+remote_dir: null

schaefer1000/schaefer1000_lr3e-4_2/eval_v2/nsd_cococlip__patch__attn/eval_log.json

@@ -0,0 +1 @@
+{"eval/epoch": 15, "eval/id_best": 26, "eval/lr_best": 0.00041999999999999996, "eval/wd_best": 0.05, "eval/train/loss": 2.0325586795806885, "eval/train/acc": 0.3826177817388365, "eval/train/acc_std": 0.002453930462401187, "eval/train/f1": 0.33547680242756384, "eval/train/f1_std": 0.002603838208510603, "eval/validation/loss": 2.3741116523742676, "eval/validation/acc": 0.29346622369878184, "eval/validation/acc_std": 0.005477150118532977, "eval/validation/f1": 0.24165857914509759, "eval/validation/f1_std": 0.005435555118429112, "eval/test/loss": 2.272183418273926, "eval/test/acc": 0.3020408163265306, "eval/test/acc_std": 0.00528761988063169, "eval/test/f1": 0.23556231602832786, "eval/test/f1_std": 0.005092943816131354, "eval/testid/loss": 2.320650577545166, "eval/testid/acc": 0.28860613071139385, "eval/testid/acc_std": 0.00544716448803037, "eval/testid/f1": 0.2395835676711849, "eval/testid/f1_std": 0.0053362635263212044}

schaefer1000/schaefer1000_lr3e-4_2/eval_v2/nsd_cococlip__patch__attn/eval_log_best.json

@@ -0,0 +1 @@
+{"eval/best/epoch": 15, "eval/best/id_best": 26, "eval/best/lr_best": 0.00041999999999999996, "eval/best/wd_best": 0.05, "eval/best/train/loss": 2.0325586795806885, "eval/best/train/acc": 0.3826177817388365, "eval/best/train/acc_std": 0.002453930462401187, "eval/best/train/f1": 0.33547680242756384, "eval/best/train/f1_std": 0.002603838208510603, "eval/best/validation/loss": 2.3741116523742676, "eval/best/validation/acc": 0.29346622369878184, "eval/best/validation/acc_std": 0.005477150118532977, "eval/best/validation/f1": 0.24165857914509759, "eval/best/validation/f1_std": 0.005435555118429112, "eval/best/test/loss": 2.272183418273926, "eval/best/test/acc": 0.3020408163265306, "eval/best/test/acc_std": 0.00528761988063169, "eval/best/test/f1": 0.23556231602832786, "eval/best/test/f1_std": 0.005092943816131354, "eval/best/testid/loss": 2.320650577545166, "eval/best/testid/acc": 0.28860613071139385, "eval/best/testid/acc_std": 0.00544716448803037, "eval/best/testid/f1": 0.2395835676711849, "eval/best/testid/f1_std": 0.0053362635263212044}

schaefer1000/schaefer1000_lr3e-4_2/eval_v2/nsd_cococlip__patch__attn/eval_log_last.json

@@ -0,0 +1 @@
+{"eval/last/epoch": 19, "eval/last/id_best": 26, "eval/last/lr_best": 0.00041999999999999996, "eval/last/wd_best": 0.05, "eval/last/train/loss": 2.0110092163085938, "eval/last/train/acc": 0.3876578874581272, "eval/last/train/acc_std": 0.0024593302999561407, "eval/last/train/f1": 0.34005298399532985, "eval/last/train/f1_std": 0.0026559664241649016, "eval/last/validation/loss": 2.3858394622802734, "eval/last/validation/acc": 0.2918050941306755, "eval/last/validation/acc_std": 0.005459882420893537, "eval/last/validation/f1": 0.23717663868678565, "eval/last/validation/f1_std": 0.005304430026200211, "eval/last/test/loss": 2.2729647159576416, "eval/last/test/acc": 0.3031539888682746, "eval/last/test/acc_std": 0.005299005099827701, "eval/last/test/f1": 0.23458008582609738, "eval/last/test/f1_std": 0.00505673303209225, "eval/last/testid/loss": 2.321321725845337, "eval/last/testid/acc": 0.28879892037786775, "eval/last/testid/acc_std": 0.005462679125203189, "eval/last/testid/f1": 0.23930567998750338, "eval/last/testid/f1_std": 0.005308332861661975}

schaefer1000/schaefer1000_lr3e-4_2/eval_v2/nsd_cococlip__patch__attn/eval_table.csv

@@ -0,0 +1,5 @@
+model,repr,clf,dataset,ckpt,epoch,lr,wd,hparam_id,hparam,split,loss,acc,acc_std,f1,f1_std
+schaefer1000_mae,patch,attn,nsd_cococlip,best,15,0.00041999999999999996,0.05,26,"[1.4, 1.0]",train,2.0325586795806885,0.3826177817388365,0.002453930462401187,0.33547680242756384,0.002603838208510603
+schaefer1000_mae,patch,attn,nsd_cococlip,best,15,0.00041999999999999996,0.05,26,"[1.4, 1.0]",validation,2.3741116523742676,0.29346622369878184,0.005477150118532977,0.24165857914509759,0.005435555118429112
+schaefer1000_mae,patch,attn,nsd_cococlip,best,15,0.00041999999999999996,0.05,26,"[1.4, 1.0]",test,2.272183418273926,0.3020408163265306,0.00528761988063169,0.23556231602832786,0.005092943816131354
+schaefer1000_mae,patch,attn,nsd_cococlip,best,15,0.00041999999999999996,0.05,26,"[1.4, 1.0]",testid,2.320650577545166,0.28860613071139385,0.00544716448803037,0.2395835676711849,0.0053362635263212044

schaefer1000/schaefer1000_lr3e-4_2/eval_v2/nsd_cococlip__patch__attn/eval_table_best.csv

@@ -0,0 +1,5 @@
+model,repr,clf,dataset,ckpt,epoch,lr,wd,hparam_id,hparam,split,loss,acc,acc_std,f1,f1_std
+schaefer1000_mae,patch,attn,nsd_cococlip,best,15,0.00041999999999999996,0.05,26,"[1.4, 1.0]",train,2.0325586795806885,0.3826177817388365,0.002453930462401187,0.33547680242756384,0.002603838208510603
+schaefer1000_mae,patch,attn,nsd_cococlip,best,15,0.00041999999999999996,0.05,26,"[1.4, 1.0]",validation,2.3741116523742676,0.29346622369878184,0.005477150118532977,0.24165857914509759,0.005435555118429112
+schaefer1000_mae,patch,attn,nsd_cococlip,best,15,0.00041999999999999996,0.05,26,"[1.4, 1.0]",test,2.272183418273926,0.3020408163265306,0.00528761988063169,0.23556231602832786,0.005092943816131354
+schaefer1000_mae,patch,attn,nsd_cococlip,best,15,0.00041999999999999996,0.05,26,"[1.4, 1.0]",testid,2.320650577545166,0.28860613071139385,0.00544716448803037,0.2395835676711849,0.0053362635263212044

schaefer1000/schaefer1000_lr3e-4_2/eval_v2/nsd_cococlip__patch__attn/eval_table_last.csv

@@ -0,0 +1,5 @@
+model,repr,clf,dataset,ckpt,epoch,lr,wd,hparam_id,hparam,split,loss,acc,acc_std,f1,f1_std
+schaefer1000_mae,patch,attn,nsd_cococlip,last,19,0.00041999999999999996,0.05,26,"[1.4, 1.0]",train,2.0110092163085938,0.3876578874581272,0.0024593302999561407,0.34005298399532985,0.0026559664241649016
+schaefer1000_mae,patch,attn,nsd_cococlip,last,19,0.00041999999999999996,0.05,26,"[1.4, 1.0]",validation,2.3858394622802734,0.2918050941306755,0.005459882420893537,0.23717663868678565,0.005304430026200211
+schaefer1000_mae,patch,attn,nsd_cococlip,last,19,0.00041999999999999996,0.05,26,"[1.4, 1.0]",test,2.2729647159576416,0.3031539888682746,0.005299005099827701,0.23458008582609738,0.00505673303209225
+schaefer1000_mae,patch,attn,nsd_cococlip,last,19,0.00041999999999999996,0.05,26,"[1.4, 1.0]",testid,2.321321725845337,0.28879892037786775,0.005462679125203189,0.23930567998750338,0.005308332861661975

schaefer1000/schaefer1000_lr3e-4_2/eval_v2/nsd_cococlip__patch__attn/log.txt

@@ -0,0 +1,970 @@
+fMRI foundation model probe eval
+version: 0.1.dev95+g65be98d36
+sha: 87e31aaa465443ed5f0da58176ac8395447cdbd0, status: has uncommitted changes, branch: dev/clane9
+cwd: /data/connor/fmri-fm
+start: 2026-05-12 20:54:49
+config:
+output_root: experiments/schaefer1000/output
+name_prefix: eval_probe
+remote_root: null
+notes: schaefer1000 ablation schaefer1000_lr3e-4_2; eval v2 (nsd_cococlip patch attn)
+model_kwargs:
+  ckpt_path: experiments/schaefer1000/output/schaefer1000/schaefer1000_lr3e-4_2/pretrain/checkpoint-last.pth
+dataset_kwargs: {}
+classifier_kwargs:
+  embed_dim: null
+  dropout: 0.0
+  xavier_init: true
+  norm: true
+lr_scale_grid:
+- 0.02
+- 0.023
+- 0.028
+- 0.033
+- 0.038
+- 0.045
+- 0.053
+- 0.062
+- 0.074
+- 0.087
+- 0.1
+- 0.12
+- 0.14
+- 0.17
+- 0.2
+- 0.23
+- 0.27
+- 0.32
+- 0.38
+- 0.44
+- 0.52
+- 0.61
+- 0.72
+- 0.85
+- 1
+- 1.2
+- 1.4
+- 1.6
+- 1.9
+- 2.3
+- 2.7
+- 3.1
+- 3.7
+- 4.3
+- 5.1
+- 6
+- 7.1
+- 8.3
+- 9.8
+- 12
+- 14
+- 16
+- 19
+- 22
+- 26
+- 31
+- 36
+- 43
+- 50
+wd_scale_grid:
+- 1.0
+num_workers: 8
+prefetch_factor: null
+balanced_sampling: false
+epochs: 20
+steps_per_epoch: 200
+batch_size: 64
+accum_iter: 2
+lr: 0.0003
+warmup_epochs: 5
+no_decay: false
+weight_decay: 0.05
+clip_grad: 1.0
+metrics:
+- acc
+- f1
+cv_metric: acc
+early_stopping: true
+amp: true
+device: cuda
+seed: 4466
+debug: false
+wandb: false
+wandb_entity: null
+wandb_project: fMRI-fm-eval
+name: schaefer1000/schaefer1000_lr3e-4_2/eval_v2/nsd_cococlip__patch__attn
+model: schaefer1000_mae
+representation: patch
+classifier: attn
+dataset: nsd_cococlip
+distributed: false
+output_dir: experiments/schaefer1000/output/schaefer1000/schaefer1000_lr3e-4_2/eval_v2/nsd_cococlip__patch__attn
+remote_dir: null
+
+creating frozen backbone model: schaefer1000_mae
+backbone:
+MaskedEncoderWrapper(
+  (model): MaskedEncoder(
+    class_token=True, reg_tokens=0, no_embed_class=True, mask_drop_scale=False
+    (patchify): Patchify3D((16, 1000, 1), (4, 1, 1), in_chans=1)
+    (patch_embed): Linear(in_features=4, out_features=768, bias=True)
+    (pos_embed): SeparablePosEmbed(768, (4, 1000, 1))
+    (blocks): ModuleList(
+      (0-11): 12 x Block(
+        (norm1): LayerNorm((768,), eps=1e-06, elementwise_affine=True)
+        (attn): Attention(
+          num_heads=12
+          (q): Linear(in_features=768, out_features=768, bias=True)
+          (k): Linear(in_features=768, out_features=768, bias=True)
+          (v): Linear(in_features=768, out_features=768, bias=True)
+          (proj): Linear(in_features=768, out_features=768, bias=True)
+        )
+        (drop_path1): Identity()
+        (norm2): LayerNorm((768,), eps=1e-06, elementwise_affine=True)
+        (mlp): Mlp(
+          (fc1): Linear(in_features=768, out_features=3072, bias=True)
+          (act): GELU(approximate='none')
+          (fc2): Linear(in_features=3072, out_features=768, bias=True)
+        )
+        (drop_path2): Identity()
+      )
+    )
+    (norm): LayerNorm((768,), eps=1e-06, elementwise_affine=True)
+  )
+)
+creating dataset: nsd_cococlip (schaefer1000)
+train (n=32539):
+HFDataset(
+    dataset=Dataset({
+    features: ['sub', 'ses', 'run', 'trial_id', 'nsd_id', 'category_id', 'path', 'start', 'end', 'n_frames', 'tr', 'bold', 'mean', 'std'],
+    num_rows: 32539
+}),
+    labels=[ 3  4  5  6 10 11 17 18 19 20 22 23 25 30 31 33 36 37 38 53 55 59 61 74],
+    counts=[1286 1180 1639 1868  834  824 1026 1042  913 1853 1503 2092 1001 1410
+  794 1241 1904 1872 2267 1428  889  904 1447 1322]
+)
+
+validation (n=5418):
+HFDataset(
+    dataset=Dataset({
+    features: ['sub', 'ses', 'run', 'trial_id', 'nsd_id', 'category_id', 'path', 'start', 'end', 'n_frames', 'tr', 'bold', 'mean', 'std'],
+    num_rows: 5418
+}),
+    labels=[ 3  4  5  6 10 11 17 18 19 20 22 23 25 30 31 33 36 37 38 53 55 59 61 74],
+    counts=[197 161 276 345 126 142 143 185 112 295 285 387 169 250 159 193 316 334
+ 343 215 172 141 226 246]
+)
+
+test (n=5390):
+HFDataset(
+    dataset=Dataset({
+    features: ['sub', 'ses', 'run', 'trial_id', 'nsd_id', 'category_id', 'path', 'start', 'end', 'n_frames', 'tr', 'bold', 'mean', 'std'],
+    num_rows: 5390
+}),
+    labels=[ 3  4  5  6 10 11 17 18 19 20 22 23 25 30 31 33 36 37 38 53 55 59 61 74],
+    counts=[202 172 274 298 144 180 134 182 186 293 218 343 165 185 140 177 346 333
+ 345 271 165 140 251 246]
+)
+
+testid (n=5187):
+HFDataset(
+    dataset=Dataset({
+    features: ['sub', 'ses', 'run', 'trial_id', 'nsd_id', 'category_id', 'path', 'start', 'end', 'n_frames', 'tr', 'bold', 'mean', 'std'],
+    num_rows: 5187
+}),
+    labels=[ 3  4  5  6 10 11 17 18 19 20 22 23 25 30 31 33 36 37 38 53 55 59 61 74],
+    counts=[197 159 267 273 123 153 175 184 139 310 215 386 153 230 118 192 330 306
+ 349 223 143 127 249 186]
+)
+
+running backbone on example batch to get embedding dim
+embedding feature dim (patch): 768
+initializing sweep of classifier heads
+classifiers:
+ModuleList(
+  (0-48): 49 x AttnPoolClassifier(
+    (kv): Linear(in_features=768, out_features=1536, bias=True)
+    (linear): Linear(in_features=768, out_features=24, bias=True)
+  )
+)
+classifier params (train): 58.8M (58.8M)
+setting up optimizer
+total batch size: 128 = 64 bs per gpu x 2 accum
+lr: 3.00e-04
+full schedule: epochs = 20 (steps = 4000) (decay = True)
+warmup: epochs = 5 (steps = 1000)
+start training for 20 epochs
+train: [0]  [  0/400]  eta: 0:10:25  lr: nan  time: 1.5627  data: 0.6251  max mem: 56639
+train: [0]  [ 20/400]  eta: 0:04:27  lr: 0.000003  loss: 3.2090 (3.2151)  grad: 0.1984 (0.2093)  time: 0.6602  data: 0.0025  max mem: 57344
+train: [0]  [ 40/400]  eta: 0:04:03  lr: 0.000006  loss: 3.2090 (3.2081)  grad: 0.1984 (0.2010)  time: 0.6487  data: 0.0037  max mem: 57344
+train: [0]  [ 60/400]  eta: 0:03:46  lr: 0.000009  loss: 3.1874 (3.2027)  grad: 0.1972 (0.2026)  time: 0.6480  data: 0.0034  max mem: 57344
+train: [0]  [ 80/400]  eta: 0:03:32  lr: 0.000012  loss: 3.1831 (3.1964)  grad: 0.1951 (0.2001)  time: 0.6494  data: 0.0035  max mem: 57344
+train: [0]  [100/400]  eta: 0:03:18  lr: 0.000015  loss: 3.1781 (3.1917)  grad: 0.1854 (0.1987)  time: 0.6509  data: 0.0038  max mem: 57344
+train: [0]  [120/400]  eta: 0:03:04  lr: 0.000018  loss: 3.1652 (3.1889)  grad: 0.1884 (0.1966)  time: 0.6499  data: 0.0037  max mem: 57344
+train: [0]  [140/400]  eta: 0:02:50  lr: 0.000021  loss: 3.1575 (3.1846)  grad: 0.1884 (0.1957)  time: 0.6502  data: 0.0037  max mem: 57344
+train: [0]  [160/400]  eta: 0:02:37  lr: 0.000024  loss: 3.1585 (3.1830)  grad: 0.1813 (0.1937)  time: 0.6490  data: 0.0037  max mem: 57344
+train: [0]  [180/400]  eta: 0:02:24  lr: 0.000027  loss: 3.1763 (3.1812)  grad: 0.1684 (0.1915)  time: 0.6487  data: 0.0037  max mem: 57344
+train: [0]  [200/400]  eta: 0:02:10  lr: 0.000030  loss: 3.1518 (3.1780)  grad: 0.1678 (0.1904)  time: 0.6493  data: 0.0037  max mem: 57344
+train: [0]  [220/400]  eta: 0:01:57  lr: 0.000033  loss: 3.1509 (3.1757)  grad: 0.1891 (0.1905)  time: 0.6493  data: 0.0037  max mem: 57344
+train: [0]  [240/400]  eta: 0:01:44  lr: 0.000036  loss: 3.1527 (3.1739)  grad: 0.1830 (0.1894)  time: 0.6501  data: 0.0037  max mem: 57344
+train: [0]  [260/400]  eta: 0:01:31  lr: 0.000039  loss: 3.1555 (3.1727)  grad: 0.1715 (0.1881)  time: 0.6495  data: 0.0037  max mem: 57344
+train: [0]  [280/400]  eta: 0:01:18  lr: 0.000042  loss: 3.1462 (3.1701)  grad: 0.1746 (0.1869)  time: 0.6500  data: 0.0037  max mem: 57344
+train: [0]  [300/400]  eta: 0:01:05  lr: 0.000045  loss: 3.1526 (3.1696)  grad: 0.1746 (0.1861)  time: 0.6495  data: 0.0037  max mem: 57344
+train: [0]  [320/400]  eta: 0:00:52  lr: 0.000048  loss: 3.1546 (3.1679)  grad: 0.1762 (0.1861)  time: 0.6491  data: 0.0037  max mem: 57344
+train: [0]  [340/400]  eta: 0:00:39  lr: 0.000051  loss: 3.1393 (3.1662)  grad: 0.1762 (0.1853)  time: 0.6486  data: 0.0037  max mem: 57344
+train: [0]  [360/400]  eta: 0:00:26  lr: 0.000054  loss: 3.1350 (3.1654)  grad: 0.1738 (0.1847)  time: 0.6496  data: 0.0038  max mem: 57344
+train: [0]  [380/400]  eta: 0:00:13  lr: 0.000057  loss: 3.1416 (3.1633)  grad: 0.1681 (0.1840)  time: 0.6493  data: 0.0037  max mem: 57344
+train: [0]  [399/400]  eta: 0:00:00  lr: 0.000060  loss: 3.1306 (3.1616)  grad: 0.1681 (0.1833)  time: 0.6491  data: 0.0037  max mem: 57344
+train: [0] Total time: 0:04:21 (0.6525 s / it)
+train: [0] Summary: lr: 0.000060  loss: 3.1306 (3.1616)  grad: 0.1681 (0.1833)
+eval (validation): [0]  [ 0/85]  eta: 0:01:22    time: 0.9680  data: 0.6065  max mem: 57344
+eval (validation): [0]  [20/85]  eta: 0:00:25    time: 0.3706  data: 0.0036  max mem: 57344
+eval (validation): [0]  [40/85]  eta: 0:00:17    time: 0.3715  data: 0.0037  max mem: 57344
+eval (validation): [0]  [60/85]  eta: 0:00:09    time: 0.3709  data: 0.0036  max mem: 57344
+eval (validation): [0]  [80/85]  eta: 0:00:01    time: 0.3706  data: 0.0036  max mem: 57344
+eval (validation): [0]  [84/85]  eta: 0:00:00    time: 0.3640  data: 0.0036  max mem: 57344
+eval (validation): [0] Total time: 0:00:32 (0.3777 s / it)
+cv: [0]  best hparam: (43, 1.0) (047) ('047_lr4.3e+01_wd1.0e+00')  loss: 2.726  acc: 0.171  f1: 0.103
+saving checkpoint experiments/schaefer1000/output/schaefer1000/schaefer1000_lr3e-4_2/eval_v2/nsd_cococlip__patch__attn/checkpoint-last.pth
+saving best checkpoint experiments/schaefer1000/output/schaefer1000/schaefer1000_lr3e-4_2/eval_v2/nsd_cococlip__patch__attn/checkpoint-best.pth
+train: [1]  [  0/400]  eta: 0:07:48  lr: nan  time: 1.1723  data: 0.5386  max mem: 57344
+train: [1]  [ 20/400]  eta: 0:04:15  lr: 0.000063  loss: 3.0650 (3.0828)  grad: 0.1638 (0.1733)  time: 0.6484  data: 0.0032  max mem: 57344
+train: [1]  [ 40/400]  eta: 0:03:58  lr: 0.000066  loss: 3.0658 (3.0884)  grad: 0.1637 (0.1681)  time: 0.6501  data: 0.0037  max mem: 57344
+train: [1]  [ 60/400]  eta: 0:03:43  lr: 0.000069  loss: 3.0812 (3.0899)  grad: 0.1681 (0.1734)  time: 0.6502  data: 0.0039  max mem: 57344
+train: [1]  [ 80/400]  eta: 0:03:29  lr: 0.000072  loss: 3.0954 (3.0957)  grad: 0.1803 (0.1760)  time: 0.6492  data: 0.0036  max mem: 57344
+train: [1]  [100/400]  eta: 0:03:16  lr: 0.000075  loss: 3.0865 (3.0922)  grad: 0.1803 (0.1771)  time: 0.6483  data: 0.0036  max mem: 57344
+train: [1]  [120/400]  eta: 0:03:02  lr: 0.000078  loss: 3.0628 (3.0881)  grad: 0.1769 (0.1770)  time: 0.6491  data: 0.0036  max mem: 57344
+train: [1]  [140/400]  eta: 0:02:49  lr: 0.000081  loss: 3.0533 (3.0813)  grad: 0.1753 (0.1784)  time: 0.6486  data: 0.0036  max mem: 57344
+train: [1]  [160/400]  eta: 0:02:36  lr: 0.000084  loss: 3.0533 (3.0782)  grad: 0.1776 (0.1784)  time: 0.6488  data: 0.0037  max mem: 57344
+train: [1]  [180/400]  eta: 0:02:23  lr: 0.000087  loss: 3.0542 (3.0782)  grad: 0.1793 (0.1789)  time: 0.6487  data: 0.0037  max mem: 57344
+train: [1]  [200/400]  eta: 0:02:10  lr: 0.000090  loss: 3.0654 (3.0764)  grad: 0.1887 (0.1803)  time: 0.6495  data: 0.0037  max mem: 57344
+train: [1]  [220/400]  eta: 0:01:57  lr: 0.000093  loss: 3.0572 (3.0749)  grad: 0.1914 (0.1811)  time: 0.6494  data: 0.0037  max mem: 57344
+train: [1]  [240/400]  eta: 0:01:44  lr: 0.000096  loss: 3.0326 (3.0707)  grad: 0.1793 (0.1816)  time: 0.6501  data: 0.0037  max mem: 57344
+train: [1]  [260/400]  eta: 0:01:31  lr: 0.000099  loss: 3.0070 (3.0681)  grad: 0.1930 (0.1831)  time: 0.6487  data: 0.0036  max mem: 57344
+train: [1]  [280/400]  eta: 0:01:18  lr: 0.000102  loss: 3.0515 (3.0668)  grad: 0.2030 (0.1847)  time: 0.6489  data: 0.0037  max mem: 57344
+train: [1]  [300/400]  eta: 0:01:05  lr: 0.000105  loss: 3.0171 (3.0628)  grad: 0.2070 (0.1865)  time: 0.6487  data: 0.0037  max mem: 57344
+train: [1]  [320/400]  eta: 0:00:52  lr: 0.000108  loss: 3.0139 (3.0615)  grad: 0.2040 (0.1872)  time: 0.6489  data: 0.0037  max mem: 57344
+train: [1]  [340/400]  eta: 0:00:39  lr: 0.000111  loss: 3.0092 (3.0589)  grad: 0.2059 (0.1892)  time: 0.6497  data: 0.0038  max mem: 57344
+train: [1]  [360/400]  eta: 0:00:26  lr: 0.000114  loss: 3.0326 (3.0585)  grad: 0.2217 (0.1913)  time: 0.6493  data: 0.0037  max mem: 57344
+train: [1]  [380/400]  eta: 0:00:13  lr: 0.000117  loss: 3.0571 (3.0577)  grad: 0.2253 (0.1933)  time: 0.6485  data: 0.0036  max mem: 57344
+train: [1]  [399/400]  eta: 0:00:00  lr: 0.000120  loss: 3.0588 (3.0592)  grad: 0.2476 (0.2003)  time: 0.6492  data: 0.0037  max mem: 57344
+train: [1] Total time: 0:04:20 (0.6507 s / it)
+train: [1] Summary: lr: 0.000120  loss: 3.0588 (3.0592)  grad: 0.2476 (0.2003)
+eval (validation): [1]  [ 0/85]  eta: 0:01:25    time: 1.0095  data: 0.6474  max mem: 57344
+eval (validation): [1]  [20/85]  eta: 0:00:26    time: 0.3728  data: 0.0033  max mem: 57344
+eval (validation): [1]  [40/85]  eta: 0:00:17    time: 0.3742  data: 0.0040  max mem: 57344
+eval (validation): [1]  [60/85]  eta: 0:00:09    time: 0.3740  data: 0.0039  max mem: 57344
+eval (validation): [1]  [80/85]  eta: 0:00:01    time: 0.3716  data: 0.0039  max mem: 57344
+eval (validation): [1]  [84/85]  eta: 0:00:00    time: 0.3652  data: 0.0038  max mem: 57344
+eval (validation): [1] Total time: 0:00:32 (0.3804 s / it)
+cv: [1]  best hparam: (19, 1.0) (042) ('042_lr1.9e+01_wd1.0e+00')  loss: 2.558  acc: 0.221  f1: 0.160
+saving checkpoint experiments/schaefer1000/output/schaefer1000/schaefer1000_lr3e-4_2/eval_v2/nsd_cococlip__patch__attn/checkpoint-last.pth
+saving best checkpoint experiments/schaefer1000/output/schaefer1000/schaefer1000_lr3e-4_2/eval_v2/nsd_cococlip__patch__attn/checkpoint-best.pth
+train: [2]  [  0/400]  eta: 0:08:18  lr: nan  time: 1.2462  data: 0.6093  max mem: 57344
+train: [2]  [ 20/400]  eta: 0:04:17  lr: 0.000123  loss: 3.4532 (3.4319)  grad: 1.1697 (1.1508)  time: 0.6489  data: 0.0026  max mem: 57344
+WARNING: classifier 48 (50, 1.0) diverged (loss=78.79 > 63.56) at step 415. Freezing.
+train: [2]  [ 40/400]  eta: 0:03:58  lr: 0.000126  loss: 3.4139 (3.4032)  grad: 1.0654 (1.0460)  time: 0.6467  data: 0.0036  max mem: 57344
+train: [2]  [ 60/400]  eta: 0:03:43  lr: 0.000129  loss: 2.9790 (3.2565)  grad: 0.2045 (0.7650)  time: 0.6432  data: 0.0038  max mem: 57344
+train: [2]  [ 80/400]  eta: 0:03:28  lr: 0.000132  loss: 2.9911 (3.1937)  grad: 0.2072 (0.6293)  time: 0.6422  data: 0.0035  max mem: 57344
+train: [2]  [100/400]  eta: 0:03:15  lr: 0.000135  loss: 2.9926 (3.1532)  grad: 0.2056 (0.5426)  time: 0.6435  data: 0.0037  max mem: 57344
+train: [2]  [120/400]  eta: 0:03:01  lr: 0.000138  loss: 3.0032 (3.1302)  grad: 0.2061 (0.4906)  time: 0.6434  data: 0.0037  max mem: 57344
+train: [2]  [140/400]  eta: 0:02:48  lr: 0.000141  loss: 3.0164 (3.1165)  grad: 0.2652 (0.4753)  time: 0.6438  data: 0.0038  max mem: 57344
+train: [2]  [160/400]  eta: 0:02:35  lr: 0.000144  loss: 3.0868 (3.1707)  grad: 0.4787 (0.5939)  time: 0.6439  data: 0.0038  max mem: 57344
+WARNING: classifier 47 (43, 1.0) diverged (loss=72.37 > 63.56) at step 482. Freezing.
+train: [2]  [180/400]  eta: 0:02:22  lr: 0.000147  loss: 3.2231 (3.1734)  grad: 0.9515 (0.6172)  time: 0.6396  data: 0.0039  max mem: 57344
+train: [2]  [200/400]  eta: 0:02:09  lr: 0.000150  loss: 2.9710 (3.1515)  grad: 0.2221 (0.5770)  time: 0.6378  data: 0.0037  max mem: 57344
+train: [2]  [220/400]  eta: 0:01:56  lr: 0.000153  loss: 2.9639 (3.1349)  grad: 0.2240 (0.5481)  time: 0.6379  data: 0.0038  max mem: 57344
+train: [2]  [240/400]  eta: 0:01:43  lr: 0.000156  loss: 3.0139 (3.1412)  grad: 0.3065 (0.5686)  time: 0.6380  data: 0.0038  max mem: 57344
+WARNING: classifier 46 (36, 1.0) diverged (loss=72.08 > 63.56) at step 527. Freezing.
+train: [2]  [260/400]  eta: 0:01:30  lr: 0.000159  loss: 3.2287 (3.1851)  grad: 0.9400 (0.6638)  time: 0.6361  data: 0.0037  max mem: 57344
+train: [2]  [280/400]  eta: 0:01:17  lr: 0.000162  loss: 3.0737 (3.1723)  grad: 0.2338 (0.6323)  time: 0.6325  data: 0.0039  max mem: 57344
+train: [2]  [300/400]  eta: 0:01:04  lr: 0.000165  loss: 2.9957 (3.1607)  grad: 0.2189 (0.6049)  time: 0.6327  data: 0.0039  max mem: 57344
+train: [2]  [320/400]  eta: 0:00:51  lr: 0.000168  loss: 2.9960 (3.1507)  grad: 0.2225 (0.5816)  time: 0.6319  data: 0.0038  max mem: 57344
+train: [2]  [340/400]  eta: 0:00:38  lr: 0.000171  loss: 2.9953 (3.1410)  grad: 0.2265 (0.5608)  time: 0.6319  data: 0.0037  max mem: 57344
+train: [2]  [360/400]  eta: 0:00:25  lr: 0.000174  loss: 3.0002 (3.1341)  grad: 0.2397 (0.5449)  time: 0.6323  data: 0.0038  max mem: 57344
+train: [2]  [380/400]  eta: 0:00:12  lr: 0.000177  loss: 3.0323 (3.1289)  grad: 0.3061 (0.5368)  time: 0.6327  data: 0.0038  max mem: 57344
+train: [2]  [399/400]  eta: 0:00:00  lr: 0.000180  loss: 3.0969 (3.1490)  grad: 0.5112 (0.5774)  time: 0.6324  data: 0.0039  max mem: 57344
+train: [2] Total time: 0:04:16 (0.6404 s / it)
+train: [2] Summary: lr: 0.000180  loss: 3.0969 (3.1490)  grad: 0.5112 (0.5774)
+eval (validation): [2]  [ 0/85]  eta: 0:01:31    time: 1.0747  data: 0.7111  max mem: 57344
+eval (validation): [2]  [20/85]  eta: 0:00:26    time: 0.3689  data: 0.0031  max mem: 57344
+eval (validation): [2]  [40/85]  eta: 0:00:17    time: 0.3700  data: 0.0036  max mem: 57344
+eval (validation): [2]  [60/85]  eta: 0:00:09    time: 0.3700  data: 0.0036  max mem: 57344
+eval (validation): [2]  [80/85]  eta: 0:00:01    time: 0.3694  data: 0.0036  max mem: 57344
+eval (validation): [2]  [84/85]  eta: 0:00:00    time: 0.3631  data: 0.0037  max mem: 57344
+eval (validation): [2] Total time: 0:00:32 (0.3777 s / it)
+cv: [2]  best hparam: (9.8, 1.0) (038) ('038_lr9.8e+00_wd1.0e+00')  loss: 2.510  acc: 0.235  f1: 0.174
+saving checkpoint experiments/schaefer1000/output/schaefer1000/schaefer1000_lr3e-4_2/eval_v2/nsd_cococlip__patch__attn/checkpoint-last.pth
+saving best checkpoint experiments/schaefer1000/output/schaefer1000/schaefer1000_lr3e-4_2/eval_v2/nsd_cococlip__patch__attn/checkpoint-best.pth
+train: [3]  [  0/400]  eta: 0:08:25  lr: nan  time: 1.2640  data: 0.6462  max mem: 57344
+WARNING: classifier 45 (31, 1.0) diverged (loss=64.82 > 63.56) at step 603. Freezing.
+train: [3]  [ 20/400]  eta: 0:04:09  lr: 0.000183  loss: 2.9208 (3.2454)  grad: 0.2131 (0.7254)  time: 0.6263  data: 0.0021  max mem: 57344
+train: [3]  [ 40/400]  eta: 0:03:51  lr: 0.000186  loss: 2.9417 (3.1151)  grad: 0.2131 (0.4696)  time: 0.6272  data: 0.0036  max mem: 57344
+train: [3]  [ 60/400]  eta: 0:03:36  lr: 0.000189  loss: 2.9472 (3.0600)  grad: 0.2044 (0.3784)  time: 0.6262  data: 0.0037  max mem: 57344
+train: [3]  [ 80/400]  eta: 0:03:22  lr: 0.000192  loss: 2.9253 (3.0218)  grad: 0.1974 (0.3335)  time: 0.6252  data: 0.0035  max mem: 57344
+train: [3]  [100/400]  eta: 0:03:09  lr: 0.000195  loss: 2.9335 (3.0070)  grad: 0.2009 (0.3086)  time: 0.6250  data: 0.0035  max mem: 57344
+train: [3]  [120/400]  eta: 0:02:56  lr: 0.000198  loss: 2.9508 (2.9951)  grad: 0.2027 (0.2905)  time: 0.6258  data: 0.0036  max mem: 57344
+train: [3]  [140/400]  eta: 0:02:43  lr: 0.000201  loss: 2.9580 (2.9877)  grad: 0.1947 (0.2776)  time: 0.6261  data: 0.0036  max mem: 57344
+train: [3]  [160/400]  eta: 0:02:31  lr: 0.000204  loss: 2.8809 (2.9717)  grad: 0.1947 (0.2670)  time: 0.6259  data: 0.0036  max mem: 57344
+train: [3]  [180/400]  eta: 0:02:18  lr: 0.000207  loss: 2.9202 (2.9716)  grad: 0.2077 (0.2619)  time: 0.6262  data: 0.0036  max mem: 57344
+train: [3]  [200/400]  eta: 0:02:05  lr: 0.000210  loss: 2.9501 (2.9702)  grad: 0.2207 (0.2579)  time: 0.6262  data: 0.0037  max mem: 57344
+train: [3]  [220/400]  eta: 0:01:53  lr: 0.000213  loss: 2.9268 (2.9664)  grad: 0.2235 (0.2554)  time: 0.6263  data: 0.0038  max mem: 57344
+train: [3]  [240/400]  eta: 0:01:40  lr: 0.000216  loss: 2.9032 (2.9633)  grad: 0.2269 (0.2532)  time: 0.6260  data: 0.0037  max mem: 57344
+train: [3]  [260/400]  eta: 0:01:27  lr: 0.000219  loss: 2.9332 (2.9606)  grad: 0.2336 (0.2520)  time: 0.6258  data: 0.0037  max mem: 57344
+train: [3]  [280/400]  eta: 0:01:15  lr: 0.000222  loss: 2.9415 (2.9578)  grad: 0.2288 (0.2505)  time: 0.6268  data: 0.0038  max mem: 57344
+train: [3]  [300/400]  eta: 0:01:02  lr: 0.000225  loss: 2.8925 (2.9547)  grad: 0.2301 (0.2491)  time: 0.6260  data: 0.0037  max mem: 57344
+train: [3]  [320/400]  eta: 0:00:50  lr: 0.000228  loss: 2.8984 (2.9532)  grad: 0.2245 (0.2475)  time: 0.6261  data: 0.0037  max mem: 57344
+train: [3]  [340/400]  eta: 0:00:37  lr: 0.000231  loss: 2.9056 (2.9509)  grad: 0.2199 (0.2458)  time: 0.6254  data: 0.0037  max mem: 57344
+train: [3]  [360/400]  eta: 0:00:25  lr: 0.000234  loss: 2.9062 (2.9493)  grad: 0.2215 (0.2455)  time: 0.6260  data: 0.0038  max mem: 57344
+train: [3]  [380/400]  eta: 0:00:12  lr: 0.000237  loss: 2.9062 (2.9465)  grad: 0.2400 (0.2451)  time: 0.6264  data: 0.0038  max mem: 57344
+train: [3]  [399/400]  eta: 0:00:00  lr: 0.000240  loss: 2.9187 (2.9450)  grad: 0.2385 (0.2447)  time: 0.6263  data: 0.0037  max mem: 57344
+train: [3] Total time: 0:04:11 (0.6279 s / it)
+train: [3] Summary: lr: 0.000240  loss: 2.9187 (2.9450)  grad: 0.2385 (0.2447)
+eval (validation): [3]  [ 0/85]  eta: 0:01:26    time: 1.0185  data: 0.6576  max mem: 57344
+eval (validation): [3]  [20/85]  eta: 0:00:26    time: 0.3706  data: 0.0027  max mem: 57344
+eval (validation): [3]  [40/85]  eta: 0:00:17    time: 0.3715  data: 0.0036  max mem: 57344
+eval (validation): [3]  [60/85]  eta: 0:00:09    time: 0.3732  data: 0.0037  max mem: 57344
+eval (validation): [3]  [80/85]  eta: 0:00:01    time: 0.3720  data: 0.0037  max mem: 57344
+eval (validation): [3]  [84/85]  eta: 0:00:00    time: 0.3661  data: 0.0036  max mem: 57344
+eval (validation): [3] Total time: 0:00:32 (0.3792 s / it)
+cv: [3]  best hparam: (5.1, 1.0) (034) ('034_lr5.1e+00_wd1.0e+00')  loss: 2.538  acc: 0.243  f1: 0.186
+saving checkpoint experiments/schaefer1000/output/schaefer1000/schaefer1000_lr3e-4_2/eval_v2/nsd_cococlip__patch__attn/checkpoint-last.pth
+saving best checkpoint experiments/schaefer1000/output/schaefer1000/schaefer1000_lr3e-4_2/eval_v2/nsd_cococlip__patch__attn/checkpoint-best.pth
+train: [4]  [  0/400]  eta: 0:08:29  lr: nan  time: 1.2735  data: 0.6616  max mem: 57344
+train: [4]  [ 20/400]  eta: 0:04:09  lr: 0.000243  loss: 2.8975 (2.9113)  grad: 0.2334 (0.2360)  time: 0.6258  data: 0.0026  max mem: 57344
+train: [4]  [ 40/400]  eta: 0:03:51  lr: 0.000246  loss: 2.8975 (2.8959)  grad: 0.2246 (0.2276)  time: 0.6262  data: 0.0035  max mem: 57344
+train: [4]  [ 60/400]  eta: 0:03:36  lr: 0.000249  loss: 2.8891 (2.8927)  grad: 0.2264 (0.2297)  time: 0.6257  data: 0.0037  max mem: 57344
+train: [4]  [ 80/400]  eta: 0:03:22  lr: 0.000252  loss: 2.8891 (2.8897)  grad: 0.2310 (0.2291)  time: 0.6266  data: 0.0037  max mem: 57344
+train: [4]  [100/400]  eta: 0:03:09  lr: 0.000255  loss: 2.8627 (2.8856)  grad: 0.2347 (0.2318)  time: 0.6257  data: 0.0037  max mem: 57344
+train: [4]  [120/400]  eta: 0:02:56  lr: 0.000258  loss: 2.8715 (2.8828)  grad: 0.2330 (0.2300)  time: 0.6246  data: 0.0034  max mem: 57344
+train: [4]  [140/400]  eta: 0:02:43  lr: 0.000261  loss: 2.8715 (2.8786)  grad: 0.2207 (0.2297)  time: 0.6265  data: 0.0038  max mem: 57344
+train: [4]  [160/400]  eta: 0:02:31  lr: 0.000264  loss: 2.8272 (2.8775)  grad: 0.2289 (0.2307)  time: 0.6266  data: 0.0038  max mem: 57344
+train: [4]  [180/400]  eta: 0:02:18  lr: 0.000267  loss: 2.9003 (2.8812)  grad: 0.2595 (0.2429)  time: 0.6263  data: 0.0037  max mem: 57344
+WARNING: classifier 44 (26, 1.0) diverged (loss=63.77 > 63.56) at step 899. Freezing.
+train: [4]  [200/400]  eta: 0:02:05  lr: 0.000270  loss: 2.9922 (2.9377)  grad: 0.4441 (0.3711)  time: 0.6258  data: 0.0038  max mem: 57344
+train: [4]  [220/400]  eta: 0:01:53  lr: 0.000273  loss: 2.9505 (2.9327)  grad: 0.2453 (0.3584)  time: 0.6198  data: 0.0038  max mem: 57344
+train: [4]  [240/400]  eta: 0:01:40  lr: 0.000276  loss: 2.8617 (2.9275)  grad: 0.2214 (0.3470)  time: 0.6202  data: 0.0039  max mem: 57344
+train: [4]  [260/400]  eta: 0:01:27  lr: 0.000279  loss: 2.8625 (2.9262)  grad: 0.2214 (0.3379)  time: 0.6202  data: 0.0038  max mem: 57344
+train: [4]  [280/400]  eta: 0:01:15  lr: 0.000282  loss: 2.8841 (2.9231)  grad: 0.2260 (0.3303)  time: 0.6209  data: 0.0039  max mem: 57344
+train: [4]  [300/400]  eta: 0:01:02  lr: 0.000285  loss: 2.8841 (2.9213)  grad: 0.2108 (0.3221)  time: 0.6205  data: 0.0038  max mem: 57344
+train: [4]  [320/400]  eta: 0:00:50  lr: 0.000288  loss: 2.8683 (2.9186)  grad: 0.2178 (0.3164)  time: 0.6208  data: 0.0041  max mem: 57344
+train: [4]  [340/400]  eta: 0:00:37  lr: 0.000291  loss: 2.8872 (2.9177)  grad: 0.2360 (0.3132)  time: 0.6205  data: 0.0039  max mem: 57344
+train: [4]  [360/400]  eta: 0:00:25  lr: 0.000294  loss: 2.9237 (2.9309)  grad: 0.3016 (0.3509)  time: 0.6213  data: 0.0039  max mem: 57344
+WARNING: classifier 42 (19, 1.0) diverged (loss=68.55 > 63.56) at step 986. Freezing.
+train: [4]  [380/400]  eta: 0:00:12  lr: 0.000297  loss: 3.0970 (2.9616)  grad: 0.8665 (0.4061)  time: 0.6182  data: 0.0039  max mem: 57344
+train: [4]  [399/400]  eta: 0:00:00  lr: 0.000300  loss: 2.8509 (2.9566)  grad: 0.2354 (0.3969)  time: 0.6142  data: 0.0038  max mem: 57344
+train: [4] Total time: 0:04:09 (0.6247 s / it)
+train: [4] Summary: lr: 0.000300  loss: 2.8509 (2.9566)  grad: 0.2354 (0.3969)
+eval (validation): [4]  [ 0/85]  eta: 0:01:31    time: 1.0743  data: 0.7106  max mem: 57344
+eval (validation): [4]  [20/85]  eta: 0:00:26    time: 0.3688  data: 0.0038  max mem: 57344
+eval (validation): [4]  [40/85]  eta: 0:00:17    time: 0.3700  data: 0.0039  max mem: 57344
+eval (validation): [4]  [60/85]  eta: 0:00:09    time: 0.3696  data: 0.0038  max mem: 57344
+eval (validation): [4]  [80/85]  eta: 0:00:01    time: 0.3690  data: 0.0038  max mem: 57344
+eval (validation): [4]  [84/85]  eta: 0:00:00    time: 0.3629  data: 0.0038  max mem: 57344
+eval (validation): [4] Total time: 0:00:32 (0.3774 s / it)
+cv: [4]  best hparam: (4.3, 1.0) (033) ('033_lr4.3e+00_wd1.0e+00')  loss: 2.521  acc: 0.251  f1: 0.185
+saving checkpoint experiments/schaefer1000/output/schaefer1000/schaefer1000_lr3e-4_2/eval_v2/nsd_cococlip__patch__attn/checkpoint-last.pth
+saving best checkpoint experiments/schaefer1000/output/schaefer1000/schaefer1000_lr3e-4_2/eval_v2/nsd_cococlip__patch__attn/checkpoint-best.pth
+train: [5]  [  0/400]  eta: 0:08:25  lr: nan  time: 1.2631  data: 0.6625  max mem: 57344
+train: [5]  [ 20/400]  eta: 0:04:05  lr: 0.000300  loss: 2.8627 (2.8762)  grad: 0.2192 (0.2252)  time: 0.6141  data: 0.0030  max mem: 57344
+train: [5]  [ 40/400]  eta: 0:03:46  lr: 0.000300  loss: 2.8606 (2.8653)  grad: 0.2065 (0.2173)  time: 0.6152  data: 0.0037  max mem: 57344
+train: [5]  [ 60/400]  eta: 0:03:32  lr: 0.000300  loss: 2.8710 (2.8712)  grad: 0.2161 (0.2270)  time: 0.6153  data: 0.0037  max mem: 57344
+train: [5]  [ 80/400]  eta: 0:03:19  lr: 0.000300  loss: 2.8793 (2.8728)  grad: 0.2689 (0.2406)  time: 0.6146  data: 0.0037  max mem: 57344
+train: [5]  [100/400]  eta: 0:03:06  lr: 0.000300  loss: 2.9699 (2.9191)  grad: 0.3168 (0.3617)  time: 0.6141  data: 0.0037  max mem: 57344
+WARNING: classifier 41 (16, 1.0) diverged (loss=75.05 > 63.56) at step 1056. Freezing.
+train: [5]  [120/400]  eta: 0:02:53  lr: 0.000300  loss: 3.0914 (3.0014)  grad: 0.7333 (0.5513)  time: 0.6103  data: 0.0034  max mem: 57344
+train: [5]  [140/400]  eta: 0:02:40  lr: 0.000300  loss: 2.9025 (2.9813)  grad: 0.2297 (0.5045)  time: 0.6076  data: 0.0036  max mem: 57344
+train: [5]  [160/400]  eta: 0:02:27  lr: 0.000299  loss: 2.8478 (2.9633)  grad: 0.2244 (0.4688)  time: 0.6087  data: 0.0038  max mem: 57344
+train: [5]  [180/400]  eta: 0:02:15  lr: 0.000299  loss: 2.8518 (2.9517)  grad: 0.2115 (0.4408)  time: 0.6080  data: 0.0037  max mem: 57344
+train: [5]  [200/400]  eta: 0:02:02  lr: 0.000299  loss: 2.8522 (2.9413)  grad: 0.2208 (0.4197)  time: 0.6067  data: 0.0034  max mem: 57344
+train: [5]  [220/400]  eta: 0:01:50  lr: 0.000299  loss: 2.8437 (2.9329)  grad: 0.2408 (0.4067)  time: 0.6089  data: 0.0038  max mem: 57344
+train: [5]  [240/400]  eta: 0:01:38  lr: 0.000299  loss: 2.9366 (2.9564)  grad: 0.3024 (0.4582)  time: 0.6088  data: 0.0039  max mem: 57344
+WARNING: classifier 43 (22, 1.0) diverged (loss=68.59 > 63.56) at step 1122. Freezing.
+train: [5]  [260/400]  eta: 0:01:25  lr: 0.000299  loss: 2.9366 (2.9671)  grad: 0.2897 (0.4699)  time: 0.6043  data: 0.0038  max mem: 57344
+train: [5]  [280/400]  eta: 0:01:13  lr: 0.000298  loss: 2.8575 (2.9584)  grad: 0.1987 (0.4507)  time: 0.6022  data: 0.0038  max mem: 57344
+train: [5]  [300/400]  eta: 0:01:01  lr: 0.000298  loss: 2.8475 (2.9500)  grad: 0.2038 (0.4337)  time: 0.6035  data: 0.0040  max mem: 57344
+train: [5]  [320/400]  eta: 0:00:48  lr: 0.000298  loss: 2.8822 (2.9461)  grad: 0.1861 (0.4186)  time: 0.6027  data: 0.0038  max mem: 57344
+train: [5]  [340/400]  eta: 0:00:36  lr: 0.000298  loss: 2.8800 (2.9404)  grad: 0.1907 (0.4060)  time: 0.6026  data: 0.0037  max mem: 57344
+train: [5]  [360/400]  eta: 0:00:24  lr: 0.000297  loss: 2.8579 (2.9361)  grad: 0.2066 (0.3948)  time: 0.6025  data: 0.0037  max mem: 57344
+train: [5]  [380/400]  eta: 0:00:12  lr: 0.000297  loss: 2.8484 (2.9322)  grad: 0.1941 (0.3843)  time: 0.6027  data: 0.0038  max mem: 57344
+train: [5]  [399/400]  eta: 0:00:00  lr: 0.000297  loss: 2.8446 (2.9272)  grad: 0.1834 (0.3740)  time: 0.6030  data: 0.0037  max mem: 57344
+train: [5] Total time: 0:04:03 (0.6097 s / it)
+train: [5] Summary: lr: 0.000297  loss: 2.8446 (2.9272)  grad: 0.1834 (0.3740)
+eval (validation): [5]  [ 0/85]  eta: 0:01:37    time: 1.1464  data: 0.7854  max mem: 57344
+eval (validation): [5]  [20/85]  eta: 0:00:26    time: 0.3722  data: 0.0026  max mem: 57344
+eval (validation): [5]  [40/85]  eta: 0:00:17    time: 0.3694  data: 0.0038  max mem: 57344
+eval (validation): [5]  [60/85]  eta: 0:00:09    time: 0.3699  data: 0.0037  max mem: 57344
+eval (validation): [5]  [80/85]  eta: 0:00:01    time: 0.3689  data: 0.0038  max mem: 57344
+eval (validation): [5]  [84/85]  eta: 0:00:00    time: 0.3628  data: 0.0038  max mem: 57344
+eval (validation): [5] Total time: 0:00:32 (0.3790 s / it)
+cv: [5]  best hparam: (2.7, 1.0) (030) ('030_lr2.7e+00_wd1.0e+00')  loss: 2.438  acc: 0.270  f1: 0.211
+saving checkpoint experiments/schaefer1000/output/schaefer1000/schaefer1000_lr3e-4_2/eval_v2/nsd_cococlip__patch__attn/checkpoint-last.pth
+saving best checkpoint experiments/schaefer1000/output/schaefer1000/schaefer1000_lr3e-4_2/eval_v2/nsd_cococlip__patch__attn/checkpoint-best.pth
+train: [6]  [  0/400]  eta: 0:08:15  lr: nan  time: 1.2393  data: 0.6490  max mem: 57344
+train: [6]  [ 20/400]  eta: 0:04:00  lr: 0.000296  loss: 2.8223 (2.8282)  grad: 0.1881 (0.1894)  time: 0.6037  data: 0.0034  max mem: 57344
+train: [6]  [ 40/400]  eta: 0:03:42  lr: 0.000296  loss: 2.8214 (2.8146)  grad: 0.1920 (0.1927)  time: 0.6032  data: 0.0038  max mem: 57344
+train: [6]  [ 60/400]  eta: 0:03:28  lr: 0.000296  loss: 2.8139 (2.8203)  grad: 0.1910 (0.1927)  time: 0.6025  data: 0.0038  max mem: 57344
+train: [6]  [ 80/400]  eta: 0:03:15  lr: 0.000295  loss: 2.8139 (2.8165)  grad: 0.1898 (0.1937)  time: 0.6029  data: 0.0038  max mem: 57344
+train: [6]  [100/400]  eta: 0:03:02  lr: 0.000295  loss: 2.8026 (2.8145)  grad: 0.1922 (0.1943)  time: 0.6029  data: 0.0039  max mem: 57344
+train: [6]  [120/400]  eta: 0:02:50  lr: 0.000295  loss: 2.8153 (2.8172)  grad: 0.1922 (0.1944)  time: 0.6033  data: 0.0039  max mem: 57344
+train: [6]  [140/400]  eta: 0:02:37  lr: 0.000294  loss: 2.8153 (2.8132)  grad: 0.1895 (0.1931)  time: 0.6025  data: 0.0037  max mem: 57344
+train: [6]  [160/400]  eta: 0:02:25  lr: 0.000294  loss: 2.7815 (2.8088)  grad: 0.1936 (0.1941)  time: 0.6009  data: 0.0034  max mem: 57344
+train: [6]  [180/400]  eta: 0:02:13  lr: 0.000293  loss: 2.7790 (2.8092)  grad: 0.1955 (0.1945)  time: 0.6029  data: 0.0037  max mem: 57344
+train: [6]  [200/400]  eta: 0:02:01  lr: 0.000293  loss: 2.7787 (2.8074)  grad: 0.1975 (0.1958)  time: 0.6031  data: 0.0038  max mem: 57344
+train: [6]  [220/400]  eta: 0:01:49  lr: 0.000292  loss: 2.7937 (2.8076)  grad: 0.1957 (0.1954)  time: 0.6028  data: 0.0037  max mem: 57344
+train: [6]  [240/400]  eta: 0:01:36  lr: 0.000292  loss: 2.8161 (2.8103)  grad: 0.1931 (0.1957)  time: 0.6011  data: 0.0034  max mem: 57344
+train: [6]  [260/400]  eta: 0:01:24  lr: 0.000291  loss: 2.8316 (2.8113)  grad: 0.1933 (0.1958)  time: 0.6043  data: 0.0041  max mem: 57344
+train: [6]  [280/400]  eta: 0:01:12  lr: 0.000291  loss: 2.7965 (2.8109)  grad: 0.1921 (0.1953)  time: 0.6034  data: 0.0041  max mem: 57344
+train: [6]  [300/400]  eta: 0:01:00  lr: 0.000290  loss: 2.7874 (2.8113)  grad: 0.1902 (0.1954)  time: 0.6031  data: 0.0039  max mem: 57344
+train: [6]  [320/400]  eta: 0:00:48  lr: 0.000290  loss: 2.8005 (2.8125)  grad: 0.1902 (0.1954)  time: 0.6032  data: 0.0039  max mem: 57344
+train: [6]  [340/400]  eta: 0:00:36  lr: 0.000289  loss: 2.7979 (2.8117)  grad: 0.2005 (0.1960)  time: 0.6038  data: 0.0040  max mem: 57344
+train: [6]  [360/400]  eta: 0:00:24  lr: 0.000288  loss: 2.7839 (2.8109)  grad: 0.2026 (0.1962)  time: 0.6031  data: 0.0039  max mem: 57344
+train: [6]  [380/400]  eta: 0:00:12  lr: 0.000288  loss: 2.7665 (2.8110)  grad: 0.2006 (0.1966)  time: 0.6031  data: 0.0039  max mem: 57344
+train: [6]  [399/400]  eta: 0:00:00  lr: 0.000287  loss: 2.7808 (2.8102)  grad: 0.2000 (0.1968)  time: 0.6035  data: 0.0039  max mem: 57344
+train: [6] Total time: 0:04:01 (0.6048 s / it)
+train: [6] Summary: lr: 0.000287  loss: 2.7808 (2.8102)  grad: 0.2000 (0.1968)
+eval (validation): [6]  [ 0/85]  eta: 0:01:26    time: 1.0217  data: 0.6611  max mem: 57344
+eval (validation): [6]  [20/85]  eta: 0:00:26    time: 0.3690  data: 0.0029  max mem: 57344
+eval (validation): [6]  [40/85]  eta: 0:00:17    time: 0.3696  data: 0.0039  max mem: 57344
+eval (validation): [6]  [60/85]  eta: 0:00:09    time: 0.3697  data: 0.0039  max mem: 57344
+eval (validation): [6]  [80/85]  eta: 0:00:01    time: 0.3696  data: 0.0038  max mem: 57344
+eval (validation): [6]  [84/85]  eta: 0:00:00    time: 0.3631  data: 0.0037  max mem: 57344
+eval (validation): [6] Total time: 0:00:32 (0.3770 s / it)
+cv: [6]  best hparam: (3.1, 1.0) (031) ('031_lr3.1e+00_wd1.0e+00')  loss: 2.456  acc: 0.264  f1: 0.210
+saving checkpoint experiments/schaefer1000/output/schaefer1000/schaefer1000_lr3e-4_2/eval_v2/nsd_cococlip__patch__attn/checkpoint-last.pth
+train: [7]  [  0/400]  eta: 0:08:27  lr: nan  time: 1.2682  data: 0.6777  max mem: 57344
+train: [7]  [ 20/400]  eta: 0:04:01  lr: 0.000286  loss: 2.7487 (2.7381)  grad: 0.1965 (0.1952)  time: 0.6032  data: 0.0032  max mem: 57344
+train: [7]  [ 40/400]  eta: 0:03:42  lr: 0.000286  loss: 2.7526 (2.7546)  grad: 0.1975 (0.1973)  time: 0.6025  data: 0.0039  max mem: 57344
+train: [7]  [ 60/400]  eta: 0:03:28  lr: 0.000285  loss: 2.7568 (2.7579)  grad: 0.1979 (0.1989)  time: 0.6037  data: 0.0039  max mem: 57344
+train: [7]  [ 80/400]  eta: 0:03:15  lr: 0.000284  loss: 2.7519 (2.7574)  grad: 0.1979 (0.1994)  time: 0.6029  data: 0.0038  max mem: 57344
+train: [7]  [100/400]  eta: 0:03:02  lr: 0.000284  loss: 2.7503 (2.7611)  grad: 0.1990 (0.1995)  time: 0.6035  data: 0.0038  max mem: 57344
+train: [7]  [120/400]  eta: 0:02:50  lr: 0.000283  loss: 2.7425 (2.7578)  grad: 0.2037 (0.2014)  time: 0.6030  data: 0.0038  max mem: 57344
+train: [7]  [140/400]  eta: 0:02:38  lr: 0.000282  loss: 2.7746 (2.7626)  grad: 0.2100 (0.2028)  time: 0.6033  data: 0.0039  max mem: 57344
+train: [7]  [160/400]  eta: 0:02:25  lr: 0.000282  loss: 2.7631 (2.7621)  grad: 0.1910 (0.2012)  time: 0.6034  data: 0.0038  max mem: 57344
+train: [7]  [180/400]  eta: 0:02:13  lr: 0.000281  loss: 2.7461 (2.7605)  grad: 0.1878 (0.2006)  time: 0.6034  data: 0.0038  max mem: 57344
+train: [7]  [200/400]  eta: 0:02:01  lr: 0.000280  loss: 2.7351 (2.7572)  grad: 0.1897 (0.1999)  time: 0.6024  data: 0.0036  max mem: 57344
+train: [7]  [220/400]  eta: 0:01:49  lr: 0.000279  loss: 2.7511 (2.7581)  grad: 0.1957 (0.2006)  time: 0.6020  data: 0.0035  max mem: 57344
+train: [7]  [240/400]  eta: 0:01:36  lr: 0.000278  loss: 2.7550 (2.7567)  grad: 0.1991 (0.2003)  time: 0.6008  data: 0.0033  max mem: 57344
+train: [7]  [260/400]  eta: 0:01:24  lr: 0.000278  loss: 2.7818 (2.7609)  grad: 0.2039 (0.2012)  time: 0.6025  data: 0.0037  max mem: 57344
+train: [7]  [280/400]  eta: 0:01:12  lr: 0.000277  loss: 2.8096 (2.7644)  grad: 0.2075 (0.2017)  time: 0.6024  data: 0.0037  max mem: 57344
+train: [7]  [300/400]  eta: 0:01:00  lr: 0.000276  loss: 2.7778 (2.7634)  grad: 0.2116 (0.2024)  time: 0.6022  data: 0.0036  max mem: 57344
+train: [7]  [320/400]  eta: 0:00:48  lr: 0.000275  loss: 2.7560 (2.7629)  grad: 0.2140 (0.2034)  time: 0.6012  data: 0.0035  max mem: 57344
+train: [7]  [340/400]  eta: 0:00:36  lr: 0.000274  loss: 2.7925 (2.7653)  grad: 0.2114 (0.2039)  time: 0.6028  data: 0.0036  max mem: 57344
+train: [7]  [360/400]  eta: 0:00:24  lr: 0.000273  loss: 2.7925 (2.7662)  grad: 0.2081 (0.2039)  time: 0.6023  data: 0.0037  max mem: 57344
+train: [7]  [380/400]  eta: 0:00:12  lr: 0.000272  loss: 2.7855 (2.7687)  grad: 0.2034 (0.2038)  time: 0.6030  data: 0.0039  max mem: 57344
+train: [7]  [399/400]  eta: 0:00:00  lr: 0.000271  loss: 2.8314 (2.7722)  grad: 0.2034 (0.2038)  time: 0.6028  data: 0.0039  max mem: 57344
+train: [7] Total time: 0:04:01 (0.6046 s / it)
+train: [7] Summary: lr: 0.000271  loss: 2.8314 (2.7722)  grad: 0.2034 (0.2038)
+eval (validation): [7]  [ 0/85]  eta: 0:01:23    time: 0.9821  data: 0.6225  max mem: 57344
+eval (validation): [7]  [20/85]  eta: 0:00:25    time: 0.3678  data: 0.0027  max mem: 57344
+eval (validation): [7]  [40/85]  eta: 0:00:17    time: 0.3691  data: 0.0035  max mem: 57344
+eval (validation): [7]  [60/85]  eta: 0:00:09    time: 0.3687  data: 0.0036  max mem: 57344
+eval (validation): [7]  [80/85]  eta: 0:00:01    time: 0.3690  data: 0.0036  max mem: 57344
+eval (validation): [7]  [84/85]  eta: 0:00:00    time: 0.3631  data: 0.0037  max mem: 57344
+eval (validation): [7] Total time: 0:00:31 (0.3757 s / it)
+cv: [7]  best hparam: (4.3, 1.0) (033) ('033_lr4.3e+00_wd1.0e+00')  loss: 2.451  acc: 0.262  f1: 0.197
+saving checkpoint experiments/schaefer1000/output/schaefer1000/schaefer1000_lr3e-4_2/eval_v2/nsd_cococlip__patch__attn/checkpoint-last.pth
+train: [8]  [  0/400]  eta: 0:08:05  lr: nan  time: 1.2142  data: 0.6237  max mem: 57344
+train: [8]  [ 20/400]  eta: 0:04:00  lr: 0.000270  loss: 2.7132 (2.7304)  grad: 0.1952 (0.1973)  time: 0.6030  data: 0.0027  max mem: 57344
+train: [8]  [ 40/400]  eta: 0:03:42  lr: 0.000270  loss: 2.7185 (2.7256)  grad: 0.1978 (0.1969)  time: 0.6041  data: 0.0037  max mem: 57344
+train: [8]  [ 60/400]  eta: 0:03:28  lr: 0.000269  loss: 2.7263 (2.7361)  grad: 0.1989 (0.2001)  time: 0.6046  data: 0.0039  max mem: 57344
+train: [8]  [ 80/400]  eta: 0:03:15  lr: 0.000268  loss: 2.7277 (2.7329)  grad: 0.2019 (0.2019)  time: 0.6042  data: 0.0037  max mem: 57344
+train: [8]  [100/400]  eta: 0:03:02  lr: 0.000267  loss: 2.7261 (2.7325)  grad: 0.1986 (0.2015)  time: 0.6039  data: 0.0036  max mem: 57344
+train: [8]  [120/400]  eta: 0:02:50  lr: 0.000266  loss: 2.7469 (2.7376)  grad: 0.1986 (0.2018)  time: 0.6040  data: 0.0037  max mem: 57344
+train: [8]  [140/400]  eta: 0:02:38  lr: 0.000265  loss: 2.7551 (2.7376)  grad: 0.2080 (0.2028)  time: 0.6046  data: 0.0039  max mem: 57344
+train: [8]  [160/400]  eta: 0:02:25  lr: 0.000264  loss: 2.7482 (2.7409)  grad: 0.2108 (0.2040)  time: 0.6048  data: 0.0039  max mem: 57344
+train: [8]  [180/400]  eta: 0:02:13  lr: 0.000263  loss: 2.7268 (2.7406)  grad: 0.2037 (0.2038)  time: 0.6045  data: 0.0038  max mem: 57344
+train: [8]  [200/400]  eta: 0:02:01  lr: 0.000262  loss: 2.7268 (2.7420)  grad: 0.2037 (0.2043)  time: 0.6041  data: 0.0038  max mem: 57344
+train: [8]  [220/400]  eta: 0:01:49  lr: 0.000260  loss: 2.7071 (2.7396)  grad: 0.2006 (0.2040)  time: 0.6043  data: 0.0038  max mem: 57344
+train: [8]  [240/400]  eta: 0:01:37  lr: 0.000259  loss: 2.7135 (2.7394)  grad: 0.1994 (0.2037)  time: 0.6045  data: 0.0036  max mem: 57344
+train: [8]  [260/400]  eta: 0:01:24  lr: 0.000258  loss: 2.7529 (2.7400)  grad: 0.1994 (0.2041)  time: 0.6044  data: 0.0037  max mem: 57344
+train: [8]  [280/400]  eta: 0:01:12  lr: 0.000257  loss: 2.7509 (2.7401)  grad: 0.2000 (0.2037)  time: 0.6051  data: 0.0038  max mem: 57344
+train: [8]  [300/400]  eta: 0:01:00  lr: 0.000256  loss: 2.7257 (2.7386)  grad: 0.2000 (0.2041)  time: 0.6044  data: 0.0037  max mem: 57344
+train: [8]  [320/400]  eta: 0:00:48  lr: 0.000255  loss: 2.7536 (2.7390)  grad: 0.2062 (0.2045)  time: 0.6044  data: 0.0037  max mem: 57344
+train: [8]  [340/400]  eta: 0:00:36  lr: 0.000254  loss: 2.7536 (2.7390)  grad: 0.2054 (0.2045)  time: 0.6046  data: 0.0036  max mem: 57344
+train: [8]  [360/400]  eta: 0:00:24  lr: 0.000253  loss: 2.7138 (2.7384)  grad: 0.2040 (0.2044)  time: 0.6038  data: 0.0036  max mem: 57344
+train: [8]  [380/400]  eta: 0:00:12  lr: 0.000252  loss: 2.7025 (2.7376)  grad: 0.1982 (0.2042)  time: 0.6042  data: 0.0037  max mem: 57344
+train: [8]  [399/400]  eta: 0:00:00  lr: 0.000250  loss: 2.6994 (2.7356)  grad: 0.1969 (0.2043)  time: 0.6032  data: 0.0034  max mem: 57344
+train: [8] Total time: 0:04:02 (0.6060 s / it)
+train: [8] Summary: lr: 0.000250  loss: 2.6994 (2.7356)  grad: 0.1969 (0.2043)
+eval (validation): [8]  [ 0/85]  eta: 0:01:08    time: 0.8100  data: 0.4512  max mem: 57344
+eval (validation): [8]  [20/85]  eta: 0:00:25    time: 0.3695  data: 0.0028  max mem: 57344
+eval (validation): [8]  [40/85]  eta: 0:00:17    time: 0.3715  data: 0.0037  max mem: 57344
+eval (validation): [8]  [60/85]  eta: 0:00:09    time: 0.3716  data: 0.0039  max mem: 57344
+eval (validation): [8]  [80/85]  eta: 0:00:01    time: 0.3714  data: 0.0037  max mem: 57344
+eval (validation): [8]  [84/85]  eta: 0:00:00    time: 0.3655  data: 0.0039  max mem: 57344
+eval (validation): [8] Total time: 0:00:31 (0.3762 s / it)
+cv: [8]  best hparam: (6, 1.0) (035) ('035_lr6.0e+00_wd1.0e+00')  loss: 2.503  acc: 0.265  f1: 0.206
+saving checkpoint experiments/schaefer1000/output/schaefer1000/schaefer1000_lr3e-4_2/eval_v2/nsd_cococlip__patch__attn/checkpoint-last.pth
+train: [9]  [  0/400]  eta: 0:08:48  lr: nan  time: 1.3211  data: 0.7296  max mem: 57344
+train: [9]  [ 20/400]  eta: 0:04:02  lr: 0.000249  loss: 2.7310 (2.7150)  grad: 0.1981 (0.1983)  time: 0.6041  data: 0.0038  max mem: 57344
+train: [9]  [ 40/400]  eta: 0:03:43  lr: 0.000248  loss: 2.7230 (2.6939)  grad: 0.1981 (0.1981)  time: 0.6028  data: 0.0036  max mem: 57344
+train: [9]  [ 60/400]  eta: 0:03:29  lr: 0.000247  loss: 2.7230 (2.7044)  grad: 0.1969 (0.1977)  time: 0.6024  data: 0.0037  max mem: 57344
+train: [9]  [ 80/400]  eta: 0:03:15  lr: 0.000246  loss: 2.7242 (2.7169)  grad: 0.2004 (0.1995)  time: 0.6030  data: 0.0040  max mem: 57344
+train: [9]  [100/400]  eta: 0:03:03  lr: 0.000244  loss: 2.6916 (2.7142)  grad: 0.2004 (0.1991)  time: 0.6034  data: 0.0041  max mem: 57344
+train: [9]  [120/400]  eta: 0:02:50  lr: 0.000243  loss: 2.6881 (2.7123)  grad: 0.1963 (0.1987)  time: 0.6026  data: 0.0037  max mem: 57344
+train: [9]  [140/400]  eta: 0:02:38  lr: 0.000242  loss: 2.6568 (2.7046)  grad: 0.2039 (0.1998)  time: 0.6023  data: 0.0037  max mem: 57344
+train: [9]  [160/400]  eta: 0:02:25  lr: 0.000241  loss: 2.6744 (2.7083)  grad: 0.2062 (0.2000)  time: 0.6028  data: 0.0038  max mem: 57344
+train: [9]  [180/400]  eta: 0:02:13  lr: 0.000240  loss: 2.7312 (2.7133)  grad: 0.2034 (0.2001)  time: 0.6023  data: 0.0038  max mem: 57344
+train: [9]  [200/400]  eta: 0:02:01  lr: 0.000238  loss: 2.7010 (2.7124)  grad: 0.2000 (0.1998)  time: 0.6009  data: 0.0033  max mem: 57344
+train: [9]  [220/400]  eta: 0:01:49  lr: 0.000237  loss: 2.7010 (2.7148)  grad: 0.1979 (0.1999)  time: 0.5997  data: 0.0031  max mem: 57344
+train: [9]  [240/400]  eta: 0:01:36  lr: 0.000236  loss: 2.7358 (2.7165)  grad: 0.2012 (0.2003)  time: 0.5998  data: 0.0032  max mem: 57344
+train: [9]  [260/400]  eta: 0:01:24  lr: 0.000234  loss: 2.7399 (2.7178)  grad: 0.1976 (0.2000)  time: 0.6000  data: 0.0032  max mem: 57344
+train: [9]  [280/400]  eta: 0:01:12  lr: 0.000233  loss: 2.6883 (2.7168)  grad: 0.1882 (0.1990)  time: 0.5994  data: 0.0032  max mem: 57344
+train: [9]  [300/400]  eta: 0:01:00  lr: 0.000232  loss: 2.6883 (2.7162)  grad: 0.1940 (0.1996)  time: 0.5999  data: 0.0032  max mem: 57344
+train: [9]  [320/400]  eta: 0:00:48  lr: 0.000230  loss: 2.7108 (2.7161)  grad: 0.2072 (0.1998)  time: 0.6002  data: 0.0032  max mem: 57344
+train: [9]  [340/400]  eta: 0:00:36  lr: 0.000229  loss: 2.7260 (2.7177)  grad: 0.2071 (0.2003)  time: 0.5997  data: 0.0032  max mem: 57344
+train: [9]  [360/400]  eta: 0:00:24  lr: 0.000228  loss: 2.6971 (2.7143)  grad: 0.1963 (0.2002)  time: 0.5996  data: 0.0032  max mem: 57344
+train: [9]  [380/400]  eta: 0:00:12  lr: 0.000226  loss: 2.6695 (2.7128)  grad: 0.1950 (0.2000)  time: 0.5999  data: 0.0032  max mem: 57344
+train: [9]  [399/400]  eta: 0:00:00  lr: 0.000225  loss: 2.7051 (2.7132)  grad: 0.1963 (0.1998)  time: 0.5995  data: 0.0031  max mem: 57344
+train: [9] Total time: 0:04:01 (0.6033 s / it)
+train: [9] Summary: lr: 0.000225  loss: 2.7051 (2.7132)  grad: 0.1963 (0.1998)
+eval (validation): [9]  [ 0/85]  eta: 0:01:06    time: 0.7843  data: 0.4245  max mem: 57344
+eval (validation): [9]  [20/85]  eta: 0:00:25    time: 0.3674  data: 0.0028  max mem: 57344
+eval (validation): [9]  [40/85]  eta: 0:00:17    time: 0.3681  data: 0.0031  max mem: 57344
+eval (validation): [9]  [60/85]  eta: 0:00:09    time: 0.3678  data: 0.0032  max mem: 57344
+eval (validation): [9]  [80/85]  eta: 0:00:01    time: 0.3683  data: 0.0032  max mem: 57344
+eval (validation): [9]  [84/85]  eta: 0:00:00    time: 0.3619  data: 0.0031  max mem: 57344
+eval (validation): [9] Total time: 0:00:31 (0.3725 s / it)
+cv: [9]  best hparam: (12, 1.0) (039) ('039_lr1.2e+01_wd1.0e+00')  loss: 2.481  acc: 0.276  f1: 0.220
+saving checkpoint experiments/schaefer1000/output/schaefer1000/schaefer1000_lr3e-4_2/eval_v2/nsd_cococlip__patch__attn/checkpoint-last.pth
+saving best checkpoint experiments/schaefer1000/output/schaefer1000/schaefer1000_lr3e-4_2/eval_v2/nsd_cococlip__patch__attn/checkpoint-best.pth
+train: [10]  [  0/400]  eta: 0:07:14  lr: nan  time: 1.0862  data: 0.4954  max mem: 57344
+train: [10]  [ 20/400]  eta: 0:03:57  lr: 0.000224  loss: 2.6897 (2.6799)  grad: 0.1858 (0.1916)  time: 0.6011  data: 0.0025  max mem: 57344
+train: [10]  [ 40/400]  eta: 0:03:40  lr: 0.000222  loss: 2.6659 (2.6666)  grad: 0.1995 (0.1994)  time: 0.6018  data: 0.0035  max mem: 57344
+train: [10]  [ 60/400]  eta: 0:03:27  lr: 0.000221  loss: 2.6620 (2.6701)  grad: 0.2026 (0.2021)  time: 0.6033  data: 0.0037  max mem: 57344
+train: [10]  [ 80/400]  eta: 0:03:14  lr: 0.000220  loss: 2.6892 (2.6738)  grad: 0.2012 (0.2014)  time: 0.6029  data: 0.0039  max mem: 57344
+train: [10]  [100/400]  eta: 0:03:02  lr: 0.000218  loss: 2.6773 (2.6711)  grad: 0.2012 (0.2009)  time: 0.6044  data: 0.0040  max mem: 57344
+train: [10]  [120/400]  eta: 0:02:50  lr: 0.000217  loss: 2.6774 (2.6788)  grad: 0.1978 (0.2010)  time: 0.6071  data: 0.0045  max mem: 57344
+train: [10]  [140/400]  eta: 0:02:38  lr: 0.000215  loss: 2.6733 (2.6745)  grad: 0.2005 (0.2016)  time: 0.6128  data: 0.0048  max mem: 57344
+train: [10]  [160/400]  eta: 0:02:25  lr: 0.000214  loss: 2.6372 (2.6743)  grad: 0.2090 (0.2032)  time: 0.6036  data: 0.0037  max mem: 57344
+train: [10]  [180/400]  eta: 0:02:13  lr: 0.000213  loss: 2.6687 (2.6751)  grad: 0.2127 (0.2036)  time: 0.6007  data: 0.0034  max mem: 57344
+train: [10]  [200/400]  eta: 0:02:01  lr: 0.000211  loss: 2.6591 (2.6734)  grad: 0.2040 (0.2038)  time: 0.6002  data: 0.0034  max mem: 57344
+train: [10]  [220/400]  eta: 0:01:49  lr: 0.000210  loss: 2.6543 (2.6732)  grad: 0.2030 (0.2038)  time: 0.5997  data: 0.0032  max mem: 57344
+train: [10]  [240/400]  eta: 0:01:36  lr: 0.000208  loss: 2.6478 (2.6722)  grad: 0.2030 (0.2035)  time: 0.6010  data: 0.0035  max mem: 57344
+train: [10]  [260/400]  eta: 0:01:24  lr: 0.000207  loss: 2.6763 (2.6739)  grad: 0.2037 (0.2036)  time: 0.5999  data: 0.0033  max mem: 57344
+train: [10]  [280/400]  eta: 0:01:12  lr: 0.000205  loss: 2.6811 (2.6767)  grad: 0.2025 (0.2029)  time: 0.5993  data: 0.0032  max mem: 57344
+train: [10]  [300/400]  eta: 0:01:00  lr: 0.000204  loss: 2.7384 (2.6792)  grad: 0.1924 (0.2028)  time: 0.5996  data: 0.0033  max mem: 57344
+train: [10]  [320/400]  eta: 0:00:48  lr: 0.000202  loss: 2.7145 (2.6797)  grad: 0.1924 (0.2024)  time: 0.5998  data: 0.0033  max mem: 57344
+train: [10]  [340/400]  eta: 0:00:36  lr: 0.000201  loss: 2.6981 (2.6814)  grad: 0.1948 (0.2020)  time: 0.5999  data: 0.0033  max mem: 57344
+train: [10]  [360/400]  eta: 0:00:24  lr: 0.000199  loss: 2.6764 (2.6795)  grad: 0.1943 (0.2015)  time: 0.6030  data: 0.0038  max mem: 57344
+train: [10]  [380/400]  eta: 0:00:12  lr: 0.000198  loss: 2.6572 (2.6796)  grad: 0.1989 (0.2018)  time: 0.6027  data: 0.0038  max mem: 57344
+train: [10]  [399/400]  eta: 0:00:00  lr: 0.000196  loss: 2.6785 (2.6799)  grad: 0.1996 (0.2016)  time: 0.6019  data: 0.0037  max mem: 57344
+train: [10] Total time: 0:04:01 (0.6037 s / it)
+train: [10] Summary: lr: 0.000196  loss: 2.6785 (2.6799)  grad: 0.1996 (0.2016)
+eval (validation): [10]  [ 0/85]  eta: 0:01:20    time: 0.9469  data: 0.5863  max mem: 57344
+eval (validation): [10]  [20/85]  eta: 0:00:25    time: 0.3714  data: 0.0049  max mem: 57344
+eval (validation): [10]  [40/85]  eta: 0:00:17    time: 0.3699  data: 0.0039  max mem: 57344
+eval (validation): [10]  [60/85]  eta: 0:00:09    time: 0.3704  data: 0.0037  max mem: 57344
+eval (validation): [10]  [80/85]  eta: 0:00:01    time: 0.3691  data: 0.0037  max mem: 57344
+eval (validation): [10]  [84/85]  eta: 0:00:00    time: 0.3624  data: 0.0037  max mem: 57344
+eval (validation): [10] Total time: 0:00:32 (0.3767 s / it)
+cv: [10]  best hparam: (1.6, 1.0) (027) ('027_lr1.6e+00_wd1.0e+00')  loss: 2.385  acc: 0.281  f1: 0.220
+saving checkpoint experiments/schaefer1000/output/schaefer1000/schaefer1000_lr3e-4_2/eval_v2/nsd_cococlip__patch__attn/checkpoint-last.pth
+saving best checkpoint experiments/schaefer1000/output/schaefer1000/schaefer1000_lr3e-4_2/eval_v2/nsd_cococlip__patch__attn/checkpoint-best.pth
+train: [11]  [  0/400]  eta: 0:08:28  lr: nan  time: 1.2704  data: 0.6803  max mem: 57344
+train: [11]  [ 20/400]  eta: 0:04:01  lr: 0.000195  loss: 2.6436 (2.6518)  grad: 0.1991 (0.2089)  time: 0.6031  data: 0.0030  max mem: 57344
+train: [11]  [ 40/400]  eta: 0:03:43  lr: 0.000193  loss: 2.6373 (2.6449)  grad: 0.1915 (0.2043)  time: 0.6058  data: 0.0047  max mem: 57344
+train: [11]  [ 60/400]  eta: 0:03:29  lr: 0.000192  loss: 2.6510 (2.6516)  grad: 0.1875 (0.1989)  time: 0.6067  data: 0.0052  max mem: 57344
+train: [11]  [ 80/400]  eta: 0:03:16  lr: 0.000190  loss: 2.6775 (2.6560)  grad: 0.1893 (0.1978)  time: 0.6053  data: 0.0038  max mem: 57344
+train: [11]  [100/400]  eta: 0:03:03  lr: 0.000189  loss: 2.6854 (2.6676)  grad: 0.1960 (0.1991)  time: 0.6055  data: 0.0035  max mem: 57344
+train: [11]  [120/400]  eta: 0:02:50  lr: 0.000187  loss: 2.7230 (2.6732)  grad: 0.1995 (0.1996)  time: 0.6046  data: 0.0035  max mem: 57344
+train: [11]  [140/400]  eta: 0:02:38  lr: 0.000186  loss: 2.6782 (2.6703)  grad: 0.1944 (0.1977)  time: 0.6029  data: 0.0036  max mem: 57344
+train: [11]  [160/400]  eta: 0:02:26  lr: 0.000184  loss: 2.6609 (2.6655)  grad: 0.1831 (0.1968)  time: 0.6036  data: 0.0036  max mem: 57344
+train: [11]  [180/400]  eta: 0:02:13  lr: 0.000183  loss: 2.5966 (2.6622)  grad: 0.1936 (0.1974)  time: 0.6032  data: 0.0036  max mem: 57344
+train: [11]  [200/400]  eta: 0:02:01  lr: 0.000181  loss: 2.6104 (2.6618)  grad: 0.2006 (0.1980)  time: 0.6033  data: 0.0035  max mem: 57344
+train: [11]  [220/400]  eta: 0:01:49  lr: 0.000180  loss: 2.6872 (2.6613)  grad: 0.1968 (0.1979)  time: 0.6033  data: 0.0036  max mem: 57344
+train: [11]  [240/400]  eta: 0:01:37  lr: 0.000178  loss: 2.6899 (2.6634)  grad: 0.1977 (0.1990)  time: 0.6034  data: 0.0036  max mem: 57344
+train: [11]  [260/400]  eta: 0:01:24  lr: 0.000177  loss: 2.6704 (2.6625)  grad: 0.2007 (0.1990)  time: 0.6036  data: 0.0035  max mem: 57344
+train: [11]  [280/400]  eta: 0:01:12  lr: 0.000175  loss: 2.6325 (2.6606)  grad: 0.2006 (0.1992)  time: 0.6023  data: 0.0034  max mem: 57344
+train: [11]  [300/400]  eta: 0:01:00  lr: 0.000174  loss: 2.6924 (2.6605)  grad: 0.2007 (0.1995)  time: 0.6018  data: 0.0034  max mem: 57344
+train: [11]  [320/400]  eta: 0:00:48  lr: 0.000172  loss: 2.6428 (2.6587)  grad: 0.2003 (0.1996)  time: 0.6017  data: 0.0033  max mem: 57344
+train: [11]  [340/400]  eta: 0:00:36  lr: 0.000170  loss: 2.6428 (2.6604)  grad: 0.2003 (0.1997)  time: 0.6015  data: 0.0033  max mem: 57344
+train: [11]  [360/400]  eta: 0:00:24  lr: 0.000169  loss: 2.6317 (2.6592)  grad: 0.2047 (0.1999)  time: 0.6015  data: 0.0033  max mem: 57344
+train: [11]  [380/400]  eta: 0:00:12  lr: 0.000167  loss: 2.6691 (2.6599)  grad: 0.2020 (0.2000)  time: 0.6017  data: 0.0032  max mem: 57344
+train: [11]  [399/400]  eta: 0:00:00  lr: 0.000166  loss: 2.6976 (2.6617)  grad: 0.2022 (0.2003)  time: 0.6017  data: 0.0033  max mem: 57344
+train: [11] Total time: 0:04:02 (0.6053 s / it)
+train: [11] Summary: lr: 0.000166  loss: 2.6976 (2.6617)  grad: 0.2022 (0.2003)
+eval (validation): [11]  [ 0/85]  eta: 0:01:08    time: 0.8033  data: 0.4418  max mem: 57344
+eval (validation): [11]  [20/85]  eta: 0:00:25    time: 0.3678  data: 0.0028  max mem: 57344
+eval (validation): [11]  [40/85]  eta: 0:00:17    time: 0.3693  data: 0.0034  max mem: 57344
+eval (validation): [11]  [60/85]  eta: 0:00:09    time: 0.3699  data: 0.0036  max mem: 57344
+eval (validation): [11]  [80/85]  eta: 0:00:01    time: 0.3694  data: 0.0036  max mem: 57344
+eval (validation): [11]  [84/85]  eta: 0:00:00    time: 0.3633  data: 0.0036  max mem: 57344
+eval (validation): [11] Total time: 0:00:31 (0.3740 s / it)
+cv: [11]  best hparam: (1.2, 1.0) (025) ('025_lr1.2e+00_wd1.0e+00')  loss: 2.397  acc: 0.275  f1: 0.221
+saving checkpoint experiments/schaefer1000/output/schaefer1000/schaefer1000_lr3e-4_2/eval_v2/nsd_cococlip__patch__attn/checkpoint-last.pth
+train: [12]  [  0/400]  eta: 0:08:34  lr: nan  time: 1.2860  data: 0.6965  max mem: 57344
+train: [12]  [ 20/400]  eta: 0:04:01  lr: 0.000164  loss: 2.6632 (2.6806)  grad: 0.1951 (0.1987)  time: 0.6023  data: 0.0034  max mem: 57344
+train: [12]  [ 40/400]  eta: 0:03:42  lr: 0.000163  loss: 2.6632 (2.6608)  grad: 0.1972 (0.1977)  time: 0.6012  data: 0.0035  max mem: 57344
+train: [12]  [ 60/400]  eta: 0:03:28  lr: 0.000161  loss: 2.6650 (2.6549)  grad: 0.1940 (0.1955)  time: 0.6008  data: 0.0034  max mem: 57344
+train: [12]  [ 80/400]  eta: 0:03:15  lr: 0.000160  loss: 2.6413 (2.6458)  grad: 0.1922 (0.1955)  time: 0.6011  data: 0.0035  max mem: 57344
+train: [12]  [100/400]  eta: 0:03:02  lr: 0.000158  loss: 2.6121 (2.6423)  grad: 0.1958 (0.1967)  time: 0.6010  data: 0.0035  max mem: 57344
+train: [12]  [120/400]  eta: 0:02:50  lr: 0.000156  loss: 2.6557 (2.6506)  grad: 0.2064 (0.1994)  time: 0.6031  data: 0.0039  max mem: 57344
+train: [12]  [140/400]  eta: 0:02:37  lr: 0.000155  loss: 2.6859 (2.6518)  grad: 0.2064 (0.2000)  time: 0.6028  data: 0.0039  max mem: 57344
+train: [12]  [160/400]  eta: 0:02:25  lr: 0.000153  loss: 2.6061 (2.6454)  grad: 0.2031 (0.1999)  time: 0.6014  data: 0.0037  max mem: 57344
+train: [12]  [180/400]  eta: 0:02:13  lr: 0.000152  loss: 2.6118 (2.6425)  grad: 0.2014 (0.2005)  time: 0.6024  data: 0.0039  max mem: 57344
+train: [12]  [200/400]  eta: 0:02:01  lr: 0.000150  loss: 2.6180 (2.6446)  grad: 0.2079 (0.2016)  time: 0.6032  data: 0.0040  max mem: 57344
+train: [12]  [220/400]  eta: 0:01:48  lr: 0.000149  loss: 2.6170 (2.6406)  grad: 0.2046 (0.2018)  time: 0.6028  data: 0.0038  max mem: 57344
+train: [12]  [240/400]  eta: 0:01:36  lr: 0.000147  loss: 2.5944 (2.6392)  grad: 0.1946 (0.2008)  time: 0.6022  data: 0.0037  max mem: 57344
+train: [12]  [260/400]  eta: 0:01:24  lr: 0.000145  loss: 2.6172 (2.6373)  grad: 0.1901 (0.1999)  time: 0.6014  data: 0.0036  max mem: 57344
+train: [12]  [280/400]  eta: 0:01:12  lr: 0.000144  loss: 2.6226 (2.6390)  grad: 0.1924 (0.1999)  time: 0.6015  data: 0.0036  max mem: 57344
+train: [12]  [300/400]  eta: 0:01:00  lr: 0.000142  loss: 2.6400 (2.6407)  grad: 0.1971 (0.2001)  time: 0.6013  data: 0.0036  max mem: 57344
+train: [12]  [320/400]  eta: 0:00:48  lr: 0.000141  loss: 2.6285 (2.6400)  grad: 0.2009 (0.2002)  time: 0.6014  data: 0.0036  max mem: 57344
+train: [12]  [340/400]  eta: 0:00:36  lr: 0.000139  loss: 2.6297 (2.6410)  grad: 0.1972 (0.2001)  time: 0.6014  data: 0.0035  max mem: 57344
+train: [12]  [360/400]  eta: 0:00:24  lr: 0.000138  loss: 2.6006 (2.6384)  grad: 0.1972 (0.2004)  time: 0.6002  data: 0.0033  max mem: 57344
+train: [12]  [380/400]  eta: 0:00:12  lr: 0.000136  loss: 2.5911 (2.6395)  grad: 0.2025 (0.2005)  time: 0.6005  data: 0.0034  max mem: 57344
+train: [12]  [399/400]  eta: 0:00:00  lr: 0.000134  loss: 2.6261 (2.6378)  grad: 0.1996 (0.2004)  time: 0.6002  data: 0.0033  max mem: 57344
+train: [12] Total time: 0:04:01 (0.6036 s / it)
+train: [12] Summary: lr: 0.000134  loss: 2.6261 (2.6378)  grad: 0.1996 (0.2004)
+eval (validation): [12]  [ 0/85]  eta: 0:01:12    time: 0.8515  data: 0.4925  max mem: 57344
+eval (validation): [12]  [20/85]  eta: 0:00:25    time: 0.3675  data: 0.0022  max mem: 57344
+eval (validation): [12]  [40/85]  eta: 0:00:17    time: 0.3675  data: 0.0030  max mem: 57344
+eval (validation): [12]  [60/85]  eta: 0:00:09    time: 0.3679  data: 0.0033  max mem: 57344
+eval (validation): [12]  [80/85]  eta: 0:00:01    time: 0.3678  data: 0.0033  max mem: 57344
+eval (validation): [12]  [84/85]  eta: 0:00:00    time: 0.3614  data: 0.0033  max mem: 57344
+eval (validation): [12] Total time: 0:00:31 (0.3730 s / it)
+cv: [12]  best hparam: (3.1, 1.0) (031) ('031_lr3.1e+00_wd1.0e+00')  loss: 2.423  acc: 0.285  f1: 0.228
+saving checkpoint experiments/schaefer1000/output/schaefer1000/schaefer1000_lr3e-4_2/eval_v2/nsd_cococlip__patch__attn/checkpoint-last.pth
+saving best checkpoint experiments/schaefer1000/output/schaefer1000/schaefer1000_lr3e-4_2/eval_v2/nsd_cococlip__patch__attn/checkpoint-best.pth
+train: [13]  [  0/400]  eta: 0:07:32  lr: nan  time: 1.1316  data: 0.5440  max mem: 57344
+train: [13]  [ 20/400]  eta: 0:03:57  lr: 0.000133  loss: 2.5743 (2.5968)  grad: 0.2038 (0.2016)  time: 0.6001  data: 0.0030  max mem: 57344
+train: [13]  [ 40/400]  eta: 0:03:40  lr: 0.000131  loss: 2.5743 (2.5974)  grad: 0.2023 (0.2003)  time: 0.6013  data: 0.0034  max mem: 57344
+train: [13]  [ 60/400]  eta: 0:03:27  lr: 0.000130  loss: 2.5732 (2.5990)  grad: 0.1956 (0.1993)  time: 0.6026  data: 0.0040  max mem: 57344
+train: [13]  [ 80/400]  eta: 0:03:14  lr: 0.000128  loss: 2.5873 (2.6026)  grad: 0.1948 (0.1989)  time: 0.6026  data: 0.0039  max mem: 57344
+train: [13]  [100/400]  eta: 0:03:02  lr: 0.000127  loss: 2.6064 (2.6096)  grad: 0.1914 (0.1979)  time: 0.6006  data: 0.0034  max mem: 57344
+train: [13]  [120/400]  eta: 0:02:49  lr: 0.000125  loss: 2.6316 (2.6206)  grad: 0.1936 (0.1998)  time: 0.6012  data: 0.0035  max mem: 57344
+train: [13]  [140/400]  eta: 0:02:37  lr: 0.000124  loss: 2.6317 (2.6244)  grad: 0.1925 (0.1988)  time: 0.6012  data: 0.0035  max mem: 57344
+train: [13]  [160/400]  eta: 0:02:25  lr: 0.000122  loss: 2.6331 (2.6281)  grad: 0.1945 (0.1992)  time: 0.6012  data: 0.0036  max mem: 57344
+train: [13]  [180/400]  eta: 0:02:12  lr: 0.000120  loss: 2.6465 (2.6285)  grad: 0.1999 (0.1994)  time: 0.6024  data: 0.0038  max mem: 57344
+train: [13]  [200/400]  eta: 0:02:00  lr: 0.000119  loss: 2.6142 (2.6235)  grad: 0.1963 (0.1997)  time: 0.6034  data: 0.0041  max mem: 57344
+train: [13]  [220/400]  eta: 0:01:48  lr: 0.000117  loss: 2.5847 (2.6204)  grad: 0.1986 (0.2000)  time: 0.6028  data: 0.0040  max mem: 57344
+train: [13]  [240/400]  eta: 0:01:36  lr: 0.000116  loss: 2.5855 (2.6183)  grad: 0.1986 (0.1996)  time: 0.6026  data: 0.0040  max mem: 57344
+train: [13]  [260/400]  eta: 0:01:24  lr: 0.000114  loss: 2.5855 (2.6159)  grad: 0.1958 (0.1991)  time: 0.6016  data: 0.0037  max mem: 57344
+train: [13]  [280/400]  eta: 0:01:12  lr: 0.000113  loss: 2.5893 (2.6147)  grad: 0.2007 (0.1997)  time: 0.6011  data: 0.0035  max mem: 57344
+train: [13]  [300/400]  eta: 0:01:00  lr: 0.000111  loss: 2.6144 (2.6147)  grad: 0.2016 (0.1999)  time: 0.6008  data: 0.0036  max mem: 57344
+train: [13]  [320/400]  eta: 0:00:48  lr: 0.000110  loss: 2.6181 (2.6142)  grad: 0.1987 (0.2001)  time: 0.6010  data: 0.0036  max mem: 57344
+train: [13]  [340/400]  eta: 0:00:36  lr: 0.000108  loss: 2.6198 (2.6171)  grad: 0.2025 (0.2006)  time: 0.6013  data: 0.0036  max mem: 57344
+train: [13]  [360/400]  eta: 0:00:24  lr: 0.000107  loss: 2.6860 (2.6189)  grad: 0.2032 (0.2010)  time: 0.6012  data: 0.0035  max mem: 57344
+train: [13]  [380/400]  eta: 0:00:12  lr: 0.000105  loss: 2.6303 (2.6194)  grad: 0.2047 (0.2014)  time: 0.6009  data: 0.0035  max mem: 57344
+train: [13]  [399/400]  eta: 0:00:00  lr: 0.000104  loss: 2.6226 (2.6200)  grad: 0.2047 (0.2015)  time: 0.6014  data: 0.0036  max mem: 57344
+train: [13] Total time: 0:04:01 (0.6031 s / it)
+train: [13] Summary: lr: 0.000104  loss: 2.6226 (2.6200)  grad: 0.2047 (0.2015)
+eval (validation): [13]  [ 0/85]  eta: 0:01:25    time: 1.0075  data: 0.6504  max mem: 57344
+eval (validation): [13]  [20/85]  eta: 0:00:25    time: 0.3667  data: 0.0023  max mem: 57344
+eval (validation): [13]  [40/85]  eta: 0:00:17    time: 0.3680  data: 0.0033  max mem: 57344
+eval (validation): [13]  [60/85]  eta: 0:00:09    time: 0.3682  data: 0.0032  max mem: 57344
+eval (validation): [13]  [80/85]  eta: 0:00:01    time: 0.3675  data: 0.0034  max mem: 57344
+eval (validation): [13]  [84/85]  eta: 0:00:00    time: 0.3614  data: 0.0033  max mem: 57344
+eval (validation): [13] Total time: 0:00:31 (0.3748 s / it)
+cv: [13]  best hparam: (1.4, 1.0) (026) ('026_lr1.4e+00_wd1.0e+00')  loss: 2.410  acc: 0.282  f1: 0.223
+saving checkpoint experiments/schaefer1000/output/schaefer1000/schaefer1000_lr3e-4_2/eval_v2/nsd_cococlip__patch__attn/checkpoint-last.pth
+train: [14]  [  0/400]  eta: 0:07:28  lr: nan  time: 1.1212  data: 0.5322  max mem: 57344
+train: [14]  [ 20/400]  eta: 0:03:57  lr: 0.000102  loss: 2.5902 (2.5858)  grad: 0.2052 (0.2021)  time: 0.5999  data: 0.0021  max mem: 57344
+train: [14]  [ 40/400]  eta: 0:03:40  lr: 0.000101  loss: 2.6198 (2.6057)  grad: 0.2004 (0.2006)  time: 0.6009  data: 0.0035  max mem: 57344
+train: [14]  [ 60/400]  eta: 0:03:27  lr: 0.000099  loss: 2.6187 (2.6006)  grad: 0.1944 (0.1967)  time: 0.6015  data: 0.0036  max mem: 57344
+train: [14]  [ 80/400]  eta: 0:03:14  lr: 0.000098  loss: 2.5789 (2.6014)  grad: 0.1886 (0.1979)  time: 0.5998  data: 0.0032  max mem: 57344
+train: [14]  [100/400]  eta: 0:03:01  lr: 0.000096  loss: 2.5637 (2.5966)  grad: 0.1959 (0.1987)  time: 0.5998  data: 0.0034  max mem: 57344
+train: [14]  [120/400]  eta: 0:02:49  lr: 0.000095  loss: 2.5651 (2.5912)  grad: 0.1959 (0.1978)  time: 0.6024  data: 0.0040  max mem: 57344
+train: [14]  [140/400]  eta: 0:02:37  lr: 0.000093  loss: 2.5695 (2.5911)  grad: 0.1937 (0.1984)  time: 0.6023  data: 0.0037  max mem: 57344
+train: [14]  [160/400]  eta: 0:02:24  lr: 0.000092  loss: 2.6419 (2.5988)  grad: 0.1944 (0.1983)  time: 0.5999  data: 0.0033  max mem: 57344
+train: [14]  [180/400]  eta: 0:02:12  lr: 0.000090  loss: 2.6191 (2.5939)  grad: 0.1997 (0.1987)  time: 0.6003  data: 0.0033  max mem: 57344
+train: [14]  [200/400]  eta: 0:02:00  lr: 0.000089  loss: 2.5727 (2.5958)  grad: 0.1986 (0.1984)  time: 0.6001  data: 0.0034  max mem: 57344
+train: [14]  [220/400]  eta: 0:01:48  lr: 0.000088  loss: 2.5968 (2.5964)  grad: 0.1903 (0.1978)  time: 0.6005  data: 0.0035  max mem: 57344
+train: [14]  [240/400]  eta: 0:01:36  lr: 0.000086  loss: 2.6001 (2.5997)  grad: 0.1945 (0.1980)  time: 0.6005  data: 0.0035  max mem: 57344
+train: [14]  [260/400]  eta: 0:01:24  lr: 0.000085  loss: 2.6034 (2.5997)  grad: 0.1976 (0.1980)  time: 0.6020  data: 0.0037  max mem: 57344
+train: [14]  [280/400]  eta: 0:01:12  lr: 0.000083  loss: 2.6034 (2.6006)  grad: 0.1980 (0.1982)  time: 0.6031  data: 0.0039  max mem: 57344
+train: [14]  [300/400]  eta: 0:01:00  lr: 0.000082  loss: 2.5882 (2.6012)  grad: 0.1973 (0.1984)  time: 0.6034  data: 0.0040  max mem: 57344
+train: [14]  [320/400]  eta: 0:00:48  lr: 0.000081  loss: 2.5762 (2.6004)  grad: 0.1957 (0.1982)  time: 0.6019  data: 0.0038  max mem: 57344
+train: [14]  [340/400]  eta: 0:00:36  lr: 0.000079  loss: 2.5748 (2.5999)  grad: 0.1900 (0.1979)  time: 0.6022  data: 0.0038  max mem: 57344
+train: [14]  [360/400]  eta: 0:00:24  lr: 0.000078  loss: 2.6153 (2.6002)  grad: 0.1900 (0.1979)  time: 0.6013  data: 0.0035  max mem: 57344
+train: [14]  [380/400]  eta: 0:00:12  lr: 0.000076  loss: 2.6001 (2.6001)  grad: 0.1910 (0.1980)  time: 0.6008  data: 0.0034  max mem: 57344
+train: [14]  [399/400]  eta: 0:00:00  lr: 0.000075  loss: 2.5951 (2.5996)  grad: 0.1953 (0.1981)  time: 0.6009  data: 0.0034  max mem: 57344
+train: [14] Total time: 0:04:01 (0.6028 s / it)
+train: [14] Summary: lr: 0.000075  loss: 2.5951 (2.5996)  grad: 0.1953 (0.1981)
+eval (validation): [14]  [ 0/85]  eta: 0:01:24    time: 0.9966  data: 0.6354  max mem: 57344
+eval (validation): [14]  [20/85]  eta: 0:00:25    time: 0.3671  data: 0.0021  max mem: 57344
+eval (validation): [14]  [40/85]  eta: 0:00:17    time: 0.3686  data: 0.0033  max mem: 57344
+eval (validation): [14]  [60/85]  eta: 0:00:09    time: 0.3688  data: 0.0034  max mem: 57344
+eval (validation): [14]  [80/85]  eta: 0:00:01    time: 0.3682  data: 0.0033  max mem: 57344
+eval (validation): [14]  [84/85]  eta: 0:00:00    time: 0.3619  data: 0.0033  max mem: 57344
+eval (validation): [14] Total time: 0:00:31 (0.3752 s / it)
+cv: [14]  best hparam: (1.2, 1.0) (025) ('025_lr1.2e+00_wd1.0e+00')  loss: 2.381  acc: 0.286  f1: 0.230
+saving checkpoint experiments/schaefer1000/output/schaefer1000/schaefer1000_lr3e-4_2/eval_v2/nsd_cococlip__patch__attn/checkpoint-last.pth
+saving best checkpoint experiments/schaefer1000/output/schaefer1000/schaefer1000_lr3e-4_2/eval_v2/nsd_cococlip__patch__attn/checkpoint-best.pth
+train: [15]  [  0/400]  eta: 0:08:09  lr: nan  time: 1.2231  data: 0.6317  max mem: 57344
+train: [15]  [ 20/400]  eta: 0:03:59  lr: 0.000074  loss: 2.5456 (2.5607)  grad: 0.1945 (0.2008)  time: 0.6018  data: 0.0027  max mem: 57344
+train: [15]  [ 40/400]  eta: 0:03:42  lr: 0.000072  loss: 2.5456 (2.5580)  grad: 0.1964 (0.2021)  time: 0.6019  data: 0.0033  max mem: 57344
+train: [15]  [ 60/400]  eta: 0:03:27  lr: 0.000071  loss: 2.5907 (2.5725)  grad: 0.2038 (0.2031)  time: 0.6001  data: 0.0034  max mem: 57344
+train: [15]  [ 80/400]  eta: 0:03:14  lr: 0.000070  loss: 2.5900 (2.5712)  grad: 0.2020 (0.2029)  time: 0.5998  data: 0.0035  max mem: 57344
+train: [15]  [100/400]  eta: 0:03:02  lr: 0.000068  loss: 2.5772 (2.5766)  grad: 0.1984 (0.2036)  time: 0.5994  data: 0.0034  max mem: 57344
+train: [15]  [120/400]  eta: 0:02:49  lr: 0.000067  loss: 2.6001 (2.5817)  grad: 0.1986 (0.2037)  time: 0.5992  data: 0.0033  max mem: 57344
+train: [15]  [140/400]  eta: 0:02:37  lr: 0.000066  loss: 2.5881 (2.5827)  grad: 0.2050 (0.2041)  time: 0.6000  data: 0.0033  max mem: 57344
+train: [15]  [160/400]  eta: 0:02:24  lr: 0.000064  loss: 2.5625 (2.5819)  grad: 0.2034 (0.2033)  time: 0.5999  data: 0.0034  max mem: 57344
+train: [15]  [180/400]  eta: 0:02:12  lr: 0.000063  loss: 2.5565 (2.5784)  grad: 0.1938 (0.2032)  time: 0.6000  data: 0.0034  max mem: 57344
+train: [15]  [200/400]  eta: 0:02:00  lr: 0.000062  loss: 2.5616 (2.5826)  grad: 0.2063 (0.2038)  time: 0.6025  data: 0.0038  max mem: 57344
+train: [15]  [220/400]  eta: 0:01:48  lr: 0.000061  loss: 2.6014 (2.5847)  grad: 0.2049 (0.2036)  time: 0.6013  data: 0.0036  max mem: 57344
+train: [15]  [240/400]  eta: 0:01:36  lr: 0.000059  loss: 2.5673 (2.5803)  grad: 0.2009 (0.2034)  time: 0.6007  data: 0.0034  max mem: 57344
+train: [15]  [260/400]  eta: 0:01:24  lr: 0.000058  loss: 2.5577 (2.5811)  grad: 0.1946 (0.2028)  time: 0.6003  data: 0.0034  max mem: 57344
+train: [15]  [280/400]  eta: 0:01:12  lr: 0.000057  loss: 2.5876 (2.5806)  grad: 0.1926 (0.2025)  time: 0.6005  data: 0.0034  max mem: 57344
+train: [15]  [300/400]  eta: 0:01:00  lr: 0.000056  loss: 2.5520 (2.5785)  grad: 0.1936 (0.2019)  time: 0.6007  data: 0.0035  max mem: 57344
+train: [15]  [320/400]  eta: 0:00:48  lr: 0.000054  loss: 2.5896 (2.5808)  grad: 0.1940 (0.2020)  time: 0.6004  data: 0.0035  max mem: 57344
+train: [15]  [340/400]  eta: 0:00:36  lr: 0.000053  loss: 2.5827 (2.5777)  grad: 0.2011 (0.2017)  time: 0.6032  data: 0.0041  max mem: 57344
+train: [15]  [360/400]  eta: 0:00:24  lr: 0.000052  loss: 2.5359 (2.5766)  grad: 0.2005 (0.2016)  time: 0.6033  data: 0.0039  max mem: 57344
+train: [15]  [380/400]  eta: 0:00:12  lr: 0.000051  loss: 2.5650 (2.5766)  grad: 0.1965 (0.2013)  time: 0.6023  data: 0.0038  max mem: 57344
+train: [15]  [399/400]  eta: 0:00:00  lr: 0.000050  loss: 2.5953 (2.5780)  grad: 0.2006 (0.2017)  time: 0.6019  data: 0.0036  max mem: 57344
+train: [15] Total time: 0:04:01 (0.6028 s / it)
+train: [15] Summary: lr: 0.000050  loss: 2.5953 (2.5780)  grad: 0.2006 (0.2017)
+eval (validation): [15]  [ 0/85]  eta: 0:01:25    time: 1.0002  data: 0.6419  max mem: 57344
+eval (validation): [15]  [20/85]  eta: 0:00:26    time: 0.3701  data: 0.0031  max mem: 57344
+eval (validation): [15]  [40/85]  eta: 0:00:17    time: 0.3705  data: 0.0036  max mem: 57344
+eval (validation): [15]  [60/85]  eta: 0:00:09    time: 0.3692  data: 0.0035  max mem: 57344
+eval (validation): [15]  [80/85]  eta: 0:00:01    time: 0.3696  data: 0.0035  max mem: 57344
+eval (validation): [15]  [84/85]  eta: 0:00:00    time: 0.3632  data: 0.0035  max mem: 57344
+eval (validation): [15] Total time: 0:00:32 (0.3769 s / it)
+cv: [15]  best hparam: (1.4, 1.0) (026) ('026_lr1.4e+00_wd1.0e+00')  loss: 2.374  acc: 0.293  f1: 0.242
+saving checkpoint experiments/schaefer1000/output/schaefer1000/schaefer1000_lr3e-4_2/eval_v2/nsd_cococlip__patch__attn/checkpoint-last.pth
+saving best checkpoint experiments/schaefer1000/output/schaefer1000/schaefer1000_lr3e-4_2/eval_v2/nsd_cococlip__patch__attn/checkpoint-best.pth
+train: [16]  [  0/400]  eta: 0:08:06  lr: nan  time: 1.2168  data: 0.6271  max mem: 57344
+train: [16]  [ 20/400]  eta: 0:04:00  lr: 0.000048  loss: 2.5482 (2.5851)  grad: 0.1982 (0.2009)  time: 0.6032  data: 0.0037  max mem: 57344
+train: [16]  [ 40/400]  eta: 0:03:42  lr: 0.000047  loss: 2.5482 (2.5690)  grad: 0.1982 (0.1984)  time: 0.6031  data: 0.0037  max mem: 57344
+train: [16]  [ 60/400]  eta: 0:03:28  lr: 0.000046  loss: 2.5300 (2.5530)  grad: 0.1905 (0.1963)  time: 0.6028  data: 0.0036  max mem: 57344
+train: [16]  [ 80/400]  eta: 0:03:15  lr: 0.000045  loss: 2.5334 (2.5536)  grad: 0.1917 (0.1968)  time: 0.6030  data: 0.0036  max mem: 57344
+train: [16]  [100/400]  eta: 0:03:02  lr: 0.000044  loss: 2.5478 (2.5488)  grad: 0.1924 (0.1959)  time: 0.6023  data: 0.0035  max mem: 57344
+train: [16]  [120/400]  eta: 0:02:50  lr: 0.000043  loss: 2.5483 (2.5595)  grad: 0.1906 (0.1962)  time: 0.6013  data: 0.0033  max mem: 57344
+train: [16]  [140/400]  eta: 0:02:37  lr: 0.000042  loss: 2.5653 (2.5586)  grad: 0.1906 (0.1955)  time: 0.6006  data: 0.0034  max mem: 57344
+train: [16]  [160/400]  eta: 0:02:25  lr: 0.000041  loss: 2.5636 (2.5596)  grad: 0.1896 (0.1955)  time: 0.5995  data: 0.0034  max mem: 57344
+train: [16]  [180/400]  eta: 0:02:13  lr: 0.000040  loss: 2.5740 (2.5606)  grad: 0.1896 (0.1953)  time: 0.5997  data: 0.0033  max mem: 57344
+train: [16]  [200/400]  eta: 0:02:00  lr: 0.000039  loss: 2.5635 (2.5616)  grad: 0.1881 (0.1954)  time: 0.6000  data: 0.0033  max mem: 57344
+train: [16]  [220/400]  eta: 0:01:48  lr: 0.000038  loss: 2.5561 (2.5623)  grad: 0.2018 (0.1964)  time: 0.5994  data: 0.0032  max mem: 57344
+train: [16]  [240/400]  eta: 0:01:36  lr: 0.000036  loss: 2.5934 (2.5652)  grad: 0.2018 (0.1963)  time: 0.5997  data: 0.0033  max mem: 57344
+train: [16]  [260/400]  eta: 0:01:24  lr: 0.000035  loss: 2.5934 (2.5642)  grad: 0.1939 (0.1966)  time: 0.6019  data: 0.0037  max mem: 57344
+train: [16]  [280/400]  eta: 0:01:12  lr: 0.000034  loss: 2.5458 (2.5637)  grad: 0.2030 (0.1974)  time: 0.6029  data: 0.0039  max mem: 57344
+train: [16]  [300/400]  eta: 0:01:00  lr: 0.000033  loss: 2.5466 (2.5624)  grad: 0.2046 (0.1977)  time: 0.6004  data: 0.0034  max mem: 57344
+train: [16]  [320/400]  eta: 0:00:48  lr: 0.000032  loss: 2.5764 (2.5659)  grad: 0.1976 (0.1974)  time: 0.6004  data: 0.0033  max mem: 57344
+train: [16]  [340/400]  eta: 0:00:36  lr: 0.000031  loss: 2.5936 (2.5673)  grad: 0.1924 (0.1978)  time: 0.5997  data: 0.0034  max mem: 57344
+train: [16]  [360/400]  eta: 0:00:24  lr: 0.000031  loss: 2.5854 (2.5683)  grad: 0.2038 (0.1980)  time: 0.6006  data: 0.0035  max mem: 57344
+train: [16]  [380/400]  eta: 0:00:12  lr: 0.000030  loss: 2.5584 (2.5683)  grad: 0.1958 (0.1978)  time: 0.6003  data: 0.0034  max mem: 57344
+train: [16]  [399/400]  eta: 0:00:00  lr: 0.000029  loss: 2.5484 (2.5692)  grad: 0.1889 (0.1977)  time: 0.6006  data: 0.0034  max mem: 57344
+train: [16] Total time: 0:04:01 (0.6029 s / it)
+train: [16] Summary: lr: 0.000029  loss: 2.5484 (2.5692)  grad: 0.1889 (0.1977)
+eval (validation): [16]  [ 0/85]  eta: 0:01:28    time: 1.0424  data: 0.6830  max mem: 57344
+eval (validation): [16]  [20/85]  eta: 0:00:26    time: 0.3700  data: 0.0038  max mem: 57344
+eval (validation): [16]  [40/85]  eta: 0:00:17    time: 0.3702  data: 0.0037  max mem: 57344
+eval (validation): [16]  [60/85]  eta: 0:00:09    time: 0.3705  data: 0.0038  max mem: 57344
+eval (validation): [16]  [80/85]  eta: 0:00:01    time: 0.3705  data: 0.0039  max mem: 57344
+eval (validation): [16]  [84/85]  eta: 0:00:00    time: 0.3637  data: 0.0037  max mem: 57344
+eval (validation): [16] Total time: 0:00:32 (0.3781 s / it)
+cv: [16]  best hparam: (1.4, 1.0) (026) ('026_lr1.4e+00_wd1.0e+00')  loss: 2.398  acc: 0.288  f1: 0.237
+saving checkpoint experiments/schaefer1000/output/schaefer1000/schaefer1000_lr3e-4_2/eval_v2/nsd_cococlip__patch__attn/checkpoint-last.pth
+train: [17]  [  0/400]  eta: 0:08:19  lr: nan  time: 1.2492  data: 0.6582  max mem: 57344
+train: [17]  [ 20/400]  eta: 0:04:01  lr: 0.000028  loss: 2.5144 (2.5679)  grad: 0.1873 (0.1918)  time: 0.6037  data: 0.0034  max mem: 57344
+train: [17]  [ 40/400]  eta: 0:03:42  lr: 0.000027  loss: 2.5144 (2.5279)  grad: 0.1924 (0.1953)  time: 0.6028  data: 0.0035  max mem: 57344
+train: [17]  [ 60/400]  eta: 0:03:28  lr: 0.000026  loss: 2.5254 (2.5261)  grad: 0.1937 (0.1951)  time: 0.6010  data: 0.0035  max mem: 57344
+train: [17]  [ 80/400]  eta: 0:03:15  lr: 0.000025  loss: 2.5656 (2.5450)  grad: 0.1937 (0.1951)  time: 0.6014  data: 0.0036  max mem: 57344
+train: [17]  [100/400]  eta: 0:03:02  lr: 0.000024  loss: 2.6087 (2.5559)  grad: 0.1975 (0.1956)  time: 0.6014  data: 0.0036  max mem: 57344
+train: [17]  [120/400]  eta: 0:02:50  lr: 0.000023  loss: 2.5855 (2.5529)  grad: 0.1983 (0.1964)  time: 0.6012  data: 0.0035  max mem: 57344
+train: [17]  [140/400]  eta: 0:02:37  lr: 0.000023  loss: 2.4961 (2.5481)  grad: 0.1987 (0.1962)  time: 0.6011  data: 0.0035  max mem: 57344
+train: [17]  [160/400]  eta: 0:02:25  lr: 0.000022  loss: 2.5507 (2.5523)  grad: 0.1997 (0.1966)  time: 0.6013  data: 0.0035  max mem: 57344
+train: [17]  [180/400]  eta: 0:02:13  lr: 0.000021  loss: 2.5676 (2.5542)  grad: 0.1997 (0.1971)  time: 0.6008  data: 0.0036  max mem: 57344
+train: [17]  [200/400]  eta: 0:02:00  lr: 0.000020  loss: 2.5498 (2.5537)  grad: 0.1891 (0.1970)  time: 0.6001  data: 0.0033  max mem: 57344
+train: [17]  [220/400]  eta: 0:01:48  lr: 0.000019  loss: 2.5532 (2.5528)  grad: 0.1893 (0.1970)  time: 0.5995  data: 0.0033  max mem: 57344
+train: [17]  [240/400]  eta: 0:01:36  lr: 0.000019  loss: 2.5532 (2.5517)  grad: 0.1913 (0.1965)  time: 0.5992  data: 0.0033  max mem: 57344
+train: [17]  [260/400]  eta: 0:01:24  lr: 0.000018  loss: 2.5358 (2.5532)  grad: 0.1924 (0.1970)  time: 0.6000  data: 0.0033  max mem: 57344
+train: [17]  [280/400]  eta: 0:01:12  lr: 0.000017  loss: 2.5432 (2.5534)  grad: 0.1975 (0.1973)  time: 0.5990  data: 0.0034  max mem: 57344
+train: [17]  [300/400]  eta: 0:01:00  lr: 0.000016  loss: 2.5432 (2.5518)  grad: 0.1983 (0.1973)  time: 0.5998  data: 0.0033  max mem: 57344
+train: [17]  [320/400]  eta: 0:00:48  lr: 0.000016  loss: 2.5652 (2.5533)  grad: 0.1983 (0.1972)  time: 0.5998  data: 0.0034  max mem: 57344
+train: [17]  [340/400]  eta: 0:00:36  lr: 0.000015  loss: 2.5740 (2.5535)  grad: 0.1933 (0.1969)  time: 0.6016  data: 0.0037  max mem: 57344
+train: [17]  [360/400]  eta: 0:00:24  lr: 0.000014  loss: 2.5491 (2.5529)  grad: 0.1955 (0.1972)  time: 0.6030  data: 0.0039  max mem: 57344
+train: [17]  [380/400]  eta: 0:00:12  lr: 0.000014  loss: 2.5636 (2.5543)  grad: 0.1964 (0.1972)  time: 0.6011  data: 0.0036  max mem: 57344
+train: [17]  [399/400]  eta: 0:00:00  lr: 0.000013  loss: 2.5555 (2.5533)  grad: 0.1937 (0.1974)  time: 0.6011  data: 0.0034  max mem: 57344
+train: [17] Total time: 0:04:01 (0.6029 s / it)
+train: [17] Summary: lr: 0.000013  loss: 2.5555 (2.5533)  grad: 0.1937 (0.1974)
+eval (validation): [17]  [ 0/85]  eta: 0:01:17    time: 0.9142  data: 0.5531  max mem: 57344
+eval (validation): [17]  [20/85]  eta: 0:00:25    time: 0.3669  data: 0.0025  max mem: 57344
+eval (validation): [17]  [40/85]  eta: 0:00:17    time: 0.3673  data: 0.0032  max mem: 57344
+eval (validation): [17]  [60/85]  eta: 0:00:09    time: 0.3679  data: 0.0035  max mem: 57344
+eval (validation): [17]  [80/85]  eta: 0:00:01    time: 0.3689  data: 0.0033  max mem: 57344
+eval (validation): [17]  [84/85]  eta: 0:00:00    time: 0.3625  data: 0.0033  max mem: 57344
+eval (validation): [17] Total time: 0:00:31 (0.3740 s / it)
+cv: [17]  best hparam: (1.4, 1.0) (026) ('026_lr1.4e+00_wd1.0e+00')  loss: 2.382  acc: 0.292  f1: 0.237
+saving checkpoint experiments/schaefer1000/output/schaefer1000/schaefer1000_lr3e-4_2/eval_v2/nsd_cococlip__patch__attn/checkpoint-last.pth
+train: [18]  [  0/400]  eta: 0:08:15  lr: nan  time: 1.2378  data: 0.6483  max mem: 57344
+train: [18]  [ 20/400]  eta: 0:03:59  lr: 0.000012  loss: 2.5094 (2.5234)  grad: 0.1916 (0.1971)  time: 0.5994  data: 0.0020  max mem: 57344
+train: [18]  [ 40/400]  eta: 0:03:41  lr: 0.000012  loss: 2.4938 (2.5076)  grad: 0.1916 (0.1943)  time: 0.6018  data: 0.0038  max mem: 57344
+train: [18]  [ 60/400]  eta: 0:03:28  lr: 0.000011  loss: 2.4846 (2.5060)  grad: 0.1922 (0.1960)  time: 0.6036  data: 0.0040  max mem: 57344
+train: [18]  [ 80/400]  eta: 0:03:15  lr: 0.000011  loss: 2.5228 (2.5110)  grad: 0.1922 (0.1950)  time: 0.6034  data: 0.0040  max mem: 57344
+train: [18]  [100/400]  eta: 0:03:02  lr: 0.000010  loss: 2.5431 (2.5240)  grad: 0.1957 (0.1975)  time: 0.6029  data: 0.0039  max mem: 57344
+train: [18]  [120/400]  eta: 0:02:50  lr: 0.000009  loss: 2.5629 (2.5310)  grad: 0.2068 (0.1979)  time: 0.6018  data: 0.0037  max mem: 57344
+train: [18]  [140/400]  eta: 0:02:37  lr: 0.000009  loss: 2.5242 (2.5280)  grad: 0.1966 (0.1965)  time: 0.6015  data: 0.0035  max mem: 57344
+train: [18]  [160/400]  eta: 0:02:25  lr: 0.000008  loss: 2.5296 (2.5327)  grad: 0.1908 (0.1968)  time: 0.6015  data: 0.0036  max mem: 57344
+train: [18]  [180/400]  eta: 0:02:13  lr: 0.000008  loss: 2.5596 (2.5356)  grad: 0.1959 (0.1971)  time: 0.6013  data: 0.0035  max mem: 57344
+train: [18]  [200/400]  eta: 0:02:01  lr: 0.000007  loss: 2.5030 (2.5309)  grad: 0.1918 (0.1967)  time: 0.6013  data: 0.0035  max mem: 57344
+train: [18]  [220/400]  eta: 0:01:48  lr: 0.000007  loss: 2.5397 (2.5338)  grad: 0.1913 (0.1971)  time: 0.6012  data: 0.0035  max mem: 57344
+train: [18]  [240/400]  eta: 0:01:36  lr: 0.000006  loss: 2.5634 (2.5365)  grad: 0.1994 (0.1973)  time: 0.6013  data: 0.0036  max mem: 57344
+train: [18]  [260/400]  eta: 0:01:24  lr: 0.000006  loss: 2.5329 (2.5327)  grad: 0.2001 (0.1979)  time: 0.6010  data: 0.0036  max mem: 57344
+train: [18]  [280/400]  eta: 0:01:12  lr: 0.000006  loss: 2.4681 (2.5305)  grad: 0.1951 (0.1974)  time: 0.5998  data: 0.0034  max mem: 57344
+train: [18]  [300/400]  eta: 0:01:00  lr: 0.000005  loss: 2.5178 (2.5324)  grad: 0.1891 (0.1977)  time: 0.5995  data: 0.0033  max mem: 57344
+train: [18]  [320/400]  eta: 0:00:48  lr: 0.000005  loss: 2.5338 (2.5312)  grad: 0.1982 (0.1981)  time: 0.5997  data: 0.0033  max mem: 57344
+train: [18]  [340/400]  eta: 0:00:36  lr: 0.000004  loss: 2.5069 (2.5303)  grad: 0.1982 (0.1982)  time: 0.5999  data: 0.0034  max mem: 57344
+train: [18]  [360/400]  eta: 0:00:24  lr: 0.000004  loss: 2.5282 (2.5317)  grad: 0.1959 (0.1978)  time: 0.5996  data: 0.0032  max mem: 57344
+train: [18]  [380/400]  eta: 0:00:12  lr: 0.000004  loss: 2.5457 (2.5321)  grad: 0.1906 (0.1978)  time: 0.5997  data: 0.0033  max mem: 57344
+train: [18]  [399/400]  eta: 0:00:00  lr: 0.000003  loss: 2.5576 (2.5340)  grad: 0.1979 (0.1982)  time: 0.5996  data: 0.0034  max mem: 57344
+train: [18] Total time: 0:04:01 (0.6029 s / it)
+train: [18] Summary: lr: 0.000003  loss: 2.5576 (2.5340)  grad: 0.1979 (0.1982)
+eval (validation): [18]  [ 0/85]  eta: 0:01:15    time: 0.8832  data: 0.5246  max mem: 57344
+eval (validation): [18]  [20/85]  eta: 0:00:25    time: 0.3692  data: 0.0020  max mem: 57344
+eval (validation): [18]  [40/85]  eta: 0:00:17    time: 0.3709  data: 0.0037  max mem: 57344
+eval (validation): [18]  [60/85]  eta: 0:00:09    time: 0.3715  data: 0.0037  max mem: 57344
+eval (validation): [18]  [80/85]  eta: 0:00:01    time: 0.3718  data: 0.0037  max mem: 57344
+eval (validation): [18]  [84/85]  eta: 0:00:00    time: 0.3656  data: 0.0037  max mem: 57344
+eval (validation): [18] Total time: 0:00:32 (0.3766 s / it)
+cv: [18]  best hparam: (1.4, 1.0) (026) ('026_lr1.4e+00_wd1.0e+00')  loss: 2.382  acc: 0.292  f1: 0.237
+saving checkpoint experiments/schaefer1000/output/schaefer1000/schaefer1000_lr3e-4_2/eval_v2/nsd_cococlip__patch__attn/checkpoint-last.pth
+train: [19]  [  0/400]  eta: 0:07:54  lr: nan  time: 1.1867  data: 0.5939  max mem: 57344
+train: [19]  [ 20/400]  eta: 0:03:58  lr: 0.000003  loss: 2.5290 (2.5453)  grad: 0.1923 (0.1965)  time: 0.6004  data: 0.0025  max mem: 57344
+train: [19]  [ 40/400]  eta: 0:03:41  lr: 0.000003  loss: 2.5338 (2.5610)  grad: 0.1954 (0.1989)  time: 0.6003  data: 0.0033  max mem: 57344
+train: [19]  [ 60/400]  eta: 0:03:27  lr: 0.000002  loss: 2.5322 (2.5480)  grad: 0.1954 (0.1985)  time: 0.6003  data: 0.0033  max mem: 57344
+train: [19]  [ 80/400]  eta: 0:03:14  lr: 0.000002  loss: 2.4911 (2.5429)  grad: 0.1979 (0.1997)  time: 0.6008  data: 0.0034  max mem: 57344
+train: [19]  [100/400]  eta: 0:03:01  lr: 0.000002  loss: 2.5451 (2.5489)  grad: 0.1977 (0.1989)  time: 0.6007  data: 0.0035  max mem: 57344
+train: [19]  [120/400]  eta: 0:02:49  lr: 0.000002  loss: 2.5569 (2.5405)  grad: 0.1899 (0.1971)  time: 0.6018  data: 0.0037  max mem: 57344
+train: [19]  [140/400]  eta: 0:02:37  lr: 0.000001  loss: 2.5068 (2.5392)  grad: 0.1914 (0.1968)  time: 0.6033  data: 0.0040  max mem: 57344
+train: [19]  [160/400]  eta: 0:02:25  lr: 0.000001  loss: 2.5068 (2.5388)  grad: 0.1966 (0.1969)  time: 0.6031  data: 0.0040  max mem: 57344
+train: [19]  [180/400]  eta: 0:02:13  lr: 0.000001  loss: 2.5282 (2.5371)  grad: 0.1970 (0.1966)  time: 0.6022  data: 0.0038  max mem: 57344
+train: [19]  [200/400]  eta: 0:02:00  lr: 0.000001  loss: 2.5334 (2.5385)  grad: 0.1970 (0.1968)  time: 0.6015  data: 0.0036  max mem: 57344
+train: [19]  [220/400]  eta: 0:01:48  lr: 0.000001  loss: 2.5145 (2.5360)  grad: 0.1944 (0.1966)  time: 0.6016  data: 0.0035  max mem: 57344
+train: [19]  [240/400]  eta: 0:01:36  lr: 0.000001  loss: 2.4893 (2.5339)  grad: 0.1912 (0.1961)  time: 0.6009  data: 0.0035  max mem: 57344
+train: [19]  [260/400]  eta: 0:01:24  lr: 0.000000  loss: 2.5116 (2.5330)  grad: 0.1884 (0.1958)  time: 0.6007  data: 0.0035  max mem: 57344
+train: [19]  [280/400]  eta: 0:01:12  lr: 0.000000  loss: 2.5490 (2.5371)  grad: 0.1930 (0.1958)  time: 0.6011  data: 0.0035  max mem: 57344
+train: [19]  [300/400]  eta: 0:01:00  lr: 0.000000  loss: 2.5765 (2.5386)  grad: 0.1906 (0.1954)  time: 0.6014  data: 0.0035  max mem: 57344
+train: [19]  [320/400]  eta: 0:00:48  lr: 0.000000  loss: 2.5425 (2.5378)  grad: 0.1939 (0.1957)  time: 0.6007  data: 0.0034  max mem: 57344
+train: [19]  [340/400]  eta: 0:00:36  lr: 0.000000  loss: 2.5400 (2.5375)  grad: 0.1923 (0.1952)  time: 0.6012  data: 0.0034  max mem: 57344
+train: [19]  [360/400]  eta: 0:00:24  lr: 0.000000  loss: 2.5507 (2.5399)  grad: 0.1896 (0.1955)  time: 0.5997  data: 0.0031  max mem: 57344
+train: [19]  [380/400]  eta: 0:00:12  lr: 0.000000  loss: 2.5549 (2.5403)  grad: 0.1963 (0.1954)  time: 0.5995  data: 0.0032  max mem: 57344
+train: [19]  [399/400]  eta: 0:00:00  lr: 0.000000  loss: 2.5472 (2.5408)  grad: 0.1890 (0.1952)  time: 0.6003  data: 0.0033  max mem: 57344
+train: [19] Total time: 0:04:01 (0.6028 s / it)
+train: [19] Summary: lr: 0.000000  loss: 2.5472 (2.5408)  grad: 0.1890 (0.1952)
+eval (validation): [19]  [ 0/85]  eta: 0:01:20    time: 0.9527  data: 0.5922  max mem: 57344
+eval (validation): [19]  [20/85]  eta: 0:00:25    time: 0.3681  data: 0.0031  max mem: 57344
+eval (validation): [19]  [40/85]  eta: 0:00:17    time: 0.3683  data: 0.0031  max mem: 57344
+eval (validation): [19]  [60/85]  eta: 0:00:09    time: 0.3680  data: 0.0035  max mem: 57344
+eval (validation): [19]  [80/85]  eta: 0:00:01    time: 0.3685  data: 0.0033  max mem: 57344
+eval (validation): [19]  [84/85]  eta: 0:00:00    time: 0.3621  data: 0.0032  max mem: 57344
+eval (validation): [19] Total time: 0:00:31 (0.3748 s / it)
+cv: [19]  best hparam: (1.4, 1.0) (026) ('026_lr1.4e+00_wd1.0e+00')  loss: 2.386  acc: 0.292  f1: 0.237
+saving checkpoint experiments/schaefer1000/output/schaefer1000/schaefer1000_lr3e-4_2/eval_v2/nsd_cococlip__patch__attn/checkpoint-last.pth
+evaluating last checkpoint: experiments/schaefer1000/output/schaefer1000/schaefer1000_lr3e-4_2/eval_v2/nsd_cococlip__patch__attn/checkpoint-last.pth
+eval model info:
+{"score": 0.2918050941306755, "hparam": [1.4, 1.0], "hparam_id": 26, "epoch": 19, "is_best": false, "best_score": 0.29346622369878184}
+eval (train): [20]  [  0/509]  eta: 0:07:34    time: 0.8938  data: 0.5364  max mem: 57344
+eval (train): [20]  [ 20/509]  eta: 0:03:12    time: 0.3688  data: 0.0027  max mem: 57344
+eval (train): [20]  [ 40/509]  eta: 0:02:59    time: 0.3701  data: 0.0031  max mem: 57344
+eval (train): [20]  [ 60/509]  eta: 0:02:49    time: 0.3706  data: 0.0033  max mem: 57344
+eval (train): [20]  [ 80/509]  eta: 0:02:41    time: 0.3713  data: 0.0035  max mem: 57344
+eval (train): [20]  [100/509]  eta: 0:02:33    time: 0.3716  data: 0.0035  max mem: 57344
+eval (train): [20]  [120/509]  eta: 0:02:25    time: 0.3723  data: 0.0037  max mem: 57344
+eval (train): [20]  [140/509]  eta: 0:02:18    time: 0.3708  data: 0.0034  max mem: 57344
+eval (train): [20]  [160/509]  eta: 0:02:10    time: 0.3708  data: 0.0032  max mem: 57344
+eval (train): [20]  [180/509]  eta: 0:02:02    time: 0.3704  data: 0.0031  max mem: 57344
+eval (train): [20]  [200/509]  eta: 0:01:55    time: 0.3714  data: 0.0034  max mem: 57344
+eval (train): [20]  [220/509]  eta: 0:01:47    time: 0.3704  data: 0.0032  max mem: 57344
+eval (train): [20]  [240/509]  eta: 0:01:40    time: 0.3696  data: 0.0033  max mem: 57344
+eval (train): [20]  [260/509]  eta: 0:01:32    time: 0.3693  data: 0.0035  max mem: 57344
+eval (train): [20]  [280/509]  eta: 0:01:25    time: 0.3698  data: 0.0035  max mem: 57344
+eval (train): [20]  [300/509]  eta: 0:01:17    time: 0.3699  data: 0.0034  max mem: 57344
+eval (train): [20]  [320/509]  eta: 0:01:10    time: 0.3705  data: 0.0037  max mem: 57344
+eval (train): [20]  [340/509]  eta: 0:01:02    time: 0.3708  data: 0.0038  max mem: 57344
+eval (train): [20]  [360/509]  eta: 0:00:55    time: 0.3706  data: 0.0038  max mem: 57344
+eval (train): [20]  [380/509]  eta: 0:00:47    time: 0.3707  data: 0.0038  max mem: 57344
+eval (train): [20]  [400/509]  eta: 0:00:40    time: 0.3706  data: 0.0037  max mem: 57344
+eval (train): [20]  [420/509]  eta: 0:00:33    time: 0.3698  data: 0.0036  max mem: 57344
+eval (train): [20]  [440/509]  eta: 0:00:25    time: 0.3694  data: 0.0037  max mem: 57344
+eval (train): [20]  [460/509]  eta: 0:00:18    time: 0.3697  data: 0.0034  max mem: 57344
+eval (train): [20]  [480/509]  eta: 0:00:10    time: 0.3691  data: 0.0035  max mem: 57344
+eval (train): [20]  [500/509]  eta: 0:00:03    time: 0.3693  data: 0.0034  max mem: 57344
+eval (train): [20]  [508/509]  eta: 0:00:00    time: 0.3588  data: 0.0035  max mem: 57344
+eval (train): [20] Total time: 0:03:08 (0.3713 s / it)
+eval (validation): [20]  [ 0/85]  eta: 0:01:27    time: 1.0257  data: 0.6668  max mem: 57344
+eval (validation): [20]  [20/85]  eta: 0:00:26    time: 0.3697  data: 0.0029  max mem: 57344
+eval (validation): [20]  [40/85]  eta: 0:00:17    time: 0.3708  data: 0.0035  max mem: 57344
+eval (validation): [20]  [60/85]  eta: 0:00:09    time: 0.3704  data: 0.0033  max mem: 57344
+eval (validation): [20]  [80/85]  eta: 0:00:01    time: 0.3706  data: 0.0035  max mem: 57344
+eval (validation): [20]  [84/85]  eta: 0:00:00    time: 0.3642  data: 0.0035  max mem: 57344
+eval (validation): [20] Total time: 0:00:32 (0.3779 s / it)
+eval (test): [20]  [ 0/85]  eta: 0:01:28    time: 1.0404  data: 0.6783  max mem: 57344
+eval (test): [20]  [20/85]  eta: 0:00:25    time: 0.3678  data: 0.0025  max mem: 57344
+eval (test): [20]  [40/85]  eta: 0:00:17    time: 0.3684  data: 0.0034  max mem: 57344
+eval (test): [20]  [60/85]  eta: 0:00:09    time: 0.3686  data: 0.0035  max mem: 57344
+eval (test): [20]  [80/85]  eta: 0:00:01    time: 0.3677  data: 0.0031  max mem: 57344
+eval (test): [20]  [84/85]  eta: 0:00:00    time: 0.3540  data: 0.0031  max mem: 57344
+eval (test): [20] Total time: 0:00:31 (0.3740 s / it)
+eval (testid): [20]  [ 0/82]  eta: 0:01:16    time: 0.9360  data: 0.5765  max mem: 57344
+eval (testid): [20]  [20/82]  eta: 0:00:24    time: 0.3670  data: 0.0025  max mem: 57344
+eval (testid): [20]  [40/82]  eta: 0:00:16    time: 0.3680  data: 0.0030  max mem: 57344
+eval (testid): [20]  [60/82]  eta: 0:00:08    time: 0.3678  data: 0.0033  max mem: 57344
+eval (testid): [20]  [80/82]  eta: 0:00:00    time: 0.3685  data: 0.0034  max mem: 57344
+eval (testid): [20]  [81/82]  eta: 0:00:00    time: 0.3515  data: 0.0034  max mem: 57344
+eval (testid): [20] Total time: 0:00:30 (0.3718 s / it)
+evaluating best checkpoint: experiments/schaefer1000/output/schaefer1000/schaefer1000_lr3e-4_2/eval_v2/nsd_cococlip__patch__attn/checkpoint-best.pth
+eval model info:
+{"score": 0.29346622369878184, "hparam": [1.4, 1.0], "hparam_id": 26, "epoch": 15, "is_best": true, "best_score": 0.29346622369878184}
+eval (train): [20]  [  0/509]  eta: 0:07:05    time: 0.8360  data: 0.4776  max mem: 57344
+eval (train): [20]  [ 20/509]  eta: 0:03:10    time: 0.3670  data: 0.0028  max mem: 57344
+eval (train): [20]  [ 40/509]  eta: 0:02:57    time: 0.3671  data: 0.0031  max mem: 57344
+eval (train): [20]  [ 60/509]  eta: 0:02:48    time: 0.3672  data: 0.0031  max mem: 57344
+eval (train): [20]  [ 80/509]  eta: 0:02:40    time: 0.3681  data: 0.0031  max mem: 57344
+eval (train): [20]  [100/509]  eta: 0:02:32    time: 0.3677  data: 0.0031  max mem: 57344
+eval (train): [20]  [120/509]  eta: 0:02:24    time: 0.3676  data: 0.0031  max mem: 57344
+eval (train): [20]  [140/509]  eta: 0:02:16    time: 0.3687  data: 0.0034  max mem: 57344
+eval (train): [20]  [160/509]  eta: 0:02:09    time: 0.3694  data: 0.0037  max mem: 57344
+eval (train): [20]  [180/509]  eta: 0:02:01    time: 0.3697  data: 0.0038  max mem: 57344
+eval (train): [20]  [200/509]  eta: 0:01:54    time: 0.3699  data: 0.0036  max mem: 57344
+eval (train): [20]  [220/509]  eta: 0:01:47    time: 0.3687  data: 0.0034  max mem: 57344
+eval (train): [20]  [240/509]  eta: 0:01:39    time: 0.3686  data: 0.0031  max mem: 57344
+eval (train): [20]  [260/509]  eta: 0:01:32    time: 0.3689  data: 0.0033  max mem: 57344
+eval (train): [20]  [280/509]  eta: 0:01:24    time: 0.3685  data: 0.0032  max mem: 57344
+eval (train): [20]  [300/509]  eta: 0:01:17    time: 0.3683  data: 0.0033  max mem: 57344
+eval (train): [20]  [320/509]  eta: 0:01:09    time: 0.3690  data: 0.0034  max mem: 57344
+eval (train): [20]  [340/509]  eta: 0:01:02    time: 0.3683  data: 0.0034  max mem: 57344
+eval (train): [20]  [360/509]  eta: 0:00:55    time: 0.3686  data: 0.0034  max mem: 57344
+eval (train): [20]  [380/509]  eta: 0:00:47    time: 0.3687  data: 0.0034  max mem: 57344
+eval (train): [20]  [400/509]  eta: 0:00:40    time: 0.3697  data: 0.0037  max mem: 57344
+eval (train): [20]  [420/509]  eta: 0:00:32    time: 0.3700  data: 0.0038  max mem: 57344
+eval (train): [20]  [440/509]  eta: 0:00:25    time: 0.3701  data: 0.0038  max mem: 57344
+eval (train): [20]  [460/509]  eta: 0:00:18    time: 0.3703  data: 0.0038  max mem: 57344
+eval (train): [20]  [480/509]  eta: 0:00:10    time: 0.3705  data: 0.0038  max mem: 57344
+eval (train): [20]  [500/509]  eta: 0:00:03    time: 0.3689  data: 0.0035  max mem: 57344
+eval (train): [20]  [508/509]  eta: 0:00:00    time: 0.3587  data: 0.0035  max mem: 57344
+eval (train): [20] Total time: 0:03:08 (0.3697 s / it)
+eval (validation): [20]  [ 0/85]  eta: 0:01:30    time: 1.0640  data: 0.7031  max mem: 57344
+eval (validation): [20]  [20/85]  eta: 0:00:26    time: 0.3684  data: 0.0023  max mem: 57344
+eval (validation): [20]  [40/85]  eta: 0:00:17    time: 0.3707  data: 0.0035  max mem: 57344
+eval (validation): [20]  [60/85]  eta: 0:00:09    time: 0.3703  data: 0.0035  max mem: 57344
+eval (validation): [20]  [80/85]  eta: 0:00:01    time: 0.3692  data: 0.0034  max mem: 57344
+eval (validation): [20]  [84/85]  eta: 0:00:00    time: 0.3627  data: 0.0033  max mem: 57344
+eval (validation): [20] Total time: 0:00:32 (0.3775 s / it)
+eval (test): [20]  [ 0/85]  eta: 0:01:26    time: 1.0235  data: 0.6625  max mem: 57344
+eval (test): [20]  [20/85]  eta: 0:00:26    time: 0.3703  data: 0.0032  max mem: 57344
+eval (test): [20]  [40/85]  eta: 0:00:17    time: 0.3714  data: 0.0036  max mem: 57344
+eval (test): [20]  [60/85]  eta: 0:00:09    time: 0.3710  data: 0.0036  max mem: 57344
+eval (test): [20]  [80/85]  eta: 0:00:01    time: 0.3710  data: 0.0035  max mem: 57344
+eval (test): [20]  [84/85]  eta: 0:00:00    time: 0.3571  data: 0.0034  max mem: 57344
+eval (test): [20] Total time: 0:00:32 (0.3766 s / it)
+eval (testid): [20]  [ 0/82]  eta: 0:01:25    time: 1.0430  data: 0.6828  max mem: 57344
+eval (testid): [20]  [20/82]  eta: 0:00:24    time: 0.3701  data: 0.0023  max mem: 57344
+eval (testid): [20]  [40/82]  eta: 0:00:16    time: 0.3710  data: 0.0034  max mem: 57344
+eval (testid): [20]  [60/82]  eta: 0:00:08    time: 0.3703  data: 0.0032  max mem: 57344
+eval (testid): [20]  [80/82]  eta: 0:00:00    time: 0.3703  data: 0.0032  max mem: 57344
+eval (testid): [20]  [81/82]  eta: 0:00:00    time: 0.3532  data: 0.0032  max mem: 57344
+eval (testid): [20] Total time: 0:00:30 (0.3757 s / it)
+eval results:
+
+| model            | repr   | clf   | dataset      | ckpt   |   epoch |      lr |   wd |   hparam_id | hparam     | split      |   loss |     acc |   acc_std |      f1 |    f1_std |
+|:-----------------|:-------|:------|:-------------|:-------|--------:|--------:|-----:|------------:|:-----------|:-----------|-------:|--------:|----------:|--------:|----------:|
+| schaefer1000_mae | patch  | attn  | nsd_cococlip | best   |      15 | 0.00042 | 0.05 |          26 | [1.4, 1.0] | train      | 2.0326 | 0.38262 | 0.0024539 | 0.33548 | 0.0026038 |
+| schaefer1000_mae | patch  | attn  | nsd_cococlip | best   |      15 | 0.00042 | 0.05 |          26 | [1.4, 1.0] | validation | 2.3741 | 0.29347 | 0.0054772 | 0.24166 | 0.0054356 |
+| schaefer1000_mae | patch  | attn  | nsd_cococlip | best   |      15 | 0.00042 | 0.05 |          26 | [1.4, 1.0] | test       | 2.2722 | 0.30204 | 0.0052876 | 0.23556 | 0.0050929 |
+| schaefer1000_mae | patch  | attn  | nsd_cococlip | best   |      15 | 0.00042 | 0.05 |          26 | [1.4, 1.0] | testid     | 2.3207 | 0.28861 | 0.0054472 | 0.23958 | 0.0053363 |
+
+
+done! total time: 1:43:38

schaefer1000/schaefer1000_lr3e-4_2/pretrain/config.yaml

@@ -0,0 +1,102 @@
+name: schaefer1000/schaefer1000_lr3e-4_2/pretrain
+notes: schaefer1000 ablation schaefer1000_lr3e-4_2 (input_space=schaefer1000 base_lr=3e-4
+  seed=5402)
+output_dir: experiments/schaefer1000/output/schaefer1000/schaefer1000_lr3e-4_2/pretrain
+input_space: schaefer1000
+patch_size: 1
+num_frames: 16
+t_patch_size: 4
+mask_ratio: 0.9
+pred_mask_ratio: null
+masking: tube
+masking_kwargs: {}
+mask_patch_size: null
+model: mae_vit_base
+model_kwargs:
+  decoding: attn
+  pos_embed: sep
+  target_norm: null
+  pca_norm_nc: 2
+  t_pred_stride: 2
+  no_decode_pos: true
+  mask_drop_scale: false
+  pred_edge_pad: 0
+  gauss_sigma: null
+  class_token: true
+  reg_tokens: 0
+  no_embed_class: true
+  head_init_scale: 0.0
+  decoder_depth: 4
+  drop_path_rate: 0.0
+datasets:
+  hcp-train:
+    type: wds
+    url: /data/fmri-datasets/pretrain/hcpya-all.${input_space}.wds/hcpya-all-${input_space}-{00000..01799}.tar
+    clipping: random
+    clipping_kwargs:
+      oversample: 4.0
+    shuffle: true
+    buffer_size: 2000
+    samples_per_epoch: 200000
+  hcp-train-subset:
+    type: arrow
+    root: s3://medarc/fmri-datasets/eval/hcpya-clips.${input_space}.arrow/validation
+    split_range:
+    - 0
+    - 2000
+    shuffle: false
+  hcp-val:
+    type: arrow
+    root: s3://medarc/fmri-datasets/eval/hcpya-clips.${input_space}.arrow/test
+    split_range:
+    - 0
+    - 2000
+    shuffle: false
+train_dataset: hcp-train
+eval_datasets:
+- hcp-train-subset
+- hcp-val
+val_dataset: null
+clip_vmax: 3.0
+normalize: frame
+tr_scale: null
+crop_scale: null
+crop_aspect: null
+gray_jitter: null
+num_workers: 16
+epochs: 100
+batch_size: 32
+accum_iter: 1
+base_lr: 0.0003
+min_lr: 0.0
+warmup_epochs: 5
+weight_decay: 0.05
+betas:
+- 0.9
+- 0.95
+clip_grad: 1.0
+amp: true
+amp_dtype: float16
+ckpt: null
+resume: true
+auto_resume: true
+start_epoch: 0
+max_checkpoints: 0
+checkpoint_period: null
+plot_period: 5
+device: cuda
+presend_cuda: false
+seed: 5402
+debug: false
+wandb: true
+wandb_entity: null
+wandb_project: fMRI-foundation-model
+rank: 0
+world_size: 1
+gpu: 0
+distributed: true
+dist_backend: nccl
+in_chans: 1
+img_size:
+- 1000
+- 1

schaefer1000/schaefer1000_lr3e-4_2/pretrain/log.json

@@ -0,0 +1,100 @@
+{"epoch": 0, "train/lr": 3.7507200230407366e-06, "train/grad": 1.916938072987045, "train/loss": 0.9862997369480133, "eval/hcp-train-subset/loss": 0.9685308077642995, "eval/hcp-val/loss": 0.9643154682651642}
+{"epoch": 1, "train/lr": 1.125096003072098e-05, "train/grad": 3.805821893228726, "train/loss": 0.9195746506977082, "eval/hcp-train-subset/loss": 0.8823787502704128, "eval/hcp-val/loss": 0.8765721465310743}
+{"epoch": 2, "train/lr": 1.875120003840123e-05, "train/grad": 3.3612876590460576, "train/loss": 0.8153612022304535, "eval/hcp-train-subset/loss": 0.7933287370589471, "eval/hcp-val/loss": 0.7836889566913727}
+{"epoch": 3, "train/lr": 2.625144004608147e-05, "train/grad": 2.0869403773161705, "train/loss": 0.7738703588485718, "eval/hcp-train-subset/loss": 0.7686068684824051, "eval/hcp-val/loss": 0.7597825056122195}
+{"epoch": 4, "train/lr": 3.375167986175557e-05, "train/grad": 1.3256996988315433, "train/loss": 0.7633260231208802, "eval/hcp-train-subset/loss": 0.7629020675536125, "eval/hcp-val/loss": 0.7528399104072202}
+{"epoch": 5, "train/lr": 3.749658191365373e-05, "train/grad": 0.8846607681024508, "train/loss": 0.7515889715290069, "eval/hcp-train-subset/loss": 0.751378366062718, "eval/hcp-val/loss": 0.7431896896131577}
+{"epoch": 6, "train/lr": 3.74760811613377e-05, "train/grad": 0.6802200482242259, "train/loss": 0.7428858069038391, "eval/hcp-train-subset/loss": 0.7463900985256318, "eval/hcp-val/loss": 0.7374620591440508}
+{"epoch": 7, "train/lr": 3.743510371420547e-05, "train/grad": 0.5756676039286226, "train/loss": 0.737853164653778, "eval/hcp-train-subset/loss": 0.7420411225288145, "eval/hcp-val/loss": 0.7320226123256068}
+{"epoch": 8, "train/lr": 3.73736943804934e-05, "train/grad": 0.5153294877028725, "train/loss": 0.733797705116272, "eval/hcp-train-subset/loss": 0.7385159730911255, "eval/hcp-val/loss": 0.7287973751944881}
+{"epoch": 9, "train/lr": 3.7291920310406644e-05, "train/grad": 0.45903559385709275, "train/loss": 0.7290213037776947, "eval/hcp-train-subset/loss": 0.7352633870417072, "eval/hcp-val/loss": 0.7264124366544908}
+{"epoch": 10, "train/lr": 3.718987092269037e-05, "train/grad": 0.4320600898876767, "train/loss": 0.7272926388168335, "eval/hcp-train-subset/loss": 0.7315984839393247, "eval/hcp-val/loss": 0.7225010712300578}
+{"epoch": 11, "train/lr": 3.706765780685143e-05, "train/grad": 0.4115456699256384, "train/loss": 0.7216341154384613, "eval/hcp-train-subset/loss": 0.7293285083386206, "eval/hcp-val/loss": 0.719501297320089}
+{"epoch": 12, "train/lr": 3.692541460113792e-05, "train/grad": 0.3930596503779013, "train/loss": 0.7213971669864655, "eval/hcp-train-subset/loss": 0.7261053006495198, "eval/hcp-val/loss": 0.7169477074376999}
+{"epoch": 13, "train/lr": 3.6763296846406675e-05, "train/grad": 0.38437780365288227, "train/loss": 0.7170913793754577, "eval/hcp-train-subset/loss": 0.7254484209322161, "eval/hcp-val/loss": 0.7162821023694931}
+{"epoch": 14, "train/lr": 3.658148181604263e-05, "train/grad": 0.3768947189791557, "train/loss": 0.7131455250167846, "eval/hcp-train-subset/loss": 0.7232727600682166, "eval/hcp-val/loss": 0.7138852650119413}
+{"epoch": 15, "train/lr": 3.6380168322111824e-05, "train/grad": 0.36852344632034245, "train/loss": 0.7137333209228516, "eval/hcp-train-subset/loss": 0.7215865215947551, "eval/hcp-val/loss": 0.7136378826633576}
+{"epoch": 16, "train/lr": 3.615957649796421e-05, "train/grad": 0.36543878670137914, "train/loss": 0.7126937080287933, "eval/hcp-train-subset/loss": 0.7207490888334089, "eval/hcp-val/loss": 0.7125403794550127}
+{"epoch": 17, "train/lr": 3.591994755752113e-05, "train/grad": 0.36124698885950407, "train/loss": 0.7090734524822235, "eval/hcp-train-subset/loss": 0.7190461600980451, "eval/hcp-val/loss": 0.7103699032337435}
+{"epoch": 18, "train/lr": 3.5661543531510486e-05, "train/grad": 0.365342448802305, "train/loss": 0.7064539412498474, "eval/hcp-train-subset/loss": 0.7197593383250698, "eval/hcp-val/loss": 0.7110550951573157}
+{"epoch": 19, "train/lr": 3.538464698094067e-05, "train/grad": 0.3633837124858453, "train/loss": 0.7041006182193756, "eval/hcp-train-subset/loss": 0.717857793454201, "eval/hcp-val/loss": 0.7095602943051246}
+{"epoch": 20, "train/lr": 3.508956068812486e-05, "train/grad": 0.3550558561551317, "train/loss": 0.7037760426521301, "eval/hcp-train-subset/loss": 0.7158851104397927, "eval/hcp-val/loss": 0.7092271456795354}
+{"epoch": 21, "train/lr": 3.4776607325591504e-05, "train/grad": 0.3589969844995202, "train/loss": 0.702181331987381, "eval/hcp-train-subset/loss": 0.7153184269705126, "eval/hcp-val/loss": 0.7079559535749497}
+{"epoch": 22, "train/lr": 3.4446129103247903e-05, "train/grad": 0.3520438471507583, "train/loss": 0.7025209135532379, "eval/hcp-train-subset/loss": 0.7163743146004216, "eval/hcp-val/loss": 0.7076330156095566}
+{"epoch": 23, "train/lr": 3.4098487394178203e-05, "train/grad": 0.3570449557355142, "train/loss": 0.6992085199069977, "eval/hcp-train-subset/loss": 0.7151443919827861, "eval/hcp-val/loss": 0.7061150343187393}
+{"epoch": 24, "train/lr": 3.37340623394871e-05, "train/grad": 0.3632052854354694, "train/loss": 0.6974006106281281, "eval/hcp-train-subset/loss": 0.7136944955395114, "eval/hcp-val/loss": 0.7067666678659378}
+{"epoch": 25, "train/lr": 3.335325243262167e-05, "train/grad": 0.35764560081709934, "train/loss": 0.6972020666599273, "eval/hcp-train-subset/loss": 0.7110791254428125, "eval/hcp-val/loss": 0.7050168543092666}
+{"epoch": 26, "train/lr": 3.295647408362393e-05, "train/grad": 0.35445368981216363, "train/loss": 0.697997944726944, "eval/hcp-train-subset/loss": 0.7124823293378276, "eval/hcp-val/loss": 0.7060774776243395}
+{"epoch": 27, "train/lr": 3.254416116379294e-05, "train/grad": 0.3511945084608019, "train/loss": 0.6960366401576996, "eval/hcp-train-subset/loss": 0.7117950724017236, "eval/hcp-val/loss": 0.7057939421746039}
+{"epoch": 28, "train/lr": 3.211676453125335e-05, "train/grad": 0.36387992564517824, "train/loss": 0.6934515490818024, "eval/hcp-train-subset/loss": 0.7122486070279153, "eval/hcp-val/loss": 0.7047717869281769}
+{"epoch": 29, "train/lr": 3.1674751537947714e-05, "train/grad": 0.3570824798553757, "train/loss": 0.6909300983333587, "eval/hcp-train-subset/loss": 0.7117206646550086, "eval/hcp-val/loss": 0.7063436373587577}
+{"epoch": 30, "train/lr": 3.1218605518594614e-05, "train/grad": 0.3599425611785837, "train/loss": 0.6953045038318634, "eval/hcp-train-subset/loss": 0.7106958358518539, "eval/hcp-val/loss": 0.703712293217259}
+{"epoch": 31, "train/lr": 3.074882526216893e-05, "train/grad": 0.3642111334664088, "train/loss": 0.6904648206615448, "eval/hcp-train-subset/loss": 0.7100344982839399, "eval/hcp-val/loss": 0.704025129156728}
+{"epoch": 32, "train/lr": 3.0265924466484304e-05, "train/grad": 0.3623973691237377, "train/loss": 0.6908761856746674, "eval/hcp-train-subset/loss": 0.7083359543354281, "eval/hcp-val/loss": 0.7024264700951115}
+{"epoch": 33, "train/lr": 2.977043117647054e-05, "train/grad": 0.35899853095238127, "train/loss": 0.6917249834918976, "eval/hcp-train-subset/loss": 0.7102101685539368, "eval/hcp-val/loss": 0.7035958738096298}
+{"epoch": 34, "train/lr": 2.9262887206765643e-05, "train/grad": 0.3652678847174556, "train/loss": 0.6888579062652588, "eval/hcp-train-subset/loss": 0.7063763833815052, "eval/hcp-val/loss": 0.7036237303287752}
+{"epoch": 35, "train/lr": 2.8743847549248834e-05, "train/grad": 0.369643939303287, "train/loss": 0.6886162902259827, "eval/hcp-train-subset/loss": 0.708023892294976, "eval/hcp-val/loss": 0.7031550590069063}
+{"epoch": 36, "train/lr": 2.821387976616516e-05, "train/grad": 0.3707799940218406, "train/loss": 0.6880217561912536, "eval/hcp-train-subset/loss": 0.7056415494411222, "eval/hcp-val/loss": 0.7027493580695121}
+{"epoch": 37, "train/lr": 2.7673563369504416e-05, "train/grad": 0.37069646798507183, "train/loss": 0.6854161455535889, "eval/hcp-train-subset/loss": 0.707313847157263, "eval/hcp-val/loss": 0.7029224999489323}
+{"epoch": 38, "train/lr": 2.7123489187313515e-05, "train/grad": 0.3740484485099815, "train/loss": 0.6863780527496338, "eval/hcp-train-subset/loss": 0.706705738459864, "eval/hcp-val/loss": 0.7024706977029}
+{"epoch": 39, "train/lr": 2.6564258717634662e-05, "train/grad": 0.3767606872044475, "train/loss": 0.6837218506717682, "eval/hcp-train-subset/loss": 0.7053717913166169, "eval/hcp-val/loss": 0.70099711994971}
+{"epoch": 40, "train/lr": 2.5996483470775956e-05, "train/grad": 0.38082504321360217, "train/loss": 0.6839035074806213, "eval/hcp-train-subset/loss": 0.7058885703163762, "eval/hcp-val/loss": 0.7020909901588194}
+{"epoch": 41, "train/lr": 2.5420784300633598e-05, "train/grad": 0.38446478720749017, "train/loss": 0.6836485733795166, "eval/hcp-train-subset/loss": 0.7049921924068082, "eval/hcp-val/loss": 0.7001899807683883}
+{"epoch": 42, "train/lr": 2.4837790725798037e-05, "train/grad": 0.38193810802374684, "train/loss": 0.6831202759456635, "eval/hcp-train-subset/loss": 0.7034412449406039, "eval/hcp-val/loss": 0.7001965315111222}
+{"epoch": 43, "train/lr": 2.4248140241183872e-05, "train/grad": 0.38203057250547967, "train/loss": 0.6820481398868561, "eval/hcp-train-subset/loss": 0.7030651050229226, "eval/hcp-val/loss": 0.7005285716825916}
+{"epoch": 44, "train/lr": 2.36524776209388e-05, "train/grad": 0.3897447898010896, "train/loss": 0.6825762571144104, "eval/hcp-train-subset/loss": 0.7036509869560119, "eval/hcp-val/loss": 0.700147605711414}
+{"epoch": 45, "train/lr": 2.305145421339275e-05, "train/grad": 0.39742657849936364, "train/loss": 0.6763025147056579, "eval/hcp-train-subset/loss": 0.7045472196994289, "eval/hcp-val/loss": 0.6998499756859194}
+{"epoch": 46, "train/lr": 2.2445727228817648e-05, "train/grad": 0.3928512307344598, "train/loss": 0.6820516934394837, "eval/hcp-train-subset/loss": 0.704522302073817, "eval/hcp-val/loss": 0.700247710750949}
+{"epoch": 47, "train/lr": 2.1835959020777867e-05, "train/grad": 0.3945744762100829, "train/loss": 0.677886083278656, "eval/hcp-train-subset/loss": 0.7030071741150271, "eval/hcp-val/loss": 0.6996841642164415}
+{"epoch": 48, "train/lr": 2.1222816361856524e-05, "train/grad": 0.3920815069236163, "train/loss": 0.680115670452118, "eval/hcp-train-subset/loss": 0.7022670766999645, "eval/hcp-val/loss": 0.6994514032717674}
+{"epoch": 49, "train/lr": 2.0606969714549076e-05, "train/grad": 0.3935913417064344, "train/loss": 0.679867234840393, "eval/hcp-train-subset/loss": 0.7012997413835218, "eval/hcp-val/loss": 0.6992642667985731}
+{"epoch": 50, "train/lr": 1.9989092498122458e-05, "train/grad": 0.4042856577836292, "train/loss": 0.6786665251064301, "eval/hcp-train-subset/loss": 0.7015336280868899, "eval/hcp-val/loss": 0.698078469884011}
+{"epoch": 51, "train/lr": 1.936986035224125e-05, "train/grad": 0.3958853328034308, "train/loss": 0.6794172270298005, "eval/hcp-train-subset/loss": 0.6997628519611974, "eval/hcp-val/loss": 0.6993282571915658}
+{"epoch": 52, "train/lr": 1.8749950398165728e-05, "train/grad": 0.40755475804922925, "train/loss": 0.6782050827884674, "eval/hcp-train-subset/loss": 0.7000859408609329, "eval/hcp-val/loss": 0.698251107046681}
+{"epoch": 53, "train/lr": 1.8130040498328958e-05, "train/grad": 0.408110498776145, "train/loss": 0.6761111216640472, "eval/hcp-train-subset/loss": 0.700757872673773, "eval/hcp-val/loss": 0.6968994909717191}
+{"epoch": 54, "train/lr": 1.751080851510517e-05, "train/grad": 0.41968068930842084, "train/loss": 0.6746697693347931, "eval/hcp-train-subset/loss": 0.7012811014729161, "eval/hcp-val/loss": 0.6977885721191284}
+{"epoch": 55, "train/lr": 1.689293156957646e-05, "train/grad": 0.4154893259763985, "train/loss": 0.6747362383270263, "eval/hcp-train-subset/loss": 0.6980924808209942, "eval/hcp-val/loss": 0.6972690676489184}
+{"epoch": 56, "train/lr": 1.627708530111125e-05, "train/grad": 0.420457946782915, "train/loss": 0.6719725515556335, "eval/hcp-train-subset/loss": 0.6993140097587339, "eval/hcp-val/loss": 0.6975935976351461}
+{"epoch": 57, "train/lr": 1.566394312856223e-05, "train/grad": 0.4377157059123873, "train/loss": 0.6740575187206268, "eval/hcp-train-subset/loss": 0.6979264751557381, "eval/hcp-val/loss": 0.6965147324146763}
+{"epoch": 58, "train/lr": 1.5054175513893184e-05, "train/grad": 0.4328440581275467, "train/loss": 0.6734611371135711, "eval/hcp-train-subset/loss": 0.6971425721722264, "eval/hcp-val/loss": 0.6961567190385634}
+{"epoch": 59, "train/lr": 1.444844922903809e-05, "train/grad": 0.4333243721653905, "train/loss": 0.6741378441143036, "eval/hcp-train-subset/loss": 0.6953672247548257, "eval/hcp-val/loss": 0.6965533062334983}
+{"epoch": 60, "train/lr": 1.3847426626796371e-05, "train/grad": 0.43694938137345063, "train/loss": 0.6730795403671265, "eval/hcp-train-subset/loss": 0.6949974817614402, "eval/hcp-val/loss": 0.6953015615863185}
+{"epoch": 61, "train/lr": 1.3251764916559323e-05, "train/grad": 0.4299473457284131, "train/loss": 0.6772745450496673, "eval/hcp-train-subset/loss": 0.6953015298612656, "eval/hcp-val/loss": 0.6957195218532316}
+{"epoch": 62, "train/lr": 1.2662115445661782e-05, "train/grad": 0.4528386157013775, "train/loss": 0.6708118471240997, "eval/hcp-train-subset/loss": 0.6953872326881655, "eval/hcp-val/loss": 0.694636445853018}
+{"epoch": 63, "train/lr": 1.2079122987143025e-05, "train/grad": 0.4616197344396941, "train/loss": 0.6699581180477142, "eval/hcp-train-subset/loss": 0.6962540745735168, "eval/hcp-val/loss": 0.6954789873092405}
+{"epoch": 64, "train/lr": 1.1503425034696866e-05, "train/grad": 0.4442125744330631, "train/loss": 0.6715892050743103, "eval/hcp-train-subset/loss": 0.6927498357911264, "eval/hcp-val/loss": 0.6945943178669098}
+{"epoch": 65, "train/lr": 1.0935651105582284e-05, "train/grad": 0.45917591268451036, "train/loss": 0.6692270779514313, "eval/hcp-train-subset/loss": 0.6926736831665039, "eval/hcp-val/loss": 0.6942044352331469}
+{"epoch": 66, "train/lr": 1.0376422052254546e-05, "train/grad": 0.45504599688471803, "train/loss": 0.6723699559974671, "eval/hcp-train-subset/loss": 0.6919098640641859, "eval/hcp-val/loss": 0.6928566069372238}
+{"epoch": 67, "train/lr": 9.82634938347307e-06, "train/grad": 0.4488341544448337, "train/loss": 0.6711826312732696, "eval/hcp-train-subset/loss": 0.6918799944462315, "eval/hcp-val/loss": 0.6937045301160505}
+{"epoch": 68, "train/lr": 9.286034595625e-06, "train/grad": 0.4549824775257587, "train/loss": 0.6733862752723694, "eval/hcp-train-subset/loss": 0.6913052258952972, "eval/hcp-val/loss": 0.6937089850825648}
+{"epoch": 69, "train/lr": 8.756068514998114e-06, "train/grad": 0.4596974102615986, "train/loss": 0.6723313470363617, "eval/hcp-train-subset/loss": 0.6911922145274377, "eval/hcp-val/loss": 0.6920118658773361}
+{"epoch": 70, "train/lr": 8.237030651720665e-06, "train/grad": 0.4631938287615299, "train/loss": 0.6746887914466858, "eval/hcp-train-subset/loss": 0.6919196113463371, "eval/hcp-val/loss": 0.6922034284760875}
+{"epoch": 71, "train/lr": 7.729488566075681e-06, "train/grad": 0.4812218841503425, "train/loss": 0.6697557126617432, "eval/hcp-train-subset/loss": 0.6893978753397542, "eval/hcp-val/loss": 0.6928907171372445}
+{"epoch": 72, "train/lr": 7.233997247882178e-06, "train/grad": 0.4848380787834697, "train/loss": 0.6663172924613953, "eval/hcp-train-subset/loss": 0.6895911068685593, "eval/hcp-val/loss": 0.6930895274685275}
+{"epoch": 73, "train/lr": 6.751098509621876e-06, "train/grad": 0.4778346759244044, "train/loss": 0.6714987691688538, "eval/hcp-train-subset/loss": 0.6884821211138079, "eval/hcp-val/loss": 0.6933126372675742}
+{"epoch": 74, "train/lr": 6.281320393975191e-06, "train/grad": 0.4854289969088326, "train/loss": 0.6722929624557495, "eval/hcp-train-subset/loss": 0.687444257159387, "eval/hcp-val/loss": 0.6918373703956604}
+{"epoch": 75, "train/lr": 5.825176596414363e-06, "train/grad": 0.4933064593989429, "train/loss": 0.6693909877538681, "eval/hcp-train-subset/loss": 0.6857752848056055, "eval/hcp-val/loss": 0.6907447826477789}
+{"epoch": 76, "train/lr": 5.3831659034850535e-06, "train/grad": 0.4930090991715768, "train/loss": 0.6700630724143982, "eval/hcp-train-subset/loss": 0.6860001702462474, "eval/hcp-val/loss": 0.6903151667887165}
+{"epoch": 77, "train/lr": 4.955771647390373e-06, "train/grad": 0.5067207231526185, "train/loss": 0.668974022064209, "eval/hcp-train-subset/loss": 0.6854748052935447, "eval/hcp-val/loss": 0.6912041546836976}
+{"epoch": 78, "train/lr": 4.543461177474475e-06, "train/grad": 0.5156114799389079, "train/loss": 0.6679129597854614, "eval/hcp-train-subset/loss": 0.6862339108220993, "eval/hcp-val/loss": 0.6909287360406691}
+{"epoch": 79, "train/lr": 4.146685349182691e-06, "train/grad": 0.4939910366740102, "train/loss": 0.6709867037296295, "eval/hcp-train-subset/loss": 0.6842609220935453, "eval/hcp-val/loss": 0.6899335893892473}
+{"epoch": 80, "train/lr": 3.7658780310578768e-06, "train/grad": 0.5090952794593443, "train/loss": 0.6692469611263275, "eval/hcp-train-subset/loss": 0.6838395076413308, "eval/hcp-val/loss": 0.6891084938280044}
+{"epoch": 81, "train/lr": 3.401455630311446e-06, "train/grad": 0.5068261019416822, "train/loss": 0.6704167773723603, "eval/hcp-train-subset/loss": 0.6841812431812286, "eval/hcp-val/loss": 0.689048437341567}
+{"epoch": 82, "train/lr": 3.0538166374881993e-06, "train/grad": 0.52340964429356, "train/loss": 0.6681853998851777, "eval/hcp-train-subset/loss": 0.6831861436367035, "eval/hcp-val/loss": 0.689653750388853}
+{"epoch": 83, "train/lr": 2.723341190722833e-06, "train/grad": 0.5251423914511884, "train/loss": 0.6689264648723602, "eval/hcp-train-subset/loss": 0.6813407886412836, "eval/hcp-val/loss": 0.6908283454756583}
+{"epoch": 84, "train/lr": 2.410390660064364e-06, "train/grad": 0.5119655285265596, "train/loss": 0.6709019479942322, "eval/hcp-train-subset/loss": 0.6814996417491667, "eval/hcp-val/loss": 0.6889690987525448}
+{"epoch": 85, "train/lr": 2.115307252323394e-06, "train/grad": 0.5026329432101913, "train/loss": 0.6719138154315949, "eval/hcp-train-subset/loss": 0.680832956106432, "eval/hcp-val/loss": 0.6884942410453674}
+{"epoch": 86, "train/lr": 1.838413636873912e-06, "train/grad": 0.526731999922852, "train/loss": 0.6713762506580353, "eval/hcp-train-subset/loss": 0.680088795000507, "eval/hcp-val/loss": 0.6879081283846209}
+{"epoch": 87, "train/lr": 1.5800125928190943e-06, "train/grad": 0.5367714690316748, "train/loss": 0.6682818307209015, "eval/hcp-train-subset/loss": 0.6803552489126882, "eval/hcp-val/loss": 0.688087671995163}
+{"epoch": 88, "train/lr": 1.3403866779068222e-06, "train/grad": 0.5232636613169918, "train/loss": 0.6749984820842743, "eval/hcp-train-subset/loss": 0.67921832011592, "eval/hcp-val/loss": 0.6889646735883528}
+{"epoch": 89, "train/lr": 1.1197979195568888e-06, "train/grad": 0.5353258199921835, "train/loss": 0.6716718782329559, "eval/hcp-train-subset/loss": 0.6801128887361095, "eval/hcp-val/loss": 0.688561596216694}
+{"epoch": 90, "train/lr": 9.184875283379039e-07, "train/grad": 0.544578510495177, "train/loss": 0.6700597220611573, "eval/hcp-train-subset/loss": 0.679934787173425, "eval/hcp-val/loss": 0.6885321361403312}
+{"epoch": 91, "train/lr": 7.366756342070463e-07, "train/grad": 0.5540215191467869, "train/loss": 0.670339083981514, "eval/hcp-train-subset/loss": 0.6782205662419719, "eval/hcp-val/loss": 0.6869672363804232}
+{"epoch": 92, "train/lr": 5.745610458012273e-07, "train/grad": 0.5439917946054551, "train/loss": 0.6712316647434234, "eval/hcp-train-subset/loss": 0.6777405421580037, "eval/hcp-val/loss": 0.6871359002205634}
+{"epoch": 93, "train/lr": 4.323210330427178e-07, "train/grad": 0.5440463166154453, "train/loss": 0.6746399412441254, "eval/hcp-train-subset/loss": 0.6767519991244039, "eval/hcp-val/loss": 0.6867612936804371}
+{"epoch": 94, "train/lr": 3.1011113329712343e-07, "train/grad": 0.5512944349617467, "train/loss": 0.675533647480011, "eval/hcp-train-subset/loss": 0.6774758340850953, "eval/hcp-val/loss": 0.6872341517479189}
+{"epoch": 95, "train/lr": 2.080649812955481e-07, "train/grad": 0.5641257733192472, "train/loss": 0.6752789007854462, "eval/hcp-train-subset/loss": 0.6770841142823619, "eval/hcp-val/loss": 0.6873802119685758}
+{"epoch": 96, "train/lr": 1.2629416300698208e-07, "train/grad": 0.5617365404896715, "train/loss": 0.6748136118888854, "eval/hcp-train-subset/loss": 0.6766494426035112, "eval/hcp-val/loss": 0.6875713679098314}
+{"epoch": 97, "train/lr": 6.488809362067338e-08, "train/grad": 0.5793533594737688, "train/loss": 0.6780129624843597, "eval/hcp-train-subset/loss": 0.6765355583160154, "eval/hcp-val/loss": 0.6869341144638677}
+{"epoch": 98, "train/lr": 2.391391977194211e-08, "train/grad": 0.5650060339168993, "train/loss": 0.6767508592510223, "eval/hcp-train-subset/loss": 0.676137400250281, "eval/hcp-val/loss": 0.686227640798015}
+{"epoch": 99, "train/lr": 3.4164461183156008e-09, "train/grad": 0.5684183976097604, "train/loss": 0.6773577667045593, "eval/hcp-train-subset/loss": 0.6764088974845025, "eval/hcp-val/loss": 0.6867091617276592}

schaefer1000/schaefer1000_lr3e-4_3/eval_v2/nsd_cococlip__patch__attn/config.yaml

@@ -0,0 +1,96 @@
+output_root: experiments/schaefer1000/output
+name_prefix: eval_probe
+remote_root: null
+notes: schaefer1000 ablation schaefer1000_lr3e-4_3; eval v2 (nsd_cococlip patch attn)
+model_kwargs:
+  ckpt_path: experiments/schaefer1000/output/schaefer1000/schaefer1000_lr3e-4_3/pretrain/checkpoint-last.pth
+dataset_kwargs: {}
+classifier_kwargs:
+  embed_dim: null
+  dropout: 0.0
+  xavier_init: true
+  norm: true
+lr_scale_grid:
+- 0.02
+- 0.023
+- 0.028
+- 0.033
+- 0.038
+- 0.045
+- 0.053
+- 0.062
+- 0.074
+- 0.087
+- 0.1
+- 0.12
+- 0.14
+- 0.17
+- 0.2
+- 0.23
+- 0.27
+- 0.32
+- 0.38
+- 0.44
+- 0.52
+- 0.61
+- 0.72
+- 0.85
+- 1
+- 1.2
+- 1.4
+- 1.6
+- 1.9
+- 2.3
+- 2.7
+- 3.1
+- 3.7
+- 4.3
+- 5.1
+- 6
+- 7.1
+- 8.3
+- 9.8
+- 12
+- 14
+- 16
+- 19
+- 22
+- 26
+- 31
+- 36
+- 43
+- 50
+wd_scale_grid:
+- 1.0
+num_workers: 8
+prefetch_factor: null
+balanced_sampling: false
+epochs: 20
+steps_per_epoch: 200
+batch_size: 64
+accum_iter: 2
+lr: 0.0003
+warmup_epochs: 5
+no_decay: false
+weight_decay: 0.05
+clip_grad: 1.0
+metrics:
+- acc
+- f1
+cv_metric: acc
+early_stopping: true
+amp: true
+device: cuda
+seed: 4466
+debug: false
+wandb: false
+wandb_entity: null
+wandb_project: fMRI-fm-eval
+name: schaefer1000/schaefer1000_lr3e-4_3/eval_v2/nsd_cococlip__patch__attn
+model: schaefer1000_mae
+representation: patch
+classifier: attn
+dataset: nsd_cococlip
+distributed: false
+output_dir: experiments/schaefer1000/output/schaefer1000/schaefer1000_lr3e-4_3/eval_v2/nsd_cococlip__patch__attn
+remote_dir: null

schaefer1000/schaefer1000_lr3e-4_3/eval_v2/nsd_cococlip__patch__attn/eval_log.json

@@ -0,0 +1 @@
+{"eval/epoch": 17, "eval/id_best": 25, "eval/lr_best": 0.00035999999999999997, "eval/wd_best": 0.05, "eval/train/loss": 2.0312817096710205, "eval/train/acc": 0.3851992993023756, "eval/train/acc_std": 0.0023882084430476586, "eval/train/f1": 0.3397757783189344, "eval/train/f1_std": 0.0025474968414894764, "eval/validation/loss": 2.3555567264556885, "eval/validation/acc": 0.29549649317091176, "eval/validation/acc_std": 0.005656262933974571, "eval/validation/f1": 0.23705767144629322, "eval/validation/f1_std": 0.005311443378165326, "eval/test/loss": 2.2627780437469482, "eval/test/acc": 0.310760667903525, "eval/test/acc_std": 0.005406987795383281, "eval/test/f1": 0.23728715667592926, "eval/test/f1_std": 0.005175306017947441, "eval/testid/loss": 2.3136305809020996, "eval/testid/acc": 0.29863119336803545, "eval/testid/acc_std": 0.005567020580789525, "eval/testid/f1": 0.2530938303255491, "eval/testid/f1_std": 0.005478202641659925}

schaefer1000/schaefer1000_lr3e-4_3/eval_v2/nsd_cococlip__patch__attn/eval_log_best.json

@@ -0,0 +1 @@
+{"eval/best/epoch": 17, "eval/best/id_best": 25, "eval/best/lr_best": 0.00035999999999999997, "eval/best/wd_best": 0.05, "eval/best/train/loss": 2.0312817096710205, "eval/best/train/acc": 0.3851992993023756, "eval/best/train/acc_std": 0.0023882084430476586, "eval/best/train/f1": 0.3397757783189344, "eval/best/train/f1_std": 0.0025474968414894764, "eval/best/validation/loss": 2.3555567264556885, "eval/best/validation/acc": 0.29549649317091176, "eval/best/validation/acc_std": 0.005656262933974571, "eval/best/validation/f1": 0.23705767144629322, "eval/best/validation/f1_std": 0.005311443378165326, "eval/best/test/loss": 2.2627780437469482, "eval/best/test/acc": 0.310760667903525, "eval/best/test/acc_std": 0.005406987795383281, "eval/best/test/f1": 0.23728715667592926, "eval/best/test/f1_std": 0.005175306017947441, "eval/best/testid/loss": 2.3136305809020996, "eval/best/testid/acc": 0.29863119336803545, "eval/best/testid/acc_std": 0.005567020580789525, "eval/best/testid/f1": 0.2530938303255491, "eval/best/testid/f1_std": 0.005478202641659925}

schaefer1000/schaefer1000_lr3e-4_3/eval_v2/nsd_cococlip__patch__attn/eval_log_last.json

@@ -0,0 +1 @@
+{"eval/last/epoch": 19, "eval/last/id_best": 25, "eval/last/lr_best": 0.00035999999999999997, "eval/last/wd_best": 0.05, "eval/last/train/loss": 2.0264649391174316, "eval/last/train/acc": 0.3847383140231722, "eval/last/train/acc_std": 0.0023872999425713927, "eval/last/train/f1": 0.33749785434090723, "eval/last/train/f1_std": 0.002576343879405602, "eval/last/validation/loss": 2.364795446395874, "eval/last/validation/acc": 0.292358803986711, "eval/last/validation/acc_std": 0.005674728313992087, "eval/last/validation/f1": 0.23463211168118656, "eval/last/validation/f1_std": 0.0053664369394046925, "eval/last/test/loss": 2.268484592437744, "eval/last/test/acc": 0.31001855287569574, "eval/last/test/acc_std": 0.005286715629380667, "eval/last/test/f1": 0.23672795415271178, "eval/last/test/f1_std": 0.005145260629233501, "eval/last/testid/loss": 2.312072992324829, "eval/last/testid/acc": 0.2984384037015616, "eval/last/testid/acc_std": 0.00556355404408502, "eval/last/testid/f1": 0.2512358125847974, "eval/last/testid/f1_std": 0.005527898972483092}

schaefer1000/schaefer1000_lr3e-4_3/eval_v2/nsd_cococlip__patch__attn/eval_table.csv

@@ -0,0 +1,5 @@
+model,repr,clf,dataset,ckpt,epoch,lr,wd,hparam_id,hparam,split,loss,acc,acc_std,f1,f1_std
+schaefer1000_mae,patch,attn,nsd_cococlip,best,17,0.00035999999999999997,0.05,25,"[1.2, 1.0]",train,2.0312817096710205,0.3851992993023756,0.0023882084430476586,0.3397757783189344,0.0025474968414894764
+schaefer1000_mae,patch,attn,nsd_cococlip,best,17,0.00035999999999999997,0.05,25,"[1.2, 1.0]",validation,2.3555567264556885,0.29549649317091176,0.005656262933974571,0.23705767144629322,0.005311443378165326
+schaefer1000_mae,patch,attn,nsd_cococlip,best,17,0.00035999999999999997,0.05,25,"[1.2, 1.0]",test,2.2627780437469482,0.310760667903525,0.005406987795383281,0.23728715667592926,0.005175306017947441
+schaefer1000_mae,patch,attn,nsd_cococlip,best,17,0.00035999999999999997,0.05,25,"[1.2, 1.0]",testid,2.3136305809020996,0.29863119336803545,0.005567020580789525,0.2530938303255491,0.005478202641659925

schaefer1000/schaefer1000_lr3e-4_3/eval_v2/nsd_cococlip__patch__attn/eval_table_best.csv

@@ -0,0 +1,5 @@
+model,repr,clf,dataset,ckpt,epoch,lr,wd,hparam_id,hparam,split,loss,acc,acc_std,f1,f1_std
+schaefer1000_mae,patch,attn,nsd_cococlip,best,17,0.00035999999999999997,0.05,25,"[1.2, 1.0]",train,2.0312817096710205,0.3851992993023756,0.0023882084430476586,0.3397757783189344,0.0025474968414894764
+schaefer1000_mae,patch,attn,nsd_cococlip,best,17,0.00035999999999999997,0.05,25,"[1.2, 1.0]",validation,2.3555567264556885,0.29549649317091176,0.005656262933974571,0.23705767144629322,0.005311443378165326
+schaefer1000_mae,patch,attn,nsd_cococlip,best,17,0.00035999999999999997,0.05,25,"[1.2, 1.0]",test,2.2627780437469482,0.310760667903525,0.005406987795383281,0.23728715667592926,0.005175306017947441
+schaefer1000_mae,patch,attn,nsd_cococlip,best,17,0.00035999999999999997,0.05,25,"[1.2, 1.0]",testid,2.3136305809020996,0.29863119336803545,0.005567020580789525,0.2530938303255491,0.005478202641659925

schaefer1000/schaefer1000_lr3e-4_3/eval_v2/nsd_cococlip__patch__attn/eval_table_last.csv

@@ -0,0 +1,5 @@
+model,repr,clf,dataset,ckpt,epoch,lr,wd,hparam_id,hparam,split,loss,acc,acc_std,f1,f1_std
+schaefer1000_mae,patch,attn,nsd_cococlip,last,19,0.00035999999999999997,0.05,25,"[1.2, 1.0]",train,2.0264649391174316,0.3847383140231722,0.0023872999425713927,0.33749785434090723,0.002576343879405602
+schaefer1000_mae,patch,attn,nsd_cococlip,last,19,0.00035999999999999997,0.05,25,"[1.2, 1.0]",validation,2.364795446395874,0.292358803986711,0.005674728313992087,0.23463211168118656,0.0053664369394046925
+schaefer1000_mae,patch,attn,nsd_cococlip,last,19,0.00035999999999999997,0.05,25,"[1.2, 1.0]",test,2.268484592437744,0.31001855287569574,0.005286715629380667,0.23672795415271178,0.005145260629233501
+schaefer1000_mae,patch,attn,nsd_cococlip,last,19,0.00035999999999999997,0.05,25,"[1.2, 1.0]",testid,2.312072992324829,0.2984384037015616,0.00556355404408502,0.2512358125847974,0.005527898972483092

schaefer1000/schaefer1000_lr3e-4_3/eval_v2/nsd_cococlip__patch__attn/log.txt

@@ -0,0 +1,968 @@
+fMRI foundation model probe eval
+version: 0.1.dev95+g65be98d36
+sha: 87e31aaa465443ed5f0da58176ac8395447cdbd0, status: has uncommitted changes, branch: dev/clane9
+cwd: /data/connor/fmri-fm
+start: 2026-05-12 20:54:49
+config:
+output_root: experiments/schaefer1000/output
+name_prefix: eval_probe
+remote_root: null
+notes: schaefer1000 ablation schaefer1000_lr3e-4_3; eval v2 (nsd_cococlip patch attn)
+model_kwargs:
+  ckpt_path: experiments/schaefer1000/output/schaefer1000/schaefer1000_lr3e-4_3/pretrain/checkpoint-last.pth
+dataset_kwargs: {}
+classifier_kwargs:
+  embed_dim: null
+  dropout: 0.0
+  xavier_init: true
+  norm: true
+lr_scale_grid:
+- 0.02
+- 0.023
+- 0.028
+- 0.033
+- 0.038
+- 0.045
+- 0.053
+- 0.062
+- 0.074
+- 0.087
+- 0.1
+- 0.12
+- 0.14
+- 0.17
+- 0.2
+- 0.23
+- 0.27
+- 0.32
+- 0.38
+- 0.44
+- 0.52
+- 0.61
+- 0.72
+- 0.85
+- 1
+- 1.2
+- 1.4
+- 1.6
+- 1.9
+- 2.3
+- 2.7
+- 3.1
+- 3.7
+- 4.3
+- 5.1
+- 6
+- 7.1
+- 8.3
+- 9.8
+- 12
+- 14
+- 16
+- 19
+- 22
+- 26
+- 31
+- 36
+- 43
+- 50
+wd_scale_grid:
+- 1.0
+num_workers: 8
+prefetch_factor: null
+balanced_sampling: false
+epochs: 20
+steps_per_epoch: 200
+batch_size: 64
+accum_iter: 2
+lr: 0.0003
+warmup_epochs: 5
+no_decay: false
+weight_decay: 0.05
+clip_grad: 1.0
+metrics:
+- acc
+- f1
+cv_metric: acc
+early_stopping: true
+amp: true
+device: cuda
+seed: 4466
+debug: false
+wandb: false
+wandb_entity: null
+wandb_project: fMRI-fm-eval
+name: schaefer1000/schaefer1000_lr3e-4_3/eval_v2/nsd_cococlip__patch__attn
+model: schaefer1000_mae
+representation: patch
+classifier: attn
+dataset: nsd_cococlip
+distributed: false
+output_dir: experiments/schaefer1000/output/schaefer1000/schaefer1000_lr3e-4_3/eval_v2/nsd_cococlip__patch__attn
+remote_dir: null
+
+creating frozen backbone model: schaefer1000_mae
+backbone:
+MaskedEncoderWrapper(
+  (model): MaskedEncoder(
+    class_token=True, reg_tokens=0, no_embed_class=True, mask_drop_scale=False
+    (patchify): Patchify3D((16, 1000, 1), (4, 1, 1), in_chans=1)
+    (patch_embed): Linear(in_features=4, out_features=768, bias=True)
+    (pos_embed): SeparablePosEmbed(768, (4, 1000, 1))
+    (blocks): ModuleList(
+      (0-11): 12 x Block(
+        (norm1): LayerNorm((768,), eps=1e-06, elementwise_affine=True)
+        (attn): Attention(
+          num_heads=12
+          (q): Linear(in_features=768, out_features=768, bias=True)
+          (k): Linear(in_features=768, out_features=768, bias=True)
+          (v): Linear(in_features=768, out_features=768, bias=True)
+          (proj): Linear(in_features=768, out_features=768, bias=True)
+        )
+        (drop_path1): Identity()
+        (norm2): LayerNorm((768,), eps=1e-06, elementwise_affine=True)
+        (mlp): Mlp(
+          (fc1): Linear(in_features=768, out_features=3072, bias=True)
+          (act): GELU(approximate='none')
+          (fc2): Linear(in_features=3072, out_features=768, bias=True)
+        )
+        (drop_path2): Identity()
+      )
+    )
+    (norm): LayerNorm((768,), eps=1e-06, elementwise_affine=True)
+  )
+)
+creating dataset: nsd_cococlip (schaefer1000)
+train (n=32539):
+HFDataset(
+    dataset=Dataset({
+    features: ['sub', 'ses', 'run', 'trial_id', 'nsd_id', 'category_id', 'path', 'start', 'end', 'n_frames', 'tr', 'bold', 'mean', 'std'],
+    num_rows: 32539
+}),
+    labels=[ 3  4  5  6 10 11 17 18 19 20 22 23 25 30 31 33 36 37 38 53 55 59 61 74],
+    counts=[1286 1180 1639 1868  834  824 1026 1042  913 1853 1503 2092 1001 1410
+  794 1241 1904 1872 2267 1428  889  904 1447 1322]
+)
+
+validation (n=5418):
+HFDataset(
+    dataset=Dataset({
+    features: ['sub', 'ses', 'run', 'trial_id', 'nsd_id', 'category_id', 'path', 'start', 'end', 'n_frames', 'tr', 'bold', 'mean', 'std'],
+    num_rows: 5418
+}),
+    labels=[ 3  4  5  6 10 11 17 18 19 20 22 23 25 30 31 33 36 37 38 53 55 59 61 74],
+    counts=[197 161 276 345 126 142 143 185 112 295 285 387 169 250 159 193 316 334
+ 343 215 172 141 226 246]
+)
+
+test (n=5390):
+HFDataset(
+    dataset=Dataset({
+    features: ['sub', 'ses', 'run', 'trial_id', 'nsd_id', 'category_id', 'path', 'start', 'end', 'n_frames', 'tr', 'bold', 'mean', 'std'],
+    num_rows: 5390
+}),
+    labels=[ 3  4  5  6 10 11 17 18 19 20 22 23 25 30 31 33 36 37 38 53 55 59 61 74],
+    counts=[202 172 274 298 144 180 134 182 186 293 218 343 165 185 140 177 346 333
+ 345 271 165 140 251 246]
+)
+
+testid (n=5187):
+HFDataset(
+    dataset=Dataset({
+    features: ['sub', 'ses', 'run', 'trial_id', 'nsd_id', 'category_id', 'path', 'start', 'end', 'n_frames', 'tr', 'bold', 'mean', 'std'],
+    num_rows: 5187
+}),
+    labels=[ 3  4  5  6 10 11 17 18 19 20 22 23 25 30 31 33 36 37 38 53 55 59 61 74],
+    counts=[197 159 267 273 123 153 175 184 139 310 215 386 153 230 118 192 330 306
+ 349 223 143 127 249 186]
+)
+
+running backbone on example batch to get embedding dim
+embedding feature dim (patch): 768
+initializing sweep of classifier heads
+classifiers:
+ModuleList(
+  (0-48): 49 x AttnPoolClassifier(
+    (kv): Linear(in_features=768, out_features=1536, bias=True)
+    (linear): Linear(in_features=768, out_features=24, bias=True)
+  )
+)
+classifier params (train): 58.8M (58.8M)
+setting up optimizer
+total batch size: 128 = 64 bs per gpu x 2 accum
+lr: 3.00e-04
+full schedule: epochs = 20 (steps = 4000) (decay = True)
+warmup: epochs = 5 (steps = 1000)
+start training for 20 epochs
+train: [0]  [  0/400]  eta: 0:11:11  lr: nan  time: 1.6796  data: 0.8892  max mem: 56639
+train: [0]  [ 20/400]  eta: 0:04:26  lr: 0.000003  loss: 3.1836 (3.1950)  grad: 0.2014 (0.2086)  time: 0.6530  data: 0.0030  max mem: 57344
+train: [0]  [ 40/400]  eta: 0:04:02  lr: 0.000006  loss: 3.1836 (3.1907)  grad: 0.2014 (0.2015)  time: 0.6436  data: 0.0034  max mem: 57344
+train: [0]  [ 60/400]  eta: 0:03:45  lr: 0.000009  loss: 3.1831 (3.1870)  grad: 0.2057 (0.2045)  time: 0.6435  data: 0.0033  max mem: 57344
+train: [0]  [ 80/400]  eta: 0:03:31  lr: 0.000012  loss: 3.1708 (3.1824)  grad: 0.1990 (0.2025)  time: 0.6466  data: 0.0037  max mem: 57344
+train: [0]  [100/400]  eta: 0:03:17  lr: 0.000015  loss: 3.1653 (3.1797)  grad: 0.1889 (0.2018)  time: 0.6482  data: 0.0039  max mem: 57344
+train: [0]  [120/400]  eta: 0:03:03  lr: 0.000018  loss: 3.1611 (3.1780)  grad: 0.1881 (0.1998)  time: 0.6460  data: 0.0037  max mem: 57344
+train: [0]  [140/400]  eta: 0:02:50  lr: 0.000021  loss: 3.1489 (3.1753)  grad: 0.1929 (0.1991)  time: 0.6458  data: 0.0036  max mem: 57344
+train: [0]  [160/400]  eta: 0:02:36  lr: 0.000024  loss: 3.1518 (3.1743)  grad: 0.1875 (0.1973)  time: 0.6445  data: 0.0035  max mem: 57344
+train: [0]  [180/400]  eta: 0:02:23  lr: 0.000027  loss: 3.1618 (3.1732)  grad: 0.1738 (0.1952)  time: 0.6444  data: 0.0035  max mem: 57344
+train: [0]  [200/400]  eta: 0:02:10  lr: 0.000030  loss: 3.1437 (3.1698)  grad: 0.1743 (0.1942)  time: 0.6449  data: 0.0035  max mem: 57344
+train: [0]  [220/400]  eta: 0:01:57  lr: 0.000033  loss: 3.1423 (3.1677)  grad: 0.1977 (0.1946)  time: 0.6443  data: 0.0035  max mem: 57344
+train: [0]  [240/400]  eta: 0:01:44  lr: 0.000036  loss: 3.1495 (3.1669)  grad: 0.1879 (0.1938)  time: 0.6453  data: 0.0035  max mem: 57344
+train: [0]  [260/400]  eta: 0:01:30  lr: 0.000039  loss: 3.1501 (3.1663)  grad: 0.1773 (0.1926)  time: 0.6442  data: 0.0035  max mem: 57344
+train: [0]  [280/400]  eta: 0:01:17  lr: 0.000042  loss: 3.1396 (3.1642)  grad: 0.1799 (0.1915)  time: 0.6453  data: 0.0035  max mem: 57344
+train: [0]  [300/400]  eta: 0:01:04  lr: 0.000045  loss: 3.1466 (3.1640)  grad: 0.1799 (0.1906)  time: 0.6441  data: 0.0034  max mem: 57344
+train: [0]  [320/400]  eta: 0:00:51  lr: 0.000048  loss: 3.1525 (3.1626)  grad: 0.1844 (0.1908)  time: 0.6437  data: 0.0034  max mem: 57344
+train: [0]  [340/400]  eta: 0:00:38  lr: 0.000051  loss: 3.1345 (3.1614)  grad: 0.1844 (0.1901)  time: 0.6442  data: 0.0034  max mem: 57344
+train: [0]  [360/400]  eta: 0:00:25  lr: 0.000054  loss: 3.1370 (3.1610)  grad: 0.1795 (0.1896)  time: 0.6451  data: 0.0037  max mem: 57344
+train: [0]  [380/400]  eta: 0:00:12  lr: 0.000057  loss: 3.1362 (3.1593)  grad: 0.1753 (0.1890)  time: 0.6461  data: 0.0037  max mem: 57344
+train: [0]  [399/400]  eta: 0:00:00  lr: 0.000060  loss: 3.1302 (3.1576)  grad: 0.1710 (0.1884)  time: 0.6462  data: 0.0038  max mem: 57344
+train: [0] Total time: 0:04:19 (0.6483 s / it)
+train: [0] Summary: lr: 0.000060  loss: 3.1302 (3.1576)  grad: 0.1710 (0.1884)
+eval (validation): [0]  [ 0/85]  eta: 0:01:34    time: 1.1116  data: 0.7528  max mem: 57344
+eval (validation): [0]  [20/85]  eta: 0:00:26    time: 0.3681  data: 0.0029  max mem: 57344
+eval (validation): [0]  [40/85]  eta: 0:00:17    time: 0.3689  data: 0.0038  max mem: 57344
+eval (validation): [0]  [60/85]  eta: 0:00:09    time: 0.3692  data: 0.0037  max mem: 57344
+eval (validation): [0]  [80/85]  eta: 0:00:01    time: 0.3679  data: 0.0036  max mem: 57344
+eval (validation): [0]  [84/85]  eta: 0:00:00    time: 0.3618  data: 0.0035  max mem: 57344
+eval (validation): [0] Total time: 0:00:32 (0.3772 s / it)
+cv: [0]  best hparam: (36, 1.0) (046) ('046_lr3.6e+01_wd1.0e+00')  loss: 2.787  acc: 0.171  f1: 0.106
+saving checkpoint experiments/schaefer1000/output/schaefer1000/schaefer1000_lr3e-4_3/eval_v2/nsd_cococlip__patch__attn/checkpoint-last.pth
+saving best checkpoint experiments/schaefer1000/output/schaefer1000/schaefer1000_lr3e-4_3/eval_v2/nsd_cococlip__patch__attn/checkpoint-best.pth
+train: [1]  [  0/400]  eta: 0:08:44  lr: nan  time: 1.3113  data: 0.6778  max mem: 57344
+train: [1]  [ 20/400]  eta: 0:04:16  lr: 0.000063  loss: 3.0544 (3.0822)  grad: 0.1773 (0.1830)  time: 0.6439  data: 0.0024  max mem: 57344
+train: [1]  [ 40/400]  eta: 0:03:57  lr: 0.000066  loss: 3.0611 (3.0880)  grad: 0.1748 (0.1774)  time: 0.6450  data: 0.0037  max mem: 57344
+train: [1]  [ 60/400]  eta: 0:03:43  lr: 0.000069  loss: 3.1010 (3.0933)  grad: 0.1766 (0.1833)  time: 0.6478  data: 0.0040  max mem: 57344
+train: [1]  [ 80/400]  eta: 0:03:29  lr: 0.000072  loss: 3.0987 (3.0984)  grad: 0.1909 (0.1869)  time: 0.6461  data: 0.0039  max mem: 57344
+train: [1]  [100/400]  eta: 0:03:15  lr: 0.000075  loss: 3.0851 (3.0950)  grad: 0.1952 (0.1890)  time: 0.6459  data: 0.0038  max mem: 57344
+train: [1]  [120/400]  eta: 0:03:02  lr: 0.000078  loss: 3.0712 (3.0913)  grad: 0.1943 (0.1889)  time: 0.6452  data: 0.0037  max mem: 57344
+train: [1]  [140/400]  eta: 0:02:49  lr: 0.000081  loss: 3.0629 (3.0848)  grad: 0.1861 (0.1899)  time: 0.6452  data: 0.0036  max mem: 57344
+train: [1]  [160/400]  eta: 0:02:35  lr: 0.000084  loss: 3.0522 (3.0816)  grad: 0.1899 (0.1902)  time: 0.6447  data: 0.0037  max mem: 57344
+train: [1]  [180/400]  eta: 0:02:22  lr: 0.000087  loss: 3.0533 (3.0814)  grad: 0.1915 (0.1908)  time: 0.6452  data: 0.0036  max mem: 57344
+train: [1]  [200/400]  eta: 0:02:09  lr: 0.000090  loss: 3.0690 (3.0793)  grad: 0.2033 (0.1923)  time: 0.6451  data: 0.0037  max mem: 57344
+train: [1]  [220/400]  eta: 0:01:56  lr: 0.000093  loss: 3.0568 (3.0771)  grad: 0.2040 (0.1932)  time: 0.6457  data: 0.0037  max mem: 57344
+train: [1]  [240/400]  eta: 0:01:43  lr: 0.000096  loss: 3.0272 (3.0728)  grad: 0.1987 (0.1939)  time: 0.6451  data: 0.0036  max mem: 57344
+train: [1]  [260/400]  eta: 0:01:30  lr: 0.000099  loss: 3.0131 (3.0699)  grad: 0.2060 (0.1948)  time: 0.6447  data: 0.0035  max mem: 57344
+train: [1]  [280/400]  eta: 0:01:17  lr: 0.000102  loss: 3.0482 (3.0685)  grad: 0.2124 (0.1964)  time: 0.6448  data: 0.0035  max mem: 57344
+train: [1]  [300/400]  eta: 0:01:04  lr: 0.000105  loss: 3.0377 (3.0649)  grad: 0.2189 (0.1980)  time: 0.6442  data: 0.0035  max mem: 57344
+train: [1]  [320/400]  eta: 0:00:51  lr: 0.000108  loss: 3.0174 (3.0628)  grad: 0.2100 (0.1982)  time: 0.6445  data: 0.0035  max mem: 57344
+train: [1]  [340/400]  eta: 0:00:38  lr: 0.000111  loss: 3.0174 (3.0603)  grad: 0.2161 (0.2009)  time: 0.6469  data: 0.0039  max mem: 57344
+train: [1]  [360/400]  eta: 0:00:25  lr: 0.000114  loss: 3.0346 (3.0642)  grad: 0.2757 (0.2177)  time: 0.6462  data: 0.0037  max mem: 57344
+train: [1]  [380/400]  eta: 0:00:12  lr: 0.000117  loss: 3.2333 (3.0969)  grad: 0.6188 (0.2922)  time: 0.6441  data: 0.0034  max mem: 57344
+WARNING: classifier 48 (50, 1.0) diverged (loss=83.53 > 63.56) at step 391. Freezing.
+train: [1]  [399/400]  eta: 0:00:00  lr: 0.000120  loss: 3.3893 (3.1025)  grad: 0.6188 (0.3025)  time: 0.6395  data: 0.0035  max mem: 57344
+train: [1] Total time: 0:04:18 (0.6470 s / it)
+train: [1] Summary: lr: 0.000120  loss: 3.3893 (3.1025)  grad: 0.6188 (0.3025)
+eval (validation): [1]  [ 0/85]  eta: 0:01:20    time: 0.9503  data: 0.5936  max mem: 57344
+eval (validation): [1]  [20/85]  eta: 0:00:25    time: 0.3671  data: 0.0031  max mem: 57344
+eval (validation): [1]  [40/85]  eta: 0:00:17    time: 0.3692  data: 0.0039  max mem: 57344
+eval (validation): [1]  [60/85]  eta: 0:00:09    time: 0.3691  data: 0.0041  max mem: 57344
+eval (validation): [1]  [80/85]  eta: 0:00:01    time: 0.3689  data: 0.0040  max mem: 57344
+eval (validation): [1]  [84/85]  eta: 0:00:00    time: 0.3622  data: 0.0039  max mem: 57344
+eval (validation): [1] Total time: 0:00:31 (0.3754 s / it)
+cv: [1]  best hparam: (14, 1.0) (040) ('040_lr1.4e+01_wd1.0e+00')  loss: 2.589  acc: 0.217  f1: 0.158
+saving checkpoint experiments/schaefer1000/output/schaefer1000/schaefer1000_lr3e-4_3/eval_v2/nsd_cococlip__patch__attn/checkpoint-last.pth
+saving best checkpoint experiments/schaefer1000/output/schaefer1000/schaefer1000_lr3e-4_3/eval_v2/nsd_cococlip__patch__attn/checkpoint-best.pth
+train: [2]  [  0/400]  eta: 0:08:37  lr: nan  time: 1.2948  data: 0.6681  max mem: 57344
+train: [2]  [ 20/400]  eta: 0:04:15  lr: 0.000123  loss: 2.9646 (2.9868)  grad: 0.1901 (0.1949)  time: 0.6399  data: 0.0034  max mem: 57344
+train: [2]  [ 40/400]  eta: 0:03:56  lr: 0.000126  loss: 2.9624 (2.9686)  grad: 0.1936 (0.1972)  time: 0.6395  data: 0.0037  max mem: 57344
+train: [2]  [ 60/400]  eta: 0:03:41  lr: 0.000129  loss: 2.9529 (2.9687)  grad: 0.2096 (0.2074)  time: 0.6398  data: 0.0037  max mem: 57344
+train: [2]  [ 80/400]  eta: 0:03:27  lr: 0.000132  loss: 2.9764 (2.9775)  grad: 0.2282 (0.2161)  time: 0.6390  data: 0.0036  max mem: 57344
+train: [2]  [100/400]  eta: 0:03:13  lr: 0.000135  loss: 2.9913 (2.9809)  grad: 0.2282 (0.2177)  time: 0.6391  data: 0.0035  max mem: 57344
+train: [2]  [120/400]  eta: 0:03:00  lr: 0.000138  loss: 2.9913 (2.9854)  grad: 0.2341 (0.2219)  time: 0.6390  data: 0.0035  max mem: 57344
+train: [2]  [140/400]  eta: 0:02:47  lr: 0.000141  loss: 3.0092 (2.9942)  grad: 0.2684 (0.2526)  time: 0.6392  data: 0.0035  max mem: 57344
+WARNING: classifier 47 (43, 1.0) diverged (loss=73.73 > 63.56) at step 478. Freezing.
+train: [2]  [160/400]  eta: 0:02:34  lr: 0.000144  loss: 3.0850 (3.0559)  grad: 0.5039 (0.4046)  time: 0.6380  data: 0.0034  max mem: 57344
+train: [2]  [180/400]  eta: 0:02:21  lr: 0.000147  loss: 2.9898 (3.0455)  grad: 0.2675 (0.3869)  time: 0.6334  data: 0.0035  max mem: 57344
+train: [2]  [200/400]  eta: 0:02:08  lr: 0.000150  loss: 2.9720 (3.0456)  grad: 0.2780 (0.3974)  time: 0.6329  data: 0.0034  max mem: 57344
+WARNING: classifier 46 (36, 1.0) diverged (loss=63.99 > 63.56) at step 505. Freezing.
+train: [2]  [220/400]  eta: 0:01:55  lr: 0.000153  loss: 3.0153 (3.0629)  grad: 0.3992 (0.4287)  time: 0.6300  data: 0.0033  max mem: 57344
+train: [2]  [240/400]  eta: 0:01:42  lr: 0.000156  loss: 2.9962 (3.0567)  grad: 0.2157 (0.4113)  time: 0.6269  data: 0.0033  max mem: 57344
+train: [2]  [260/400]  eta: 0:01:29  lr: 0.000159  loss: 2.9772 (3.0511)  grad: 0.2157 (0.3971)  time: 0.6285  data: 0.0038  max mem: 57344
+train: [2]  [280/400]  eta: 0:01:16  lr: 0.000162  loss: 2.9818 (3.0474)  grad: 0.2259 (0.3853)  time: 0.6297  data: 0.0039  max mem: 57344
+train: [2]  [300/400]  eta: 0:01:03  lr: 0.000165  loss: 2.9859 (3.0438)  grad: 0.2245 (0.3751)  time: 0.6297  data: 0.0039  max mem: 57344
+train: [2]  [320/400]  eta: 0:00:50  lr: 0.000168  loss: 2.9901 (3.0411)  grad: 0.2422 (0.3683)  time: 0.6273  data: 0.0034  max mem: 57344
+train: [2]  [340/400]  eta: 0:00:38  lr: 0.000171  loss: 3.0304 (3.0465)  grad: 0.3623 (0.3864)  time: 0.6277  data: 0.0035  max mem: 57344
+train: [2]  [360/400]  eta: 0:00:25  lr: 0.000174  loss: 3.3769 (3.0853)  grad: 0.9698 (0.4560)  time: 0.6276  data: 0.0036  max mem: 57344
+WARNING: classifier 45 (31, 1.0) diverged (loss=83.76 > 63.56) at step 583. Freezing.
+train: [2]  [380/400]  eta: 0:00:12  lr: 0.000177  loss: 3.4685 (3.1026)  grad: 1.1359 (0.4870)  time: 0.6253  data: 0.0039  max mem: 57344
+train: [2]  [399/400]  eta: 0:00:00  lr: 0.000180  loss: 3.2143 (3.1140)  grad: 0.9069 (0.5142)  time: 0.6239  data: 0.0039  max mem: 57344
+train: [2] Total time: 0:04:13 (0.6348 s / it)
+train: [2] Summary: lr: 0.000180  loss: 3.2143 (3.1140)  grad: 0.9069 (0.5142)
+eval (validation): [2]  [ 0/85]  eta: 0:01:37    time: 1.1429  data: 0.7832  max mem: 57344
+eval (validation): [2]  [20/85]  eta: 0:00:26    time: 0.3670  data: 0.0028  max mem: 57344
+eval (validation): [2]  [40/85]  eta: 0:00:17    time: 0.3676  data: 0.0036  max mem: 57344
+eval (validation): [2]  [60/85]  eta: 0:00:09    time: 0.3667  data: 0.0037  max mem: 57344
+eval (validation): [2]  [80/85]  eta: 0:00:01    time: 0.3673  data: 0.0036  max mem: 57344
+eval (validation): [2]  [84/85]  eta: 0:00:00    time: 0.3609  data: 0.0036  max mem: 57344
+eval (validation): [2] Total time: 0:00:31 (0.3761 s / it)
+cv: [2]  best hparam: (8.3, 1.0) (037) ('037_lr8.3e+00_wd1.0e+00')  loss: 2.477  acc: 0.240  f1: 0.178
+saving checkpoint experiments/schaefer1000/output/schaefer1000/schaefer1000_lr3e-4_3/eval_v2/nsd_cococlip__patch__attn/checkpoint-last.pth
+saving best checkpoint experiments/schaefer1000/output/schaefer1000/schaefer1000_lr3e-4_3/eval_v2/nsd_cococlip__patch__attn/checkpoint-best.pth
+train: [3]  [  0/400]  eta: 0:08:27  lr: nan  time: 1.2679  data: 0.6593  max mem: 57344
+train: [3]  [ 20/400]  eta: 0:04:07  lr: 0.000183  loss: 3.7034 (3.7287)  grad: 1.5401 (1.5309)  time: 0.6217  data: 0.0034  max mem: 57344
+WARNING: classifier 44 (26, 1.0) diverged (loss=65.85 > 63.56) at step 618. Freezing.
+train: [3]  [ 40/400]  eta: 0:03:49  lr: 0.000186  loss: 3.7691 (3.7288)  grad: 1.5438 (1.5322)  time: 0.6215  data: 0.0037  max mem: 57344
+train: [3]  [ 60/400]  eta: 0:03:34  lr: 0.000189  loss: 3.0184 (3.4719)  grad: 0.2166 (1.0884)  time: 0.6159  data: 0.0036  max mem: 57344
+train: [3]  [ 80/400]  eta: 0:03:20  lr: 0.000192  loss: 2.9546 (3.3345)  grad: 0.2018 (0.8683)  time: 0.6155  data: 0.0036  max mem: 57344
+train: [3]  [100/400]  eta: 0:03:07  lr: 0.000195  loss: 2.9577 (3.2585)  grad: 0.2101 (0.7377)  time: 0.6160  data: 0.0037  max mem: 57344
+train: [3]  [120/400]  eta: 0:02:54  lr: 0.000198  loss: 2.9665 (3.2075)  grad: 0.2142 (0.6501)  time: 0.6160  data: 0.0036  max mem: 57344
+train: [3]  [140/400]  eta: 0:02:41  lr: 0.000201  loss: 2.9665 (3.1701)  grad: 0.2017 (0.5858)  time: 0.6157  data: 0.0035  max mem: 57344
+train: [3]  [160/400]  eta: 0:02:29  lr: 0.000204  loss: 2.8837 (3.1335)  grad: 0.2000 (0.5373)  time: 0.6153  data: 0.0035  max mem: 57344
+train: [3]  [180/400]  eta: 0:02:16  lr: 0.000207  loss: 2.9194 (3.1151)  grad: 0.2082 (0.5020)  time: 0.6159  data: 0.0035  max mem: 57344
+train: [3]  [200/400]  eta: 0:02:04  lr: 0.000210  loss: 2.9607 (3.1003)  grad: 0.2259 (0.4750)  time: 0.6150  data: 0.0035  max mem: 57344
+train: [3]  [220/400]  eta: 0:01:51  lr: 0.000213  loss: 2.9327 (3.0841)  grad: 0.2258 (0.4521)  time: 0.6162  data: 0.0037  max mem: 57344
+train: [3]  [240/400]  eta: 0:01:39  lr: 0.000216  loss: 2.9139 (3.0707)  grad: 0.2211 (0.4325)  time: 0.6166  data: 0.0039  max mem: 57344
+train: [3]  [260/400]  eta: 0:01:26  lr: 0.000219  loss: 2.9055 (3.0584)  grad: 0.2067 (0.4152)  time: 0.6165  data: 0.0038  max mem: 57344
+train: [3]  [280/400]  eta: 0:01:14  lr: 0.000222  loss: 2.9017 (3.0478)  grad: 0.2062 (0.4007)  time: 0.6181  data: 0.0040  max mem: 57344
+train: [3]  [300/400]  eta: 0:01:01  lr: 0.000225  loss: 2.8924 (3.0386)  grad: 0.2158 (0.3890)  time: 0.6160  data: 0.0037  max mem: 57344
+train: [3]  [320/400]  eta: 0:00:49  lr: 0.000228  loss: 2.9000 (3.0318)  grad: 0.2225 (0.3783)  time: 0.6155  data: 0.0035  max mem: 57344
+train: [3]  [340/400]  eta: 0:00:37  lr: 0.000231  loss: 2.8963 (3.0237)  grad: 0.2110 (0.3682)  time: 0.6156  data: 0.0035  max mem: 57344
+train: [3]  [360/400]  eta: 0:00:24  lr: 0.000234  loss: 2.8930 (3.0170)  grad: 0.2143 (0.3603)  time: 0.6176  data: 0.0040  max mem: 57344
+train: [3]  [380/400]  eta: 0:00:12  lr: 0.000237  loss: 2.8864 (3.0112)  grad: 0.2222 (0.3533)  time: 0.6193  data: 0.0042  max mem: 57344
+train: [3]  [399/400]  eta: 0:00:00  lr: 0.000240  loss: 2.9233 (3.0066)  grad: 0.2286 (0.3472)  time: 0.6171  data: 0.0039  max mem: 57344
+train: [3] Total time: 0:04:07 (0.6187 s / it)
+train: [3] Summary: lr: 0.000240  loss: 2.9233 (3.0066)  grad: 0.2286 (0.3472)
+eval (validation): [3]  [ 0/85]  eta: 0:01:25    time: 1.0034  data: 0.6479  max mem: 57344
+eval (validation): [3]  [20/85]  eta: 0:00:25    time: 0.3665  data: 0.0035  max mem: 57344
+eval (validation): [3]  [40/85]  eta: 0:00:17    time: 0.3665  data: 0.0036  max mem: 57344
+eval (validation): [3]  [60/85]  eta: 0:00:09    time: 0.3667  data: 0.0037  max mem: 57344
+eval (validation): [3]  [80/85]  eta: 0:00:01    time: 0.3668  data: 0.0035  max mem: 57344
+eval (validation): [3]  [84/85]  eta: 0:00:00    time: 0.3604  data: 0.0035  max mem: 57344
+eval (validation): [3] Total time: 0:00:31 (0.3739 s / it)
+cv: [3]  best hparam: (6, 1.0) (035) ('035_lr6.0e+00_wd1.0e+00')  loss: 2.511  acc: 0.238  f1: 0.193
+saving checkpoint experiments/schaefer1000/output/schaefer1000/schaefer1000_lr3e-4_3/eval_v2/nsd_cococlip__patch__attn/checkpoint-last.pth
+train: [4]  [  0/400]  eta: 0:08:44  lr: nan  time: 1.3107  data: 0.7072  max mem: 57344
+train: [4]  [ 20/400]  eta: 0:04:06  lr: 0.000243  loss: 2.8988 (2.9040)  grad: 0.2444 (0.2489)  time: 0.6153  data: 0.0028  max mem: 57344
+train: [4]  [ 40/400]  eta: 0:03:47  lr: 0.000246  loss: 2.8988 (2.9055)  grad: 0.2483 (0.2729)  time: 0.6161  data: 0.0037  max mem: 57344
+train: [4]  [ 60/400]  eta: 0:03:33  lr: 0.000249  loss: 2.9500 (3.0633)  grad: 0.5123 (0.6052)  time: 0.6154  data: 0.0036  max mem: 57344
+WARNING: classifier 43 (22, 1.0) diverged (loss=82.14 > 63.56) at step 833. Freezing.
+train: [4]  [ 80/400]  eta: 0:03:19  lr: 0.000252  loss: 3.1366 (3.0959)  grad: 0.8124 (0.6471)  time: 0.6120  data: 0.0036  max mem: 57344
+train: [4]  [100/400]  eta: 0:03:06  lr: 0.000255  loss: 2.9012 (3.0557)  grad: 0.2264 (0.5628)  time: 0.6099  data: 0.0035  max mem: 57344
+train: [4]  [120/400]  eta: 0:02:53  lr: 0.000258  loss: 2.9012 (3.0264)  grad: 0.2112 (0.5030)  time: 0.6101  data: 0.0036  max mem: 57344
+train: [4]  [140/400]  eta: 0:02:40  lr: 0.000261  loss: 2.8769 (3.0055)  grad: 0.2168 (0.4642)  time: 0.6100  data: 0.0035  max mem: 57344
+train: [4]  [160/400]  eta: 0:02:27  lr: 0.000264  loss: 2.8661 (2.9896)  grad: 0.2249 (0.4346)  time: 0.6093  data: 0.0035  max mem: 57344
+train: [4]  [180/400]  eta: 0:02:15  lr: 0.000267  loss: 2.8911 (2.9775)  grad: 0.2264 (0.4132)  time: 0.6094  data: 0.0035  max mem: 57344
+train: [4]  [200/400]  eta: 0:02:03  lr: 0.000270  loss: 2.9009 (2.9708)  grad: 0.2297 (0.3951)  time: 0.6107  data: 0.0035  max mem: 57344
+train: [4]  [220/400]  eta: 0:01:50  lr: 0.000273  loss: 2.9009 (2.9655)  grad: 0.2403 (0.3835)  time: 0.6115  data: 0.0039  max mem: 57344
+train: [4]  [240/400]  eta: 0:01:38  lr: 0.000276  loss: 2.9170 (2.9678)  grad: 0.3267 (0.3991)  time: 0.6100  data: 0.0038  max mem: 57344
+train: [4]  [260/400]  eta: 0:01:25  lr: 0.000279  loss: 3.2238 (3.0166)  grad: 0.9389 (0.4894)  time: 0.6107  data: 0.0038  max mem: 57344
+WARNING: classifier 42 (19, 1.0) diverged (loss=76.61 > 63.56) at step 937. Freezing.
+train: [4]  [280/400]  eta: 0:01:13  lr: 0.000282  loss: 3.6586 (3.0617)  grad: 1.6077 (0.5563)  time: 0.6093  data: 0.0038  max mem: 57344
+train: [4]  [300/400]  eta: 0:01:01  lr: 0.000285  loss: 2.9853 (3.0509)  grad: 0.2132 (0.5327)  time: 0.6061  data: 0.0040  max mem: 57344
+train: [4]  [320/400]  eta: 0:00:49  lr: 0.000288  loss: 2.8953 (3.0404)  grad: 0.2071 (0.5130)  time: 0.6050  data: 0.0038  max mem: 57344
+train: [4]  [340/400]  eta: 0:00:36  lr: 0.000291  loss: 2.8923 (3.0333)  grad: 0.2192 (0.4962)  time: 0.6038  data: 0.0036  max mem: 57344
+train: [4]  [360/400]  eta: 0:00:24  lr: 0.000294  loss: 2.8923 (3.0254)  grad: 0.2174 (0.4803)  time: 0.6038  data: 0.0035  max mem: 57344
+train: [4]  [380/400]  eta: 0:00:12  lr: 0.000297  loss: 2.8485 (3.0157)  grad: 0.2158 (0.4667)  time: 0.6060  data: 0.0038  max mem: 57344
+train: [4]  [399/400]  eta: 0:00:00  lr: 0.000300  loss: 2.8414 (3.0081)  grad: 0.2204 (0.4545)  time: 0.6052  data: 0.0039  max mem: 57344
+train: [4] Total time: 0:04:04 (0.6115 s / it)
+train: [4] Summary: lr: 0.000300  loss: 2.8414 (3.0081)  grad: 0.2204 (0.4545)
+eval (validation): [4]  [ 0/85]  eta: 0:01:27    time: 1.0320  data: 0.6748  max mem: 57344
+eval (validation): [4]  [20/85]  eta: 0:00:25    time: 0.3664  data: 0.0031  max mem: 57344
+eval (validation): [4]  [40/85]  eta: 0:00:17    time: 0.3679  data: 0.0040  max mem: 57344
+eval (validation): [4]  [60/85]  eta: 0:00:09    time: 0.3673  data: 0.0040  max mem: 57344
+eval (validation): [4]  [80/85]  eta: 0:00:01    time: 0.3669  data: 0.0038  max mem: 57344
+eval (validation): [4]  [84/85]  eta: 0:00:00    time: 0.3610  data: 0.0038  max mem: 57344
+eval (validation): [4] Total time: 0:00:31 (0.3749 s / it)
+cv: [4]  best hparam: (3.7, 1.0) (032) ('032_lr3.7e+00_wd1.0e+00')  loss: 2.471  acc: 0.257  f1: 0.192
+saving checkpoint experiments/schaefer1000/output/schaefer1000/schaefer1000_lr3e-4_3/eval_v2/nsd_cococlip__patch__attn/checkpoint-last.pth
+saving best checkpoint experiments/schaefer1000/output/schaefer1000/schaefer1000_lr3e-4_3/eval_v2/nsd_cococlip__patch__attn/checkpoint-best.pth
+train: [5]  [  0/400]  eta: 0:08:55  lr: nan  time: 1.3376  data: 0.7453  max mem: 57344
+train: [5]  [ 20/400]  eta: 0:04:02  lr: 0.000300  loss: 2.8644 (2.8856)  grad: 0.2227 (0.2272)  time: 0.6042  data: 0.0026  max mem: 57344
+train: [5]  [ 40/400]  eta: 0:03:43  lr: 0.000300  loss: 2.8575 (2.8760)  grad: 0.2156 (0.2199)  time: 0.6044  data: 0.0038  max mem: 57344
+train: [5]  [ 60/400]  eta: 0:03:29  lr: 0.000300  loss: 2.8685 (2.8736)  grad: 0.2161 (0.2240)  time: 0.6040  data: 0.0037  max mem: 57344
+train: [5]  [ 80/400]  eta: 0:03:16  lr: 0.000300  loss: 2.8739 (2.8717)  grad: 0.2268 (0.2244)  time: 0.6051  data: 0.0037  max mem: 57344
+train: [5]  [100/400]  eta: 0:03:03  lr: 0.000300  loss: 2.8644 (2.8711)  grad: 0.2290 (0.2240)  time: 0.6046  data: 0.0037  max mem: 57344
+train: [5]  [120/400]  eta: 0:02:50  lr: 0.000300  loss: 2.8624 (2.8683)  grad: 0.2301 (0.2249)  time: 0.6047  data: 0.0037  max mem: 57344
+train: [5]  [140/400]  eta: 0:02:38  lr: 0.000300  loss: 2.8474 (2.8668)  grad: 0.2320 (0.2261)  time: 0.6050  data: 0.0037  max mem: 57344
+train: [5]  [160/400]  eta: 0:02:26  lr: 0.000299  loss: 2.8280 (2.8611)  grad: 0.2266 (0.2258)  time: 0.6038  data: 0.0036  max mem: 57344
+train: [5]  [180/400]  eta: 0:02:13  lr: 0.000299  loss: 2.8220 (2.8592)  grad: 0.2162 (0.2252)  time: 0.6035  data: 0.0035  max mem: 57344
+train: [5]  [200/400]  eta: 0:02:01  lr: 0.000299  loss: 2.8346 (2.8568)  grad: 0.2224 (0.2258)  time: 0.6034  data: 0.0035  max mem: 57344
+train: [5]  [220/400]  eta: 0:01:49  lr: 0.000299  loss: 2.8166 (2.8532)  grad: 0.2286 (0.2261)  time: 0.6035  data: 0.0035  max mem: 57344
+train: [5]  [240/400]  eta: 0:01:37  lr: 0.000299  loss: 2.8219 (2.8495)  grad: 0.2243 (0.2253)  time: 0.6043  data: 0.0036  max mem: 57344
+train: [5]  [260/400]  eta: 0:01:24  lr: 0.000299  loss: 2.8411 (2.8507)  grad: 0.2176 (0.2254)  time: 0.6045  data: 0.0036  max mem: 57344
+train: [5]  [280/400]  eta: 0:01:12  lr: 0.000298  loss: 2.8437 (2.8499)  grad: 0.2222 (0.2257)  time: 0.6055  data: 0.0039  max mem: 57344
+train: [5]  [300/400]  eta: 0:01:00  lr: 0.000298  loss: 2.8368 (2.8484)  grad: 0.2222 (0.2250)  time: 0.6057  data: 0.0038  max mem: 57344
+train: [5]  [320/400]  eta: 0:00:48  lr: 0.000298  loss: 2.8723 (2.8500)  grad: 0.2057 (0.2241)  time: 0.6061  data: 0.0039  max mem: 57344
+train: [5]  [340/400]  eta: 0:00:36  lr: 0.000298  loss: 2.8627 (2.8485)  grad: 0.2060 (0.2240)  time: 0.6061  data: 0.0039  max mem: 57344
+train: [5]  [360/400]  eta: 0:00:24  lr: 0.000297  loss: 2.8443 (2.8485)  grad: 0.2258 (0.2241)  time: 0.6046  data: 0.0036  max mem: 57344
+train: [5]  [380/400]  eta: 0:00:12  lr: 0.000297  loss: 2.8443 (2.8490)  grad: 0.2248 (0.2240)  time: 0.6038  data: 0.0035  max mem: 57344
+train: [5]  [399/400]  eta: 0:00:00  lr: 0.000297  loss: 2.8122 (2.8478)  grad: 0.2063 (0.2228)  time: 0.6041  data: 0.0035  max mem: 57344
+train: [5] Total time: 0:04:02 (0.6067 s / it)
+train: [5] Summary: lr: 0.000297  loss: 2.8122 (2.8478)  grad: 0.2063 (0.2228)
+eval (validation): [5]  [ 0/85]  eta: 0:01:22    time: 0.9680  data: 0.6085  max mem: 57344
+eval (validation): [5]  [20/85]  eta: 0:00:25    time: 0.3667  data: 0.0028  max mem: 57344
+eval (validation): [5]  [40/85]  eta: 0:00:17    time: 0.3671  data: 0.0038  max mem: 57344
+eval (validation): [5]  [60/85]  eta: 0:00:09    time: 0.3668  data: 0.0037  max mem: 57344
+eval (validation): [5]  [80/85]  eta: 0:00:01    time: 0.3679  data: 0.0038  max mem: 57344
+eval (validation): [5]  [84/85]  eta: 0:00:00    time: 0.3618  data: 0.0038  max mem: 57344
+eval (validation): [5] Total time: 0:00:31 (0.3744 s / it)
+cv: [5]  best hparam: (3.1, 1.0) (031) ('031_lr3.1e+00_wd1.0e+00')  loss: 2.396  acc: 0.272  f1: 0.203
+saving checkpoint experiments/schaefer1000/output/schaefer1000/schaefer1000_lr3e-4_3/eval_v2/nsd_cococlip__patch__attn/checkpoint-last.pth
+saving best checkpoint experiments/schaefer1000/output/schaefer1000/schaefer1000_lr3e-4_3/eval_v2/nsd_cococlip__patch__attn/checkpoint-best.pth
+train: [6]  [  0/400]  eta: 0:08:52  lr: nan  time: 1.3313  data: 0.7420  max mem: 57344
+train: [6]  [ 20/400]  eta: 0:04:03  lr: 0.000296  loss: 2.7840 (2.8149)  grad: 0.2073 (0.2082)  time: 0.6060  data: 0.0032  max mem: 57344
+train: [6]  [ 40/400]  eta: 0:03:44  lr: 0.000296  loss: 2.8003 (2.8060)  grad: 0.2154 (0.2160)  time: 0.6051  data: 0.0039  max mem: 57344
+train: [6]  [ 60/400]  eta: 0:03:29  lr: 0.000296  loss: 2.8138 (2.8057)  grad: 0.2136 (0.2159)  time: 0.6043  data: 0.0037  max mem: 57344
+train: [6]  [ 80/400]  eta: 0:03:16  lr: 0.000295  loss: 2.7748 (2.7946)  grad: 0.2096 (0.2157)  time: 0.6043  data: 0.0037  max mem: 57344
+train: [6]  [100/400]  eta: 0:03:03  lr: 0.000295  loss: 2.7676 (2.7943)  grad: 0.2147 (0.2173)  time: 0.6043  data: 0.0037  max mem: 57344
+train: [6]  [120/400]  eta: 0:02:51  lr: 0.000295  loss: 2.7997 (2.7966)  grad: 0.2174 (0.2169)  time: 0.6045  data: 0.0037  max mem: 57344
+train: [6]  [140/400]  eta: 0:02:38  lr: 0.000294  loss: 2.7858 (2.7902)  grad: 0.2057 (0.2145)  time: 0.6042  data: 0.0037  max mem: 57344
+train: [6]  [160/400]  eta: 0:02:26  lr: 0.000294  loss: 2.7663 (2.7844)  grad: 0.2069 (0.2150)  time: 0.6045  data: 0.0036  max mem: 57344
+train: [6]  [180/400]  eta: 0:02:13  lr: 0.000293  loss: 2.7552 (2.7845)  grad: 0.2162 (0.2152)  time: 0.6035  data: 0.0035  max mem: 57344
+train: [6]  [200/400]  eta: 0:02:01  lr: 0.000293  loss: 2.7552 (2.7832)  grad: 0.2203 (0.2169)  time: 0.6032  data: 0.0034  max mem: 57344
+train: [6]  [220/400]  eta: 0:01:49  lr: 0.000292  loss: 2.7827 (2.7844)  grad: 0.2185 (0.2161)  time: 0.6030  data: 0.0033  max mem: 57344
+train: [6]  [240/400]  eta: 0:01:37  lr: 0.000292  loss: 2.7895 (2.7854)  grad: 0.2088 (0.2158)  time: 0.6031  data: 0.0033  max mem: 57344
+train: [6]  [260/400]  eta: 0:01:24  lr: 0.000291  loss: 2.7895 (2.7870)  grad: 0.2103 (0.2154)  time: 0.6026  data: 0.0033  max mem: 57344
+train: [6]  [280/400]  eta: 0:01:12  lr: 0.000291  loss: 2.7860 (2.7869)  grad: 0.2083 (0.2147)  time: 0.6059  data: 0.0037  max mem: 57344
+train: [6]  [300/400]  eta: 0:01:00  lr: 0.000290  loss: 2.7746 (2.7872)  grad: 0.2082 (0.2148)  time: 0.6048  data: 0.0039  max mem: 57344
+train: [6]  [320/400]  eta: 0:00:48  lr: 0.000290  loss: 2.7829 (2.7889)  grad: 0.2082 (0.2147)  time: 0.6055  data: 0.0038  max mem: 57344
+train: [6]  [340/400]  eta: 0:00:36  lr: 0.000289  loss: 2.7829 (2.7878)  grad: 0.2090 (0.2150)  time: 0.6057  data: 0.0038  max mem: 57344
+train: [6]  [360/400]  eta: 0:00:24  lr: 0.000288  loss: 2.7803 (2.7870)  grad: 0.2150 (0.2150)  time: 0.6064  data: 0.0039  max mem: 57344
+train: [6]  [380/400]  eta: 0:00:12  lr: 0.000288  loss: 2.7467 (2.7870)  grad: 0.2133 (0.2152)  time: 0.6058  data: 0.0040  max mem: 57344
+train: [6]  [399/400]  eta: 0:00:00  lr: 0.000287  loss: 2.7404 (2.7858)  grad: 0.2133 (0.2153)  time: 0.6038  data: 0.0034  max mem: 57344
+train: [6] Total time: 0:04:02 (0.6066 s / it)
+train: [6] Summary: lr: 0.000287  loss: 2.7404 (2.7858)  grad: 0.2133 (0.2153)
+eval (validation): [6]  [ 0/85]  eta: 0:01:21    time: 0.9588  data: 0.6015  max mem: 57344
+eval (validation): [6]  [20/85]  eta: 0:00:25    time: 0.3666  data: 0.0032  max mem: 57344
+eval (validation): [6]  [40/85]  eta: 0:00:17    time: 0.3663  data: 0.0034  max mem: 57344
+eval (validation): [6]  [60/85]  eta: 0:00:09    time: 0.3674  data: 0.0039  max mem: 57344
+eval (validation): [6]  [80/85]  eta: 0:00:01    time: 0.3665  data: 0.0036  max mem: 57344
+eval (validation): [6]  [84/85]  eta: 0:00:00    time: 0.3601  data: 0.0036  max mem: 57344
+eval (validation): [6] Total time: 0:00:31 (0.3736 s / it)
+cv: [6]  best hparam: (4.3, 1.0) (033) ('033_lr4.3e+00_wd1.0e+00')  loss: 2.444  acc: 0.268  f1: 0.213
+saving checkpoint experiments/schaefer1000/output/schaefer1000/schaefer1000_lr3e-4_3/eval_v2/nsd_cococlip__patch__attn/checkpoint-last.pth
+train: [7]  [  0/400]  eta: 0:09:10  lr: nan  time: 1.3769  data: 0.7809  max mem: 57344
+train: [7]  [ 20/400]  eta: 0:04:03  lr: 0.000286  loss: 2.7122 (2.7048)  grad: 0.2135 (0.2121)  time: 0.6049  data: 0.0033  max mem: 57344
+train: [7]  [ 40/400]  eta: 0:03:44  lr: 0.000286  loss: 2.7157 (2.7271)  grad: 0.2145 (0.2141)  time: 0.6060  data: 0.0039  max mem: 57344
+train: [7]  [ 60/400]  eta: 0:03:30  lr: 0.000285  loss: 2.7380 (2.7340)  grad: 0.2152 (0.2168)  time: 0.6063  data: 0.0039  max mem: 57344
+train: [7]  [ 80/400]  eta: 0:03:16  lr: 0.000284  loss: 2.7422 (2.7350)  grad: 0.2166 (0.2186)  time: 0.6049  data: 0.0038  max mem: 57344
+train: [7]  [100/400]  eta: 0:03:03  lr: 0.000284  loss: 2.7388 (2.7393)  grad: 0.2190 (0.2187)  time: 0.6052  data: 0.0038  max mem: 57344
+train: [7]  [120/400]  eta: 0:02:51  lr: 0.000283  loss: 2.7388 (2.7353)  grad: 0.2223 (0.2204)  time: 0.6044  data: 0.0036  max mem: 57344
+train: [7]  [140/400]  eta: 0:02:38  lr: 0.000282  loss: 2.7264 (2.7390)  grad: 0.2281 (0.2216)  time: 0.6045  data: 0.0036  max mem: 57344
+train: [7]  [160/400]  eta: 0:02:26  lr: 0.000282  loss: 2.7420 (2.7397)  grad: 0.2067 (0.2202)  time: 0.6045  data: 0.0037  max mem: 57344
+train: [7]  [180/400]  eta: 0:02:14  lr: 0.000281  loss: 2.7230 (2.7366)  grad: 0.2052 (0.2191)  time: 0.6039  data: 0.0036  max mem: 57344
+train: [7]  [200/400]  eta: 0:02:01  lr: 0.000280  loss: 2.7230 (2.7339)  grad: 0.2067 (0.2186)  time: 0.6041  data: 0.0037  max mem: 57344
+train: [7]  [220/400]  eta: 0:01:49  lr: 0.000279  loss: 2.7350 (2.7343)  grad: 0.2121 (0.2189)  time: 0.6039  data: 0.0036  max mem: 57344
+train: [7]  [240/400]  eta: 0:01:37  lr: 0.000278  loss: 2.7308 (2.7321)  grad: 0.2149 (0.2184)  time: 0.6030  data: 0.0033  max mem: 57344
+train: [7]  [260/400]  eta: 0:01:25  lr: 0.000278  loss: 2.7413 (2.7361)  grad: 0.2197 (0.2190)  time: 0.6021  data: 0.0033  max mem: 57344
+train: [7]  [280/400]  eta: 0:01:12  lr: 0.000277  loss: 2.7749 (2.7396)  grad: 0.2252 (0.2195)  time: 0.6029  data: 0.0033  max mem: 57344
+train: [7]  [300/400]  eta: 0:01:00  lr: 0.000276  loss: 2.7581 (2.7399)  grad: 0.2284 (0.2205)  time: 0.6029  data: 0.0034  max mem: 57344
+train: [7]  [320/400]  eta: 0:00:48  lr: 0.000275  loss: 2.7374 (2.7405)  grad: 0.2346 (0.2216)  time: 0.6028  data: 0.0033  max mem: 57344
+train: [7]  [340/400]  eta: 0:00:36  lr: 0.000274  loss: 2.7646 (2.7430)  grad: 0.2352 (0.2224)  time: 0.6050  data: 0.0037  max mem: 57344
+train: [7]  [360/400]  eta: 0:00:24  lr: 0.000273  loss: 2.7729 (2.7438)  grad: 0.2297 (0.2226)  time: 0.6053  data: 0.0036  max mem: 57344
+train: [7]  [380/400]  eta: 0:00:12  lr: 0.000272  loss: 2.7729 (2.7460)  grad: 0.2226 (0.2225)  time: 0.6043  data: 0.0036  max mem: 57344
+train: [7]  [399/400]  eta: 0:00:00  lr: 0.000271  loss: 2.7964 (2.7497)  grad: 0.2205 (0.2226)  time: 0.6049  data: 0.0038  max mem: 57344
+train: [7] Total time: 0:04:02 (0.6065 s / it)
+train: [7] Summary: lr: 0.000271  loss: 2.7964 (2.7497)  grad: 0.2205 (0.2226)
+eval (validation): [7]  [ 0/85]  eta: 0:01:21    time: 0.9589  data: 0.6034  max mem: 57344
+eval (validation): [7]  [20/85]  eta: 0:00:30    time: 0.4421  data: 0.0801  max mem: 57344
+eval (validation): [7]  [40/85]  eta: 0:00:18    time: 0.3661  data: 0.0038  max mem: 57344
+eval (validation): [7]  [60/85]  eta: 0:00:10    time: 0.3659  data: 0.0034  max mem: 57344
+eval (validation): [7]  [80/85]  eta: 0:00:01    time: 0.3659  data: 0.0035  max mem: 57344
+eval (validation): [7]  [84/85]  eta: 0:00:00    time: 0.3595  data: 0.0035  max mem: 57344
+eval (validation): [7] Total time: 0:00:33 (0.3908 s / it)
+cv: [7]  best hparam: (5.1, 1.0) (034) ('034_lr5.1e+00_wd1.0e+00')  loss: 2.455  acc: 0.267  f1: 0.208
+saving checkpoint experiments/schaefer1000/output/schaefer1000/schaefer1000_lr3e-4_3/eval_v2/nsd_cococlip__patch__attn/checkpoint-last.pth
+train: [8]  [  0/400]  eta: 0:07:59  lr: nan  time: 1.1988  data: 0.6070  max mem: 57344
+train: [8]  [ 20/400]  eta: 0:04:00  lr: 0.000270  loss: 2.6846 (2.7040)  grad: 0.2175 (0.2193)  time: 0.6046  data: 0.0035  max mem: 57344
+train: [8]  [ 40/400]  eta: 0:03:42  lr: 0.000270  loss: 2.7029 (2.7017)  grad: 0.2187 (0.2172)  time: 0.6047  data: 0.0038  max mem: 57344
+train: [8]  [ 60/400]  eta: 0:03:28  lr: 0.000269  loss: 2.7168 (2.7133)  grad: 0.2191 (0.2192)  time: 0.6049  data: 0.0037  max mem: 57344
+train: [8]  [ 80/400]  eta: 0:03:15  lr: 0.000268  loss: 2.6994 (2.7084)  grad: 0.2190 (0.2197)  time: 0.6061  data: 0.0041  max mem: 57344
+train: [8]  [100/400]  eta: 0:03:03  lr: 0.000267  loss: 2.6904 (2.7070)  grad: 0.2140 (0.2181)  time: 0.6047  data: 0.0038  max mem: 57344
+train: [8]  [120/400]  eta: 0:02:50  lr: 0.000266  loss: 2.7211 (2.7105)  grad: 0.2158 (0.2176)  time: 0.6034  data: 0.0034  max mem: 57344
+train: [8]  [140/400]  eta: 0:02:38  lr: 0.000265  loss: 2.7252 (2.7100)  grad: 0.2182 (0.2188)  time: 0.6040  data: 0.0035  max mem: 57344
+train: [8]  [160/400]  eta: 0:02:26  lr: 0.000264  loss: 2.6889 (2.7134)  grad: 0.2316 (0.2211)  time: 0.6063  data: 0.0041  max mem: 57344
+train: [8]  [180/400]  eta: 0:02:13  lr: 0.000263  loss: 2.7031 (2.7142)  grad: 0.2264 (0.2213)  time: 0.6069  data: 0.0041  max mem: 57344
+train: [8]  [200/400]  eta: 0:02:01  lr: 0.000262  loss: 2.7031 (2.7142)  grad: 0.2229 (0.2216)  time: 0.6047  data: 0.0038  max mem: 57344
+train: [8]  [220/400]  eta: 0:01:49  lr: 0.000260  loss: 2.6725 (2.7114)  grad: 0.2242 (0.2216)  time: 0.6048  data: 0.0037  max mem: 57344
+train: [8]  [240/400]  eta: 0:01:37  lr: 0.000259  loss: 2.6956 (2.7119)  grad: 0.2179 (0.2214)  time: 0.6041  data: 0.0036  max mem: 57344
+train: [8]  [260/400]  eta: 0:01:24  lr: 0.000258  loss: 2.7297 (2.7141)  grad: 0.2204 (0.2220)  time: 0.6038  data: 0.0036  max mem: 57344
+train: [8]  [280/400]  eta: 0:01:12  lr: 0.000257  loss: 2.7319 (2.7147)  grad: 0.2176 (0.2217)  time: 0.6039  data: 0.0036  max mem: 57344
+train: [8]  [300/400]  eta: 0:01:00  lr: 0.000256  loss: 2.7218 (2.7134)  grad: 0.2183 (0.2221)  time: 0.6044  data: 0.0036  max mem: 57344
+train: [8]  [320/400]  eta: 0:00:48  lr: 0.000255  loss: 2.7199 (2.7138)  grad: 0.2248 (0.2224)  time: 0.6041  data: 0.0036  max mem: 57344
+train: [8]  [340/400]  eta: 0:00:36  lr: 0.000254  loss: 2.7071 (2.7143)  grad: 0.2199 (0.2223)  time: 0.6041  data: 0.0035  max mem: 57344
+train: [8]  [360/400]  eta: 0:00:24  lr: 0.000253  loss: 2.7065 (2.7142)  grad: 0.2179 (0.2222)  time: 0.6033  data: 0.0034  max mem: 57344
+train: [8]  [380/400]  eta: 0:00:12  lr: 0.000252  loss: 2.7013 (2.7139)  grad: 0.2162 (0.2220)  time: 0.6030  data: 0.0033  max mem: 57344
+train: [8]  [399/400]  eta: 0:00:00  lr: 0.000250  loss: 2.6832 (2.7119)  grad: 0.2153 (0.2223)  time: 0.6023  data: 0.0033  max mem: 57344
+train: [8] Total time: 0:04:02 (0.6062 s / it)
+train: [8] Summary: lr: 0.000250  loss: 2.6832 (2.7119)  grad: 0.2153 (0.2223)
+eval (validation): [8]  [ 0/85]  eta: 0:01:07    time: 0.7999  data: 0.4432  max mem: 57344
+eval (validation): [8]  [20/85]  eta: 0:00:25    time: 0.3649  data: 0.0029  max mem: 57344
+eval (validation): [8]  [40/85]  eta: 0:00:16    time: 0.3659  data: 0.0033  max mem: 57344
+eval (validation): [8]  [60/85]  eta: 0:00:09    time: 0.3657  data: 0.0033  max mem: 57344
+eval (validation): [8]  [80/85]  eta: 0:00:01    time: 0.3655  data: 0.0032  max mem: 57344
+eval (validation): [8]  [84/85]  eta: 0:00:00    time: 0.3596  data: 0.0034  max mem: 57344
+eval (validation): [8] Total time: 0:00:31 (0.3706 s / it)
+cv: [8]  best hparam: (1.4, 1.0) (026) ('026_lr1.4e+00_wd1.0e+00')  loss: 2.396  acc: 0.280  f1: 0.220
+saving checkpoint experiments/schaefer1000/output/schaefer1000/schaefer1000_lr3e-4_3/eval_v2/nsd_cococlip__patch__attn/checkpoint-last.pth
+saving best checkpoint experiments/schaefer1000/output/schaefer1000/schaefer1000_lr3e-4_3/eval_v2/nsd_cococlip__patch__attn/checkpoint-best.pth
+train: [9]  [  0/400]  eta: 0:08:26  lr: nan  time: 1.2663  data: 0.6752  max mem: 57344
+train: [9]  [ 20/400]  eta: 0:04:02  lr: 0.000249  loss: 2.6683 (2.6817)  grad: 0.2111 (0.2139)  time: 0.6060  data: 0.0035  max mem: 57344
+train: [9]  [ 40/400]  eta: 0:03:43  lr: 0.000248  loss: 2.6303 (2.6588)  grad: 0.2126 (0.2142)  time: 0.6048  data: 0.0037  max mem: 57344
+train: [9]  [ 60/400]  eta: 0:03:29  lr: 0.000247  loss: 2.6818 (2.6714)  grad: 0.2126 (0.2148)  time: 0.6035  data: 0.0034  max mem: 57344
+train: [9]  [ 80/400]  eta: 0:03:16  lr: 0.000246  loss: 2.7220 (2.6870)  grad: 0.2169 (0.2171)  time: 0.6032  data: 0.0034  max mem: 57344
+train: [9]  [100/400]  eta: 0:03:03  lr: 0.000244  loss: 2.7271 (2.6870)  grad: 0.2192 (0.2171)  time: 0.6053  data: 0.0037  max mem: 57344
+train: [9]  [120/400]  eta: 0:02:50  lr: 0.000243  loss: 2.6731 (2.6855)  grad: 0.2175 (0.2172)  time: 0.6072  data: 0.0040  max mem: 57344
+train: [9]  [140/400]  eta: 0:02:38  lr: 0.000242  loss: 2.6255 (2.6763)  grad: 0.2244 (0.2183)  time: 0.6048  data: 0.0037  max mem: 57344
+train: [9]  [160/400]  eta: 0:02:26  lr: 0.000241  loss: 2.6715 (2.6805)  grad: 0.2183 (0.2180)  time: 0.6043  data: 0.0038  max mem: 57344
+train: [9]  [180/400]  eta: 0:02:13  lr: 0.000240  loss: 2.7041 (2.6857)  grad: 0.2137 (0.2177)  time: 0.6037  data: 0.0035  max mem: 57344
+train: [9]  [200/400]  eta: 0:02:01  lr: 0.000238  loss: 2.6871 (2.6838)  grad: 0.2146 (0.2173)  time: 0.6036  data: 0.0035  max mem: 57344
+train: [9]  [220/400]  eta: 0:01:49  lr: 0.000237  loss: 2.6819 (2.6874)  grad: 0.2144 (0.2175)  time: 0.6037  data: 0.0035  max mem: 57344
+train: [9]  [240/400]  eta: 0:01:37  lr: 0.000236  loss: 2.7024 (2.6889)  grad: 0.2167 (0.2178)  time: 0.6030  data: 0.0033  max mem: 57344
+train: [9]  [260/400]  eta: 0:01:24  lr: 0.000234  loss: 2.7043 (2.6908)  grad: 0.2171 (0.2175)  time: 0.6027  data: 0.0032  max mem: 57344
+train: [9]  [280/400]  eta: 0:01:12  lr: 0.000233  loss: 2.6698 (2.6895)  grad: 0.2028 (0.2162)  time: 0.6025  data: 0.0032  max mem: 57344
+train: [9]  [300/400]  eta: 0:01:00  lr: 0.000232  loss: 2.6698 (2.6889)  grad: 0.2149 (0.2168)  time: 0.6026  data: 0.0033  max mem: 57344
+train: [9]  [320/400]  eta: 0:00:48  lr: 0.000230  loss: 2.6785 (2.6893)  grad: 0.2224 (0.2171)  time: 0.6023  data: 0.0032  max mem: 57344
+train: [9]  [340/400]  eta: 0:00:36  lr: 0.000229  loss: 2.6785 (2.6905)  grad: 0.2163 (0.2173)  time: 0.6028  data: 0.0032  max mem: 57344
+train: [9]  [360/400]  eta: 0:00:24  lr: 0.000228  loss: 2.6632 (2.6868)  grad: 0.2119 (0.2172)  time: 0.6028  data: 0.0032  max mem: 57344
+train: [9]  [380/400]  eta: 0:00:12  lr: 0.000226  loss: 2.6537 (2.6846)  grad: 0.2098 (0.2170)  time: 0.6023  data: 0.0032  max mem: 57344
+train: [9]  [399/400]  eta: 0:00:00  lr: 0.000225  loss: 2.6875 (2.6851)  grad: 0.2156 (0.2170)  time: 0.6022  data: 0.0032  max mem: 57344
+train: [9] Total time: 0:04:02 (0.6056 s / it)
+train: [9] Summary: lr: 0.000225  loss: 2.6875 (2.6851)  grad: 0.2156 (0.2170)
+eval (validation): [9]  [ 0/85]  eta: 0:01:09    time: 0.8218  data: 0.4651  max mem: 57344
+eval (validation): [9]  [20/85]  eta: 0:00:25    time: 0.3648  data: 0.0028  max mem: 57344
+eval (validation): [9]  [40/85]  eta: 0:00:16    time: 0.3653  data: 0.0031  max mem: 57344
+eval (validation): [9]  [60/85]  eta: 0:00:09    time: 0.3658  data: 0.0031  max mem: 57344
+eval (validation): [9]  [80/85]  eta: 0:00:01    time: 0.3656  data: 0.0033  max mem: 57344
+eval (validation): [9]  [84/85]  eta: 0:00:00    time: 0.3597  data: 0.0033  max mem: 57344
+eval (validation): [9] Total time: 0:00:31 (0.3707 s / it)
+cv: [9]  best hparam: (5.1, 1.0) (034) ('034_lr5.1e+00_wd1.0e+00')  loss: 2.395  acc: 0.282  f1: 0.217
+saving checkpoint experiments/schaefer1000/output/schaefer1000/schaefer1000_lr3e-4_3/eval_v2/nsd_cococlip__patch__attn/checkpoint-last.pth
+saving best checkpoint experiments/schaefer1000/output/schaefer1000/schaefer1000_lr3e-4_3/eval_v2/nsd_cococlip__patch__attn/checkpoint-best.pth
+train: [10]  [  0/400]  eta: 0:07:31  lr: nan  time: 1.1288  data: 0.5385  max mem: 57344
+train: [10]  [ 20/400]  eta: 0:03:58  lr: 0.000224  loss: 2.6490 (2.6415)  grad: 0.2046 (0.2105)  time: 0.6013  data: 0.0025  max mem: 57344
+train: [10]  [ 40/400]  eta: 0:03:41  lr: 0.000222  loss: 2.6187 (2.6308)  grad: 0.2190 (0.2170)  time: 0.6023  data: 0.0033  max mem: 57344
+train: [10]  [ 60/400]  eta: 0:03:27  lr: 0.000221  loss: 2.6196 (2.6347)  grad: 0.2225 (0.2191)  time: 0.6027  data: 0.0034  max mem: 57344
+train: [10]  [ 80/400]  eta: 0:03:14  lr: 0.000220  loss: 2.6437 (2.6431)  grad: 0.2236 (0.2188)  time: 0.6036  data: 0.0035  max mem: 57344
+train: [10]  [100/400]  eta: 0:03:02  lr: 0.000218  loss: 2.6581 (2.6403)  grad: 0.2224 (0.2183)  time: 0.6044  data: 0.0036  max mem: 57344
+train: [10]  [120/400]  eta: 0:02:50  lr: 0.000217  loss: 2.6621 (2.6488)  grad: 0.2134 (0.2179)  time: 0.6046  data: 0.0036  max mem: 57344
+train: [10]  [140/400]  eta: 0:02:37  lr: 0.000215  loss: 2.6512 (2.6468)  grad: 0.2168 (0.2183)  time: 0.6039  data: 0.0037  max mem: 57344
+train: [10]  [160/400]  eta: 0:02:25  lr: 0.000214  loss: 2.6427 (2.6489)  grad: 0.2206 (0.2197)  time: 0.6044  data: 0.0038  max mem: 57344
+train: [10]  [180/400]  eta: 0:02:13  lr: 0.000213  loss: 2.6467 (2.6511)  grad: 0.2280 (0.2200)  time: 0.6032  data: 0.0035  max mem: 57344
+train: [10]  [200/400]  eta: 0:02:01  lr: 0.000211  loss: 2.6373 (2.6482)  grad: 0.2196 (0.2202)  time: 0.6026  data: 0.0034  max mem: 57344
+train: [10]  [220/400]  eta: 0:01:49  lr: 0.000210  loss: 2.6454 (2.6491)  grad: 0.2166 (0.2202)  time: 0.6027  data: 0.0034  max mem: 57344
+train: [10]  [240/400]  eta: 0:01:36  lr: 0.000208  loss: 2.6454 (2.6472)  grad: 0.2164 (0.2198)  time: 0.6027  data: 0.0033  max mem: 57344
+train: [10]  [260/400]  eta: 0:01:24  lr: 0.000207  loss: 2.6378 (2.6482)  grad: 0.2154 (0.2197)  time: 0.6029  data: 0.0033  max mem: 57344
+train: [10]  [280/400]  eta: 0:01:12  lr: 0.000205  loss: 2.6718 (2.6511)  grad: 0.2149 (0.2193)  time: 0.6029  data: 0.0033  max mem: 57344
+train: [10]  [300/400]  eta: 0:01:00  lr: 0.000204  loss: 2.6850 (2.6528)  grad: 0.2132 (0.2190)  time: 0.6026  data: 0.0034  max mem: 57344
+train: [10]  [320/400]  eta: 0:00:48  lr: 0.000202  loss: 2.6730 (2.6536)  grad: 0.2138 (0.2189)  time: 0.6030  data: 0.0034  max mem: 57344
+train: [10]  [340/400]  eta: 0:00:36  lr: 0.000201  loss: 2.6599 (2.6545)  grad: 0.2149 (0.2187)  time: 0.6029  data: 0.0034  max mem: 57344
+train: [10]  [360/400]  eta: 0:00:24  lr: 0.000199  loss: 2.6505 (2.6522)  grad: 0.2073 (0.2181)  time: 0.6031  data: 0.0034  max mem: 57344
+train: [10]  [380/400]  eta: 0:00:12  lr: 0.000198  loss: 2.6227 (2.6525)  grad: 0.2092 (0.2183)  time: 0.6044  data: 0.0036  max mem: 57344
+train: [10]  [399/400]  eta: 0:00:00  lr: 0.000196  loss: 2.6427 (2.6523)  grad: 0.2158 (0.2181)  time: 0.6043  data: 0.0036  max mem: 57344
+train: [10] Total time: 0:04:01 (0.6048 s / it)
+train: [10] Summary: lr: 0.000196  loss: 2.6427 (2.6523)  grad: 0.2158 (0.2181)
+eval (validation): [10]  [ 0/85]  eta: 0:01:19    time: 0.9322  data: 0.5767  max mem: 57344
+eval (validation): [10]  [20/85]  eta: 0:00:25    time: 0.3645  data: 0.0027  max mem: 57344
+eval (validation): [10]  [40/85]  eta: 0:00:17    time: 0.3664  data: 0.0033  max mem: 57344
+eval (validation): [10]  [60/85]  eta: 0:00:09    time: 0.3660  data: 0.0034  max mem: 57344
+eval (validation): [10]  [80/85]  eta: 0:00:01    time: 0.3655  data: 0.0033  max mem: 57344
+eval (validation): [10]  [84/85]  eta: 0:00:00    time: 0.3592  data: 0.0033  max mem: 57344
+eval (validation): [10] Total time: 0:00:31 (0.3721 s / it)
+cv: [10]  best hparam: (1.9, 1.0) (028) ('028_lr1.9e+00_wd1.0e+00')  loss: 2.365  acc: 0.289  f1: 0.229
+saving checkpoint experiments/schaefer1000/output/schaefer1000/schaefer1000_lr3e-4_3/eval_v2/nsd_cococlip__patch__attn/checkpoint-last.pth
+saving best checkpoint experiments/schaefer1000/output/schaefer1000/schaefer1000_lr3e-4_3/eval_v2/nsd_cococlip__patch__attn/checkpoint-best.pth
+train: [11]  [  0/400]  eta: 0:07:42  lr: nan  time: 1.1575  data: 0.5670  max mem: 57344
+train: [11]  [ 20/400]  eta: 0:03:59  lr: 0.000195  loss: 2.6102 (2.6187)  grad: 0.2155 (0.2233)  time: 0.6037  data: 0.0026  max mem: 57344
+train: [11]  [ 40/400]  eta: 0:03:42  lr: 0.000193  loss: 2.6093 (2.6112)  grad: 0.2155 (0.2184)  time: 0.6027  data: 0.0033  max mem: 57344
+train: [11]  [ 60/400]  eta: 0:03:28  lr: 0.000192  loss: 2.6264 (2.6203)  grad: 0.2079 (0.2141)  time: 0.6031  data: 0.0035  max mem: 57344
+train: [11]  [ 80/400]  eta: 0:03:15  lr: 0.000190  loss: 2.6343 (2.6239)  grad: 0.2079 (0.2137)  time: 0.6031  data: 0.0035  max mem: 57344
+train: [11]  [100/400]  eta: 0:03:02  lr: 0.000189  loss: 2.6640 (2.6379)  grad: 0.2125 (0.2153)  time: 0.6032  data: 0.0034  max mem: 57344
+train: [11]  [120/400]  eta: 0:02:50  lr: 0.000187  loss: 2.6676 (2.6409)  grad: 0.2135 (0.2155)  time: 0.6027  data: 0.0034  max mem: 57344
+train: [11]  [140/400]  eta: 0:02:37  lr: 0.000186  loss: 2.6525 (2.6369)  grad: 0.2060 (0.2131)  time: 0.6027  data: 0.0034  max mem: 57344
+train: [11]  [160/400]  eta: 0:02:25  lr: 0.000184  loss: 2.6400 (2.6320)  grad: 0.1977 (0.2116)  time: 0.6032  data: 0.0034  max mem: 57344
+train: [11]  [180/400]  eta: 0:02:13  lr: 0.000183  loss: 2.5820 (2.6289)  grad: 0.2056 (0.2124)  time: 0.6027  data: 0.0034  max mem: 57344
+train: [11]  [200/400]  eta: 0:02:01  lr: 0.000181  loss: 2.6107 (2.6292)  grad: 0.2182 (0.2132)  time: 0.6034  data: 0.0034  max mem: 57344
+train: [11]  [220/400]  eta: 0:01:48  lr: 0.000180  loss: 2.6529 (2.6297)  grad: 0.2102 (0.2128)  time: 0.6026  data: 0.0034  max mem: 57344
+train: [11]  [240/400]  eta: 0:01:36  lr: 0.000178  loss: 2.6549 (2.6315)  grad: 0.2158 (0.2145)  time: 0.6032  data: 0.0035  max mem: 57344
+train: [11]  [260/400]  eta: 0:01:24  lr: 0.000177  loss: 2.6279 (2.6302)  grad: 0.2238 (0.2148)  time: 0.6035  data: 0.0035  max mem: 57344
+train: [11]  [280/400]  eta: 0:01:12  lr: 0.000175  loss: 2.6180 (2.6285)  grad: 0.2181 (0.2150)  time: 0.6034  data: 0.0034  max mem: 57344
+train: [11]  [300/400]  eta: 0:01:00  lr: 0.000174  loss: 2.6244 (2.6270)  grad: 0.2164 (0.2153)  time: 0.6030  data: 0.0033  max mem: 57344
+train: [11]  [320/400]  eta: 0:00:48  lr: 0.000172  loss: 2.6027 (2.6256)  grad: 0.2152 (0.2154)  time: 0.6028  data: 0.0033  max mem: 57344
+train: [11]  [340/400]  eta: 0:00:36  lr: 0.000170  loss: 2.6027 (2.6270)  grad: 0.2122 (0.2151)  time: 0.6028  data: 0.0033  max mem: 57344
+train: [11]  [360/400]  eta: 0:00:24  lr: 0.000169  loss: 2.5850 (2.6251)  grad: 0.2156 (0.2152)  time: 0.6027  data: 0.0033  max mem: 57344
+train: [11]  [380/400]  eta: 0:00:12  lr: 0.000167  loss: 2.6238 (2.6266)  grad: 0.2173 (0.2153)  time: 0.6030  data: 0.0033  max mem: 57344
+train: [11]  [399/400]  eta: 0:00:00  lr: 0.000166  loss: 2.6734 (2.6286)  grad: 0.2173 (0.2156)  time: 0.6028  data: 0.0034  max mem: 57344
+train: [11] Total time: 0:04:01 (0.6047 s / it)
+train: [11] Summary: lr: 0.000166  loss: 2.6734 (2.6286)  grad: 0.2173 (0.2156)
+eval (validation): [11]  [ 0/85]  eta: 0:01:04    time: 0.7591  data: 0.4002  max mem: 57344
+eval (validation): [11]  [20/85]  eta: 0:00:24    time: 0.3655  data: 0.0031  max mem: 57344
+eval (validation): [11]  [40/85]  eta: 0:00:16    time: 0.3660  data: 0.0032  max mem: 57344
+eval (validation): [11]  [60/85]  eta: 0:00:09    time: 0.3664  data: 0.0034  max mem: 57344
+eval (validation): [11]  [80/85]  eta: 0:00:01    time: 0.3665  data: 0.0034  max mem: 57344
+eval (validation): [11]  [84/85]  eta: 0:00:00    time: 0.3601  data: 0.0034  max mem: 57344
+eval (validation): [11] Total time: 0:00:31 (0.3706 s / it)
+cv: [11]  best hparam: (1.4, 1.0) (026) ('026_lr1.4e+00_wd1.0e+00')  loss: 2.383  acc: 0.284  f1: 0.234
+saving checkpoint experiments/schaefer1000/output/schaefer1000/schaefer1000_lr3e-4_3/eval_v2/nsd_cococlip__patch__attn/checkpoint-last.pth
+train: [12]  [  0/400]  eta: 0:08:32  lr: nan  time: 1.2811  data: 0.6868  max mem: 57344
+train: [12]  [ 20/400]  eta: 0:04:01  lr: 0.000164  loss: 2.6434 (2.6505)  grad: 0.2123 (0.2146)  time: 0.6023  data: 0.0025  max mem: 57344
+train: [12]  [ 40/400]  eta: 0:03:42  lr: 0.000163  loss: 2.6388 (2.6306)  grad: 0.2122 (0.2121)  time: 0.6028  data: 0.0032  max mem: 57344
+train: [12]  [ 60/400]  eta: 0:03:28  lr: 0.000161  loss: 2.6054 (2.6257)  grad: 0.2082 (0.2103)  time: 0.6022  data: 0.0033  max mem: 57344
+train: [12]  [ 80/400]  eta: 0:03:15  lr: 0.000160  loss: 2.5982 (2.6115)  grad: 0.2088 (0.2105)  time: 0.6030  data: 0.0034  max mem: 57344
+train: [12]  [100/400]  eta: 0:03:02  lr: 0.000158  loss: 2.5775 (2.6115)  grad: 0.2146 (0.2126)  time: 0.6031  data: 0.0034  max mem: 57344
+train: [12]  [120/400]  eta: 0:02:50  lr: 0.000156  loss: 2.6306 (2.6211)  grad: 0.2229 (0.2159)  time: 0.6028  data: 0.0034  max mem: 57344
+train: [12]  [140/400]  eta: 0:02:37  lr: 0.000155  loss: 2.6621 (2.6209)  grad: 0.2229 (0.2168)  time: 0.6038  data: 0.0037  max mem: 57344
+train: [12]  [160/400]  eta: 0:02:25  lr: 0.000153  loss: 2.5804 (2.6150)  grad: 0.2205 (0.2172)  time: 0.6035  data: 0.0036  max mem: 57344
+train: [12]  [180/400]  eta: 0:02:13  lr: 0.000152  loss: 2.5804 (2.6128)  grad: 0.2216 (0.2183)  time: 0.6038  data: 0.0035  max mem: 57344
+train: [12]  [200/400]  eta: 0:02:01  lr: 0.000150  loss: 2.5985 (2.6138)  grad: 0.2284 (0.2195)  time: 0.6035  data: 0.0036  max mem: 57344
+train: [12]  [220/400]  eta: 0:01:49  lr: 0.000149  loss: 2.5818 (2.6106)  grad: 0.2234 (0.2195)  time: 0.6040  data: 0.0037  max mem: 57344
+train: [12]  [240/400]  eta: 0:01:36  lr: 0.000147  loss: 2.5570 (2.6093)  grad: 0.2112 (0.2183)  time: 0.6042  data: 0.0036  max mem: 57344
+train: [12]  [260/400]  eta: 0:01:24  lr: 0.000145  loss: 2.5732 (2.6074)  grad: 0.2065 (0.2172)  time: 0.6035  data: 0.0036  max mem: 57344
+train: [12]  [280/400]  eta: 0:01:12  lr: 0.000144  loss: 2.5892 (2.6080)  grad: 0.2070 (0.2172)  time: 0.6037  data: 0.0035  max mem: 57344
+train: [12]  [300/400]  eta: 0:01:00  lr: 0.000142  loss: 2.6324 (2.6099)  grad: 0.2148 (0.2174)  time: 0.6040  data: 0.0036  max mem: 57344
+train: [12]  [320/400]  eta: 0:00:48  lr: 0.000141  loss: 2.6138 (2.6084)  grad: 0.2190 (0.2174)  time: 0.6040  data: 0.0036  max mem: 57344
+train: [12]  [340/400]  eta: 0:00:36  lr: 0.000139  loss: 2.5997 (2.6088)  grad: 0.2111 (0.2173)  time: 0.6037  data: 0.0035  max mem: 57344
+train: [12]  [360/400]  eta: 0:00:24  lr: 0.000138  loss: 2.5701 (2.6065)  grad: 0.2122 (0.2172)  time: 0.6040  data: 0.0035  max mem: 57344
+train: [12]  [380/400]  eta: 0:00:12  lr: 0.000136  loss: 2.5640 (2.6082)  grad: 0.2159 (0.2172)  time: 0.6029  data: 0.0034  max mem: 57344
+train: [12]  [399/400]  eta: 0:00:00  lr: 0.000134  loss: 2.5779 (2.6073)  grad: 0.2133 (0.2171)  time: 0.6030  data: 0.0034  max mem: 57344
+train: [12] Total time: 0:04:02 (0.6053 s / it)
+train: [12] Summary: lr: 0.000134  loss: 2.5779 (2.6073)  grad: 0.2133 (0.2171)
+eval (validation): [12]  [ 0/85]  eta: 0:01:10    time: 0.8247  data: 0.4660  max mem: 57344
+eval (validation): [12]  [20/85]  eta: 0:00:25    time: 0.3661  data: 0.0025  max mem: 57344
+eval (validation): [12]  [40/85]  eta: 0:00:16    time: 0.3658  data: 0.0033  max mem: 57344
+eval (validation): [12]  [60/85]  eta: 0:00:09    time: 0.3661  data: 0.0033  max mem: 57344
+eval (validation): [12]  [80/85]  eta: 0:00:01    time: 0.3659  data: 0.0032  max mem: 57344
+eval (validation): [12]  [84/85]  eta: 0:00:00    time: 0.3592  data: 0.0032  max mem: 57344
+eval (validation): [12] Total time: 0:00:31 (0.3712 s / it)
+cv: [12]  best hparam: (1.9, 1.0) (028) ('028_lr1.9e+00_wd1.0e+00')  loss: 2.419  acc: 0.287  f1: 0.232
+saving checkpoint experiments/schaefer1000/output/schaefer1000/schaefer1000_lr3e-4_3/eval_v2/nsd_cococlip__patch__attn/checkpoint-last.pth
+train: [13]  [  0/400]  eta: 0:07:31  lr: nan  time: 1.1284  data: 0.5389  max mem: 57344
+train: [13]  [ 20/400]  eta: 0:03:58  lr: 0.000133  loss: 2.5106 (2.5540)  grad: 0.2156 (0.2149)  time: 0.6025  data: 0.0023  max mem: 57344
+train: [13]  [ 40/400]  eta: 0:03:41  lr: 0.000131  loss: 2.5418 (2.5604)  grad: 0.2156 (0.2143)  time: 0.6031  data: 0.0033  max mem: 57344
+train: [13]  [ 60/400]  eta: 0:03:27  lr: 0.000130  loss: 2.5418 (2.5628)  grad: 0.2073 (0.2130)  time: 0.6022  data: 0.0032  max mem: 57344
+train: [13]  [ 80/400]  eta: 0:03:14  lr: 0.000128  loss: 2.5297 (2.5641)  grad: 0.2072 (0.2121)  time: 0.6026  data: 0.0034  max mem: 57344
+train: [13]  [100/400]  eta: 0:03:02  lr: 0.000127  loss: 2.5731 (2.5704)  grad: 0.2053 (0.2110)  time: 0.6030  data: 0.0034  max mem: 57344
+train: [13]  [120/400]  eta: 0:02:49  lr: 0.000125  loss: 2.6089 (2.5817)  grad: 0.2077 (0.2134)  time: 0.6028  data: 0.0033  max mem: 57344
+train: [13]  [140/400]  eta: 0:02:37  lr: 0.000124  loss: 2.5835 (2.5846)  grad: 0.2116 (0.2133)  time: 0.6030  data: 0.0034  max mem: 57344
+train: [13]  [160/400]  eta: 0:02:25  lr: 0.000122  loss: 2.5964 (2.5904)  grad: 0.2116 (0.2140)  time: 0.6031  data: 0.0035  max mem: 57344
+train: [13]  [180/400]  eta: 0:02:13  lr: 0.000120  loss: 2.6150 (2.5899)  grad: 0.2163 (0.2146)  time: 0.6031  data: 0.0034  max mem: 57344
+train: [13]  [200/400]  eta: 0:02:01  lr: 0.000119  loss: 2.5840 (2.5864)  grad: 0.2142 (0.2150)  time: 0.6038  data: 0.0036  max mem: 57344
+train: [13]  [220/400]  eta: 0:01:48  lr: 0.000117  loss: 2.5492 (2.5850)  grad: 0.2142 (0.2151)  time: 0.6040  data: 0.0035  max mem: 57344
+train: [13]  [240/400]  eta: 0:01:36  lr: 0.000116  loss: 2.5544 (2.5823)  grad: 0.2123 (0.2146)  time: 0.6041  data: 0.0035  max mem: 57344
+train: [13]  [260/400]  eta: 0:01:24  lr: 0.000114  loss: 2.5555 (2.5804)  grad: 0.2075 (0.2138)  time: 0.6038  data: 0.0035  max mem: 57344
+train: [13]  [280/400]  eta: 0:01:12  lr: 0.000113  loss: 2.5534 (2.5789)  grad: 0.2102 (0.2142)  time: 0.6033  data: 0.0034  max mem: 57344
+train: [13]  [300/400]  eta: 0:01:00  lr: 0.000111  loss: 2.5510 (2.5783)  grad: 0.2176 (0.2145)  time: 0.6030  data: 0.0034  max mem: 57344
+train: [13]  [320/400]  eta: 0:00:48  lr: 0.000110  loss: 2.5759 (2.5781)  grad: 0.2152 (0.2150)  time: 0.6032  data: 0.0034  max mem: 57344
+train: [13]  [340/400]  eta: 0:00:36  lr: 0.000108  loss: 2.5786 (2.5802)  grad: 0.2202 (0.2153)  time: 0.6029  data: 0.0034  max mem: 57344
+train: [13]  [360/400]  eta: 0:00:24  lr: 0.000107  loss: 2.6260 (2.5822)  grad: 0.2226 (0.2160)  time: 0.6033  data: 0.0035  max mem: 57344
+train: [13]  [380/400]  eta: 0:00:12  lr: 0.000105  loss: 2.6096 (2.5824)  grad: 0.2205 (0.2164)  time: 0.6030  data: 0.0034  max mem: 57344
+train: [13]  [399/400]  eta: 0:00:00  lr: 0.000104  loss: 2.5746 (2.5829)  grad: 0.2205 (0.2167)  time: 0.6033  data: 0.0034  max mem: 57344
+train: [13] Total time: 0:04:01 (0.6047 s / it)
+train: [13] Summary: lr: 0.000104  loss: 2.5746 (2.5829)  grad: 0.2205 (0.2167)
+eval (validation): [13]  [ 0/85]  eta: 0:01:17    time: 0.9116  data: 0.5583  max mem: 57344
+eval (validation): [13]  [20/85]  eta: 0:00:25    time: 0.3655  data: 0.0028  max mem: 57344
+eval (validation): [13]  [40/85]  eta: 0:00:17    time: 0.3656  data: 0.0032  max mem: 57344
+eval (validation): [13]  [60/85]  eta: 0:00:09    time: 0.3656  data: 0.0034  max mem: 57344
+eval (validation): [13]  [80/85]  eta: 0:00:01    time: 0.3656  data: 0.0033  max mem: 57344
+eval (validation): [13]  [84/85]  eta: 0:00:00    time: 0.3594  data: 0.0033  max mem: 57344
+eval (validation): [13] Total time: 0:00:31 (0.3718 s / it)
+cv: [13]  best hparam: (3.1, 1.0) (031) ('031_lr3.1e+00_wd1.0e+00')  loss: 2.401  acc: 0.288  f1: 0.238
+saving checkpoint experiments/schaefer1000/output/schaefer1000/schaefer1000_lr3e-4_3/eval_v2/nsd_cococlip__patch__attn/checkpoint-last.pth
+train: [14]  [  0/400]  eta: 0:07:17  lr: nan  time: 1.0932  data: 0.5037  max mem: 57344
+train: [14]  [ 20/400]  eta: 0:03:57  lr: 0.000102  loss: 2.5709 (2.5564)  grad: 0.2217 (0.2203)  time: 0.6020  data: 0.0027  max mem: 57344
+train: [14]  [ 40/400]  eta: 0:03:41  lr: 0.000101  loss: 2.5866 (2.5801)  grad: 0.2183 (0.2184)  time: 0.6034  data: 0.0035  max mem: 57344
+train: [14]  [ 60/400]  eta: 0:03:27  lr: 0.000099  loss: 2.5886 (2.5715)  grad: 0.2082 (0.2123)  time: 0.6027  data: 0.0034  max mem: 57344
+train: [14]  [ 80/400]  eta: 0:03:14  lr: 0.000098  loss: 2.5442 (2.5709)  grad: 0.2006 (0.2132)  time: 0.6026  data: 0.0034  max mem: 57344
+train: [14]  [100/400]  eta: 0:03:02  lr: 0.000096  loss: 2.5382 (2.5658)  grad: 0.2144 (0.2145)  time: 0.6034  data: 0.0034  max mem: 57344
+train: [14]  [120/400]  eta: 0:02:49  lr: 0.000095  loss: 2.5382 (2.5585)  grad: 0.2142 (0.2134)  time: 0.6041  data: 0.0035  max mem: 57344
+train: [14]  [140/400]  eta: 0:02:37  lr: 0.000093  loss: 2.5416 (2.5592)  grad: 0.2079 (0.2142)  time: 0.6044  data: 0.0037  max mem: 57344
+train: [14]  [160/400]  eta: 0:02:25  lr: 0.000092  loss: 2.6034 (2.5663)  grad: 0.2098 (0.2142)  time: 0.6039  data: 0.0035  max mem: 57344
+train: [14]  [180/400]  eta: 0:02:13  lr: 0.000090  loss: 2.5848 (2.5603)  grad: 0.2132 (0.2145)  time: 0.6032  data: 0.0034  max mem: 57344
+train: [14]  [200/400]  eta: 0:02:01  lr: 0.000089  loss: 2.5229 (2.5626)  grad: 0.2131 (0.2143)  time: 0.6031  data: 0.0034  max mem: 57344
+train: [14]  [220/400]  eta: 0:01:48  lr: 0.000088  loss: 2.5371 (2.5620)  grad: 0.2095 (0.2136)  time: 0.6030  data: 0.0034  max mem: 57344
+train: [14]  [240/400]  eta: 0:01:36  lr: 0.000086  loss: 2.5527 (2.5635)  grad: 0.2081 (0.2136)  time: 0.6032  data: 0.0035  max mem: 57344
+train: [14]  [260/400]  eta: 0:01:24  lr: 0.000085  loss: 2.5537 (2.5633)  grad: 0.2138 (0.2138)  time: 0.6033  data: 0.0035  max mem: 57344
+train: [14]  [280/400]  eta: 0:01:12  lr: 0.000083  loss: 2.5701 (2.5642)  grad: 0.2138 (0.2138)  time: 0.6040  data: 0.0035  max mem: 57344
+train: [14]  [300/400]  eta: 0:01:00  lr: 0.000082  loss: 2.5740 (2.5659)  grad: 0.2088 (0.2140)  time: 0.6047  data: 0.0037  max mem: 57344
+train: [14]  [320/400]  eta: 0:00:48  lr: 0.000081  loss: 2.5508 (2.5651)  grad: 0.2107 (0.2137)  time: 0.6042  data: 0.0037  max mem: 57344
+train: [14]  [340/400]  eta: 0:00:36  lr: 0.000079  loss: 2.5462 (2.5644)  grad: 0.2064 (0.2132)  time: 0.6040  data: 0.0035  max mem: 57344
+train: [14]  [360/400]  eta: 0:00:24  lr: 0.000078  loss: 2.5563 (2.5641)  grad: 0.2046 (0.2131)  time: 0.6043  data: 0.0037  max mem: 57344
+train: [14]  [380/400]  eta: 0:00:12  lr: 0.000076  loss: 2.5598 (2.5634)  grad: 0.2066 (0.2131)  time: 0.6045  data: 0.0038  max mem: 57344
+train: [14]  [399/400]  eta: 0:00:00  lr: 0.000075  loss: 2.5491 (2.5623)  grad: 0.2114 (0.2134)  time: 0.6039  data: 0.0036  max mem: 57344
+train: [14] Total time: 0:04:02 (0.6051 s / it)
+train: [14] Summary: lr: 0.000075  loss: 2.5491 (2.5623)  grad: 0.2114 (0.2134)
+eval (validation): [14]  [ 0/85]  eta: 0:01:22    time: 0.9702  data: 0.6140  max mem: 57344
+eval (validation): [14]  [20/85]  eta: 0:00:25    time: 0.3651  data: 0.0026  max mem: 57344
+eval (validation): [14]  [40/85]  eta: 0:00:17    time: 0.3660  data: 0.0035  max mem: 57344
+eval (validation): [14]  [60/85]  eta: 0:00:09    time: 0.3667  data: 0.0036  max mem: 57344
+eval (validation): [14]  [80/85]  eta: 0:00:01    time: 0.3662  data: 0.0035  max mem: 57344
+eval (validation): [14]  [84/85]  eta: 0:00:00    time: 0.3602  data: 0.0035  max mem: 57344
+eval (validation): [14] Total time: 0:00:31 (0.3730 s / it)
+cv: [14]  best hparam: (1.2, 1.0) (025) ('025_lr1.2e+00_wd1.0e+00')  loss: 2.374  acc: 0.287  f1: 0.230
+saving checkpoint experiments/schaefer1000/output/schaefer1000/schaefer1000_lr3e-4_3/eval_v2/nsd_cococlip__patch__attn/checkpoint-last.pth
+train: [15]  [  0/400]  eta: 0:08:38  lr: nan  time: 1.2968  data: 0.7039  max mem: 57344
+train: [15]  [ 20/400]  eta: 0:04:01  lr: 0.000074  loss: 2.5008 (2.5215)  grad: 0.2122 (0.2190)  time: 0.6036  data: 0.0030  max mem: 57344
+train: [15]  [ 40/400]  eta: 0:03:43  lr: 0.000072  loss: 2.4951 (2.5125)  grad: 0.2134 (0.2174)  time: 0.6041  data: 0.0036  max mem: 57344
+train: [15]  [ 60/400]  eta: 0:03:29  lr: 0.000071  loss: 2.5065 (2.5278)  grad: 0.2207 (0.2185)  time: 0.6029  data: 0.0035  max mem: 57344
+train: [15]  [ 80/400]  eta: 0:03:15  lr: 0.000070  loss: 2.5853 (2.5326)  grad: 0.2156 (0.2182)  time: 0.6028  data: 0.0034  max mem: 57344
+train: [15]  [100/400]  eta: 0:03:03  lr: 0.000068  loss: 2.5535 (2.5349)  grad: 0.2151 (0.2183)  time: 0.6027  data: 0.0033  max mem: 57344
+train: [15]  [120/400]  eta: 0:02:50  lr: 0.000067  loss: 2.5422 (2.5365)  grad: 0.2145 (0.2179)  time: 0.6030  data: 0.0033  max mem: 57344
+train: [15]  [140/400]  eta: 0:02:38  lr: 0.000066  loss: 2.5391 (2.5396)  grad: 0.2189 (0.2186)  time: 0.6027  data: 0.0034  max mem: 57344
+train: [15]  [160/400]  eta: 0:02:25  lr: 0.000064  loss: 2.5248 (2.5370)  grad: 0.2148 (0.2176)  time: 0.6021  data: 0.0033  max mem: 57344
+train: [15]  [180/400]  eta: 0:02:13  lr: 0.000063  loss: 2.4965 (2.5353)  grad: 0.2050 (0.2169)  time: 0.6029  data: 0.0032  max mem: 57344
+train: [15]  [200/400]  eta: 0:02:01  lr: 0.000062  loss: 2.5382 (2.5403)  grad: 0.2154 (0.2178)  time: 0.6028  data: 0.0034  max mem: 57344
+train: [15]  [220/400]  eta: 0:01:49  lr: 0.000061  loss: 2.5445 (2.5418)  grad: 0.2164 (0.2175)  time: 0.6038  data: 0.0035  max mem: 57344
+train: [15]  [240/400]  eta: 0:01:36  lr: 0.000059  loss: 2.5270 (2.5370)  grad: 0.2152 (0.2172)  time: 0.6036  data: 0.0035  max mem: 57344
+train: [15]  [260/400]  eta: 0:01:24  lr: 0.000058  loss: 2.5138 (2.5378)  grad: 0.2123 (0.2169)  time: 0.6033  data: 0.0035  max mem: 57344
+train: [15]  [280/400]  eta: 0:01:12  lr: 0.000057  loss: 2.5397 (2.5384)  grad: 0.2109 (0.2169)  time: 0.6025  data: 0.0033  max mem: 57344
+train: [15]  [300/400]  eta: 0:01:00  lr: 0.000056  loss: 2.5262 (2.5360)  grad: 0.2076 (0.2162)  time: 0.6035  data: 0.0034  max mem: 57344
+train: [15]  [320/400]  eta: 0:00:48  lr: 0.000054  loss: 2.5296 (2.5388)  grad: 0.2089 (0.2161)  time: 0.6034  data: 0.0034  max mem: 57344
+train: [15]  [340/400]  eta: 0:00:36  lr: 0.000053  loss: 2.5262 (2.5356)  grad: 0.2125 (0.2157)  time: 0.6031  data: 0.0034  max mem: 57344
+train: [15]  [360/400]  eta: 0:00:24  lr: 0.000052  loss: 2.5141 (2.5352)  grad: 0.2096 (0.2154)  time: 0.6048  data: 0.0036  max mem: 57344
+train: [15]  [380/400]  eta: 0:00:12  lr: 0.000051  loss: 2.5462 (2.5358)  grad: 0.2096 (0.2152)  time: 0.6045  data: 0.0036  max mem: 57344
+train: [15]  [399/400]  eta: 0:00:00  lr: 0.000050  loss: 2.5493 (2.5363)  grad: 0.2135 (0.2157)  time: 0.6046  data: 0.0036  max mem: 57344
+train: [15] Total time: 0:04:02 (0.6053 s / it)
+train: [15] Summary: lr: 0.000050  loss: 2.5493 (2.5363)  grad: 0.2135 (0.2157)
+eval (validation): [15]  [ 0/85]  eta: 0:01:14    time: 0.8778  data: 0.5196  max mem: 57344
+eval (validation): [15]  [20/85]  eta: 0:00:25    time: 0.3659  data: 0.0032  max mem: 57344
+eval (validation): [15]  [40/85]  eta: 0:00:17    time: 0.3662  data: 0.0036  max mem: 57344
+eval (validation): [15]  [60/85]  eta: 0:00:09    time: 0.3665  data: 0.0036  max mem: 57344
+eval (validation): [15]  [80/85]  eta: 0:00:01    time: 0.3667  data: 0.0036  max mem: 57344
+eval (validation): [15]  [84/85]  eta: 0:00:00    time: 0.3605  data: 0.0035  max mem: 57344
+eval (validation): [15] Total time: 0:00:31 (0.3722 s / it)
+cv: [15]  best hparam: (1.2, 1.0) (025) ('025_lr1.2e+00_wd1.0e+00')  loss: 2.353  acc: 0.295  f1: 0.239
+saving checkpoint experiments/schaefer1000/output/schaefer1000/schaefer1000_lr3e-4_3/eval_v2/nsd_cococlip__patch__attn/checkpoint-last.pth
+saving best checkpoint experiments/schaefer1000/output/schaefer1000/schaefer1000_lr3e-4_3/eval_v2/nsd_cococlip__patch__attn/checkpoint-best.pth
+train: [16]  [  0/400]  eta: 0:08:09  lr: nan  time: 1.2229  data: 0.6320  max mem: 57344
+train: [16]  [ 20/400]  eta: 0:04:00  lr: 0.000048  loss: 2.4650 (2.5355)  grad: 0.2141 (0.2164)  time: 0.6043  data: 0.0034  max mem: 57344
+train: [16]  [ 40/400]  eta: 0:03:42  lr: 0.000047  loss: 2.4718 (2.5228)  grad: 0.2104 (0.2124)  time: 0.6039  data: 0.0035  max mem: 57344
+train: [16]  [ 60/400]  eta: 0:03:28  lr: 0.000046  loss: 2.4988 (2.5118)  grad: 0.2034 (0.2105)  time: 0.6033  data: 0.0034  max mem: 57344
+train: [16]  [ 80/400]  eta: 0:03:15  lr: 0.000045  loss: 2.5118 (2.5103)  grad: 0.2081 (0.2112)  time: 0.6039  data: 0.0035  max mem: 57344
+train: [16]  [100/400]  eta: 0:03:02  lr: 0.000044  loss: 2.5114 (2.5057)  grad: 0.2111 (0.2107)  time: 0.6037  data: 0.0035  max mem: 57344
+train: [16]  [120/400]  eta: 0:02:50  lr: 0.000043  loss: 2.5270 (2.5151)  grad: 0.2049 (0.2108)  time: 0.6034  data: 0.0035  max mem: 57344
+train: [16]  [140/400]  eta: 0:02:38  lr: 0.000042  loss: 2.5470 (2.5132)  grad: 0.2022 (0.2099)  time: 0.6034  data: 0.0034  max mem: 57344
+train: [16]  [160/400]  eta: 0:02:25  lr: 0.000041  loss: 2.5072 (2.5140)  grad: 0.2041 (0.2098)  time: 0.6025  data: 0.0033  max mem: 57344
+train: [16]  [180/400]  eta: 0:02:13  lr: 0.000040  loss: 2.5072 (2.5160)  grad: 0.2044 (0.2099)  time: 0.6031  data: 0.0033  max mem: 57344
+train: [16]  [200/400]  eta: 0:02:01  lr: 0.000039  loss: 2.5248 (2.5167)  grad: 0.2046 (0.2096)  time: 0.6028  data: 0.0034  max mem: 57344
+train: [16]  [220/400]  eta: 0:01:49  lr: 0.000038  loss: 2.4860 (2.5161)  grad: 0.2115 (0.2103)  time: 0.6030  data: 0.0034  max mem: 57344
+train: [16]  [240/400]  eta: 0:01:36  lr: 0.000036  loss: 2.5360 (2.5190)  grad: 0.2114 (0.2100)  time: 0.6032  data: 0.0034  max mem: 57344
+train: [16]  [260/400]  eta: 0:01:24  lr: 0.000035  loss: 2.5362 (2.5183)  grad: 0.2079 (0.2104)  time: 0.6026  data: 0.0033  max mem: 57344
+train: [16]  [280/400]  eta: 0:01:12  lr: 0.000034  loss: 2.5143 (2.5183)  grad: 0.2158 (0.2111)  time: 0.6027  data: 0.0033  max mem: 57344
+train: [16]  [300/400]  eta: 0:01:00  lr: 0.000033  loss: 2.5143 (2.5180)  grad: 0.2179 (0.2114)  time: 0.6031  data: 0.0035  max mem: 57344
+train: [16]  [320/400]  eta: 0:00:48  lr: 0.000032  loss: 2.5420 (2.5204)  grad: 0.2075 (0.2110)  time: 0.6029  data: 0.0035  max mem: 57344
+train: [16]  [340/400]  eta: 0:00:36  lr: 0.000031  loss: 2.5472 (2.5216)  grad: 0.2083 (0.2115)  time: 0.6031  data: 0.0034  max mem: 57344
+train: [16]  [360/400]  eta: 0:00:24  lr: 0.000031  loss: 2.5309 (2.5224)  grad: 0.2129 (0.2116)  time: 0.6029  data: 0.0035  max mem: 57344
+train: [16]  [380/400]  eta: 0:00:12  lr: 0.000030  loss: 2.5106 (2.5225)  grad: 0.2035 (0.2113)  time: 0.6030  data: 0.0034  max mem: 57344
+train: [16]  [399/400]  eta: 0:00:00  lr: 0.000029  loss: 2.5247 (2.5231)  grad: 0.2030 (0.2111)  time: 0.6030  data: 0.0034  max mem: 57344
+train: [16] Total time: 0:04:01 (0.6050 s / it)
+train: [16] Summary: lr: 0.000029  loss: 2.5247 (2.5231)  grad: 0.2030 (0.2111)
+eval (validation): [16]  [ 0/85]  eta: 0:01:15    time: 0.8941  data: 0.5381  max mem: 57344
+eval (validation): [16]  [20/85]  eta: 0:00:25    time: 0.3669  data: 0.0027  max mem: 57344
+eval (validation): [16]  [40/85]  eta: 0:00:17    time: 0.3675  data: 0.0035  max mem: 57344
+eval (validation): [16]  [60/85]  eta: 0:00:09    time: 0.3671  data: 0.0037  max mem: 57344
+eval (validation): [16]  [80/85]  eta: 0:00:01    time: 0.3670  data: 0.0035  max mem: 57344
+eval (validation): [16]  [84/85]  eta: 0:00:00    time: 0.3607  data: 0.0035  max mem: 57344
+eval (validation): [16] Total time: 0:00:31 (0.3732 s / it)
+cv: [16]  best hparam: (1.9, 1.0) (028) ('028_lr1.9e+00_wd1.0e+00')  loss: 2.395  acc: 0.291  f1: 0.243
+saving checkpoint experiments/schaefer1000/output/schaefer1000/schaefer1000_lr3e-4_3/eval_v2/nsd_cococlip__patch__attn/checkpoint-last.pth
+train: [17]  [  0/400]  eta: 0:08:50  lr: nan  time: 1.3274  data: 0.7337  max mem: 57344
+train: [17]  [ 20/400]  eta: 0:04:02  lr: 0.000028  loss: 2.4728 (2.5255)  grad: 0.2032 (0.2031)  time: 0.6035  data: 0.0029  max mem: 57344
+train: [17]  [ 40/400]  eta: 0:03:43  lr: 0.000027  loss: 2.4630 (2.4860)  grad: 0.2089 (0.2097)  time: 0.6040  data: 0.0036  max mem: 57344
+train: [17]  [ 60/400]  eta: 0:03:29  lr: 0.000026  loss: 2.4630 (2.4857)  grad: 0.2086 (0.2092)  time: 0.6035  data: 0.0035  max mem: 57344
+train: [17]  [ 80/400]  eta: 0:03:16  lr: 0.000025  loss: 2.5422 (2.5052)  grad: 0.2053 (0.2090)  time: 0.6035  data: 0.0035  max mem: 57344
+train: [17]  [100/400]  eta: 0:03:03  lr: 0.000024  loss: 2.5732 (2.5156)  grad: 0.2088 (0.2098)  time: 0.6032  data: 0.0034  max mem: 57344
+train: [17]  [120/400]  eta: 0:02:50  lr: 0.000023  loss: 2.5213 (2.5116)  grad: 0.2116 (0.2105)  time: 0.6037  data: 0.0034  max mem: 57344
+train: [17]  [140/400]  eta: 0:02:38  lr: 0.000023  loss: 2.4868 (2.5048)  grad: 0.2110 (0.2100)  time: 0.6038  data: 0.0035  max mem: 57344
+train: [17]  [160/400]  eta: 0:02:25  lr: 0.000022  loss: 2.4731 (2.5075)  grad: 0.2058 (0.2104)  time: 0.6033  data: 0.0034  max mem: 57344
+train: [17]  [180/400]  eta: 0:02:13  lr: 0.000021  loss: 2.4731 (2.5081)  grad: 0.2085 (0.2104)  time: 0.6034  data: 0.0034  max mem: 57344
+train: [17]  [200/400]  eta: 0:02:01  lr: 0.000020  loss: 2.5376 (2.5101)  grad: 0.2000 (0.2103)  time: 0.6043  data: 0.0036  max mem: 57344
+train: [17]  [220/400]  eta: 0:01:49  lr: 0.000019  loss: 2.5400 (2.5094)  grad: 0.2049 (0.2101)  time: 0.6030  data: 0.0034  max mem: 57344
+train: [17]  [240/400]  eta: 0:01:37  lr: 0.000019  loss: 2.4624 (2.5076)  grad: 0.2061 (0.2096)  time: 0.6030  data: 0.0034  max mem: 57344
+train: [17]  [260/400]  eta: 0:01:24  lr: 0.000018  loss: 2.4760 (2.5079)  grad: 0.2072 (0.2100)  time: 0.6029  data: 0.0033  max mem: 57344
+train: [17]  [280/400]  eta: 0:01:12  lr: 0.000017  loss: 2.4980 (2.5074)  grad: 0.2088 (0.2104)  time: 0.6028  data: 0.0033  max mem: 57344
+train: [17]  [300/400]  eta: 0:01:00  lr: 0.000016  loss: 2.5065 (2.5060)  grad: 0.2098 (0.2105)  time: 0.6031  data: 0.0033  max mem: 57344
+train: [17]  [320/400]  eta: 0:00:48  lr: 0.000016  loss: 2.5118 (2.5072)  grad: 0.2083 (0.2101)  time: 0.6031  data: 0.0033  max mem: 57344
+train: [17]  [340/400]  eta: 0:00:36  lr: 0.000015  loss: 2.5194 (2.5067)  grad: 0.2060 (0.2098)  time: 0.6027  data: 0.0033  max mem: 57344
+train: [17]  [360/400]  eta: 0:00:24  lr: 0.000014  loss: 2.5145 (2.5060)  grad: 0.2084 (0.2101)  time: 0.6039  data: 0.0034  max mem: 57344
+train: [17]  [380/400]  eta: 0:00:12  lr: 0.000014  loss: 2.5145 (2.5076)  grad: 0.2091 (0.2101)  time: 0.6038  data: 0.0035  max mem: 57344
+train: [17]  [399/400]  eta: 0:00:00  lr: 0.000013  loss: 2.5141 (2.5072)  grad: 0.2091 (0.2102)  time: 0.6033  data: 0.0034  max mem: 57344
+train: [17] Total time: 0:04:02 (0.6055 s / it)
+train: [17] Summary: lr: 0.000013  loss: 2.5141 (2.5072)  grad: 0.2091 (0.2102)
+eval (validation): [17]  [ 0/85]  eta: 0:01:17    time: 0.9078  data: 0.5488  max mem: 57344
+eval (validation): [17]  [20/85]  eta: 0:00:25    time: 0.3653  data: 0.0021  max mem: 57344
+eval (validation): [17]  [40/85]  eta: 0:00:17    time: 0.3658  data: 0.0033  max mem: 57344
+eval (validation): [17]  [60/85]  eta: 0:00:09    time: 0.3663  data: 0.0034  max mem: 57344
+eval (validation): [17]  [80/85]  eta: 0:00:01    time: 0.3656  data: 0.0032  max mem: 57344
+eval (validation): [17]  [84/85]  eta: 0:00:00    time: 0.3593  data: 0.0032  max mem: 57344
+eval (validation): [17] Total time: 0:00:31 (0.3720 s / it)
+cv: [17]  best hparam: (1.2, 1.0) (025) ('025_lr1.2e+00_wd1.0e+00')  loss: 2.356  acc: 0.295  f1: 0.237
+saving checkpoint experiments/schaefer1000/output/schaefer1000/schaefer1000_lr3e-4_3/eval_v2/nsd_cococlip__patch__attn/checkpoint-last.pth
+saving best checkpoint experiments/schaefer1000/output/schaefer1000/schaefer1000_lr3e-4_3/eval_v2/nsd_cococlip__patch__attn/checkpoint-best.pth
+train: [18]  [  0/400]  eta: 0:07:21  lr: nan  time: 1.1038  data: 0.5107  max mem: 57344
+train: [18]  [ 20/400]  eta: 0:03:58  lr: 0.000012  loss: 2.4481 (2.4539)  grad: 0.2020 (0.2078)  time: 0.6031  data: 0.0028  max mem: 57344
+train: [18]  [ 40/400]  eta: 0:03:41  lr: 0.000012  loss: 2.4481 (2.4498)  grad: 0.2066 (0.2057)  time: 0.6032  data: 0.0034  max mem: 57344
+train: [18]  [ 60/400]  eta: 0:03:27  lr: 0.000011  loss: 2.4505 (2.4501)  grad: 0.2042 (0.2072)  time: 0.6044  data: 0.0036  max mem: 57344
+train: [18]  [ 80/400]  eta: 0:03:15  lr: 0.000011  loss: 2.4525 (2.4591)  grad: 0.2018 (0.2062)  time: 0.6047  data: 0.0037  max mem: 57344
+train: [18]  [100/400]  eta: 0:03:02  lr: 0.000010  loss: 2.5048 (2.4750)  grad: 0.2107 (0.2087)  time: 0.6048  data: 0.0037  max mem: 57344
+train: [18]  [120/400]  eta: 0:02:50  lr: 0.000009  loss: 2.5285 (2.4812)  grad: 0.2145 (0.2090)  time: 0.6041  data: 0.0036  max mem: 57344
+train: [18]  [140/400]  eta: 0:02:37  lr: 0.000009  loss: 2.4698 (2.4757)  grad: 0.2066 (0.2075)  time: 0.6040  data: 0.0035  max mem: 57344
+train: [18]  [160/400]  eta: 0:02:25  lr: 0.000008  loss: 2.4900 (2.4797)  grad: 0.2045 (0.2081)  time: 0.6036  data: 0.0035  max mem: 57344
+train: [18]  [180/400]  eta: 0:02:13  lr: 0.000008  loss: 2.5307 (2.4840)  grad: 0.2087 (0.2085)  time: 0.6041  data: 0.0035  max mem: 57344
+train: [18]  [200/400]  eta: 0:02:01  lr: 0.000007  loss: 2.4726 (2.4792)  grad: 0.2057 (0.2082)  time: 0.6039  data: 0.0036  max mem: 57344
+train: [18]  [220/400]  eta: 0:01:49  lr: 0.000007  loss: 2.4875 (2.4820)  grad: 0.1989 (0.2085)  time: 0.6037  data: 0.0035  max mem: 57344
+train: [18]  [240/400]  eta: 0:01:36  lr: 0.000006  loss: 2.5055 (2.4855)  grad: 0.2142 (0.2091)  time: 0.6042  data: 0.0036  max mem: 57344
+train: [18]  [260/400]  eta: 0:01:24  lr: 0.000006  loss: 2.4843 (2.4827)  grad: 0.2142 (0.2097)  time: 0.6039  data: 0.0035  max mem: 57344
+train: [18]  [280/400]  eta: 0:01:12  lr: 0.000006  loss: 2.4380 (2.4806)  grad: 0.2054 (0.2092)  time: 0.6036  data: 0.0036  max mem: 57344
+train: [18]  [300/400]  eta: 0:01:00  lr: 0.000005  loss: 2.4590 (2.4818)  grad: 0.2054 (0.2098)  time: 0.6029  data: 0.0034  max mem: 57344
+train: [18]  [320/400]  eta: 0:00:48  lr: 0.000005  loss: 2.4754 (2.4811)  grad: 0.2112 (0.2101)  time: 0.6030  data: 0.0033  max mem: 57344
+train: [18]  [340/400]  eta: 0:00:36  lr: 0.000004  loss: 2.4669 (2.4806)  grad: 0.2117 (0.2102)  time: 0.6031  data: 0.0034  max mem: 57344
+train: [18]  [360/400]  eta: 0:00:24  lr: 0.000004  loss: 2.4796 (2.4813)  grad: 0.2083 (0.2098)  time: 0.6029  data: 0.0034  max mem: 57344
+train: [18]  [380/400]  eta: 0:00:12  lr: 0.000004  loss: 2.4848 (2.4816)  grad: 0.2051 (0.2099)  time: 0.6032  data: 0.0033  max mem: 57344
+train: [18]  [399/400]  eta: 0:00:00  lr: 0.000003  loss: 2.5020 (2.4839)  grad: 0.2078 (0.2104)  time: 0.6029  data: 0.0034  max mem: 57344
+train: [18] Total time: 0:04:02 (0.6052 s / it)
+train: [18] Summary: lr: 0.000003  loss: 2.5020 (2.4839)  grad: 0.2078 (0.2104)
+eval (validation): [18]  [ 0/85]  eta: 0:01:05    time: 0.7763  data: 0.4196  max mem: 57344
+eval (validation): [18]  [20/85]  eta: 0:00:25    time: 0.3658  data: 0.0027  max mem: 57344
+eval (validation): [18]  [40/85]  eta: 0:00:16    time: 0.3657  data: 0.0033  max mem: 57344
+eval (validation): [18]  [60/85]  eta: 0:00:09    time: 0.3668  data: 0.0035  max mem: 57344
+eval (validation): [18]  [80/85]  eta: 0:00:01    time: 0.3668  data: 0.0036  max mem: 57344
+eval (validation): [18]  [84/85]  eta: 0:00:00    time: 0.3604  data: 0.0036  max mem: 57344
+eval (validation): [18] Total time: 0:00:31 (0.3709 s / it)
+cv: [18]  best hparam: (1.2, 1.0) (025) ('025_lr1.2e+00_wd1.0e+00')  loss: 2.362  acc: 0.294  f1: 0.235
+saving checkpoint experiments/schaefer1000/output/schaefer1000/schaefer1000_lr3e-4_3/eval_v2/nsd_cococlip__patch__attn/checkpoint-last.pth
+train: [19]  [  0/400]  eta: 0:08:30  lr: nan  time: 1.2765  data: 0.6853  max mem: 57344
+train: [19]  [ 20/400]  eta: 0:04:01  lr: 0.000003  loss: 2.4931 (2.4994)  grad: 0.2056 (0.2079)  time: 0.6023  data: 0.0031  max mem: 57344
+train: [19]  [ 40/400]  eta: 0:03:42  lr: 0.000003  loss: 2.4931 (2.5163)  grad: 0.2073 (0.2102)  time: 0.6032  data: 0.0034  max mem: 57344
+train: [19]  [ 60/400]  eta: 0:03:28  lr: 0.000002  loss: 2.4772 (2.5054)  grad: 0.2073 (0.2105)  time: 0.6023  data: 0.0033  max mem: 57344
+train: [19]  [ 80/400]  eta: 0:03:15  lr: 0.000002  loss: 2.4493 (2.5014)  grad: 0.2117 (0.2126)  time: 0.6024  data: 0.0033  max mem: 57344
+train: [19]  [100/400]  eta: 0:03:02  lr: 0.000002  loss: 2.5103 (2.5096)  grad: 0.2117 (0.2116)  time: 0.6039  data: 0.0034  max mem: 57344
+train: [19]  [120/400]  eta: 0:02:50  lr: 0.000002  loss: 2.5103 (2.5007)  grad: 0.2011 (0.2095)  time: 0.6033  data: 0.0034  max mem: 57344
+train: [19]  [140/400]  eta: 0:02:38  lr: 0.000001  loss: 2.4788 (2.4982)  grad: 0.2032 (0.2091)  time: 0.6046  data: 0.0036  max mem: 57344
+train: [19]  [160/400]  eta: 0:02:25  lr: 0.000001  loss: 2.4788 (2.4984)  grad: 0.2083 (0.2093)  time: 0.6044  data: 0.0036  max mem: 57344
+train: [19]  [180/400]  eta: 0:02:13  lr: 0.000001  loss: 2.4761 (2.4949)  grad: 0.2083 (0.2086)  time: 0.6046  data: 0.0037  max mem: 57344
+train: [19]  [200/400]  eta: 0:02:01  lr: 0.000001  loss: 2.4752 (2.4968)  grad: 0.2102 (0.2092)  time: 0.6042  data: 0.0035  max mem: 57344
+train: [19]  [220/400]  eta: 0:01:49  lr: 0.000001  loss: 2.4752 (2.4938)  grad: 0.2076 (0.2090)  time: 0.6039  data: 0.0035  max mem: 57344
+train: [19]  [240/400]  eta: 0:01:37  lr: 0.000001  loss: 2.4366 (2.4909)  grad: 0.2032 (0.2084)  time: 0.6037  data: 0.0035  max mem: 57344
+train: [19]  [260/400]  eta: 0:01:24  lr: 0.000000  loss: 2.4334 (2.4898)  grad: 0.2009 (0.2081)  time: 0.6042  data: 0.0035  max mem: 57344
+train: [19]  [280/400]  eta: 0:01:12  lr: 0.000000  loss: 2.5043 (2.4929)  grad: 0.2034 (0.2080)  time: 0.6030  data: 0.0035  max mem: 57344
+train: [19]  [300/400]  eta: 0:01:00  lr: 0.000000  loss: 2.5072 (2.4931)  grad: 0.2023 (0.2077)  time: 0.6037  data: 0.0034  max mem: 57344
+train: [19]  [320/400]  eta: 0:00:48  lr: 0.000000  loss: 2.5053 (2.4911)  grad: 0.2030 (0.2079)  time: 0.6035  data: 0.0034  max mem: 57344
+train: [19]  [340/400]  eta: 0:00:36  lr: 0.000000  loss: 2.4818 (2.4907)  grad: 0.2030 (0.2075)  time: 0.6040  data: 0.0034  max mem: 57344
+train: [19]  [360/400]  eta: 0:00:24  lr: 0.000000  loss: 2.4941 (2.4930)  grad: 0.2030 (0.2080)  time: 0.6036  data: 0.0035  max mem: 57344
+train: [19]  [380/400]  eta: 0:00:12  lr: 0.000000  loss: 2.5045 (2.4935)  grad: 0.2061 (0.2078)  time: 0.6030  data: 0.0033  max mem: 57344
+train: [19]  [399/400]  eta: 0:00:00  lr: 0.000000  loss: 2.5045 (2.4945)  grad: 0.2022 (0.2075)  time: 0.6030  data: 0.0034  max mem: 57344
+train: [19] Total time: 0:04:02 (0.6055 s / it)
+train: [19] Summary: lr: 0.000000  loss: 2.5045 (2.4945)  grad: 0.2022 (0.2075)
+eval (validation): [19]  [ 0/85]  eta: 0:01:10    time: 0.8352  data: 0.4793  max mem: 57344
+eval (validation): [19]  [20/85]  eta: 0:00:25    time: 0.3661  data: 0.0027  max mem: 57344
+eval (validation): [19]  [40/85]  eta: 0:00:16    time: 0.3657  data: 0.0035  max mem: 57344
+eval (validation): [19]  [60/85]  eta: 0:00:09    time: 0.3668  data: 0.0034  max mem: 57344
+eval (validation): [19]  [80/85]  eta: 0:00:01    time: 0.3666  data: 0.0033  max mem: 57344
+eval (validation): [19]  [84/85]  eta: 0:00:00    time: 0.3605  data: 0.0033  max mem: 57344
+eval (validation): [19] Total time: 0:00:31 (0.3716 s / it)
+cv: [19]  best hparam: (1.2, 1.0) (025) ('025_lr1.2e+00_wd1.0e+00')  loss: 2.365  acc: 0.292  f1: 0.235
+saving checkpoint experiments/schaefer1000/output/schaefer1000/schaefer1000_lr3e-4_3/eval_v2/nsd_cococlip__patch__attn/checkpoint-last.pth
+evaluating last checkpoint: experiments/schaefer1000/output/schaefer1000/schaefer1000_lr3e-4_3/eval_v2/nsd_cococlip__patch__attn/checkpoint-last.pth
+eval model info:
+{"score": 0.292358803986711, "hparam": [1.2, 1.0], "hparam_id": 25, "epoch": 19, "is_best": false, "best_score": 0.29549649317091176}
+eval (train): [20]  [  0/509]  eta: 0:07:21    time: 0.8680  data: 0.5095  max mem: 57344
+eval (train): [20]  [ 20/509]  eta: 0:03:11    time: 0.3670  data: 0.0029  max mem: 57344
+eval (train): [20]  [ 40/509]  eta: 0:02:58    time: 0.3677  data: 0.0034  max mem: 57344
+eval (train): [20]  [ 60/509]  eta: 0:02:48    time: 0.3679  data: 0.0034  max mem: 57344
+eval (train): [20]  [ 80/509]  eta: 0:02:40    time: 0.3674  data: 0.0034  max mem: 57344
+eval (train): [20]  [100/509]  eta: 0:02:32    time: 0.3682  data: 0.0035  max mem: 57344
+eval (train): [20]  [120/509]  eta: 0:02:24    time: 0.3687  data: 0.0037  max mem: 57344
+eval (train): [20]  [140/509]  eta: 0:02:17    time: 0.3685  data: 0.0035  max mem: 57344
+eval (train): [20]  [160/509]  eta: 0:02:09    time: 0.3685  data: 0.0034  max mem: 57344
+eval (train): [20]  [180/509]  eta: 0:02:01    time: 0.3680  data: 0.0034  max mem: 57344
+eval (train): [20]  [200/509]  eta: 0:01:54    time: 0.3681  data: 0.0032  max mem: 57344
+eval (train): [20]  [220/509]  eta: 0:01:47    time: 0.3679  data: 0.0033  max mem: 57344
+eval (train): [20]  [240/509]  eta: 0:01:39    time: 0.3685  data: 0.0033  max mem: 57344
+eval (train): [20]  [260/509]  eta: 0:01:32    time: 0.3685  data: 0.0034  max mem: 57344
+eval (train): [20]  [280/509]  eta: 0:01:24    time: 0.3671  data: 0.0035  max mem: 57344
+eval (train): [20]  [300/509]  eta: 0:01:17    time: 0.3665  data: 0.0033  max mem: 57344
+eval (train): [20]  [320/509]  eta: 0:01:09    time: 0.3667  data: 0.0034  max mem: 57344
+eval (train): [20]  [340/509]  eta: 0:01:02    time: 0.3679  data: 0.0035  max mem: 57344
+eval (train): [20]  [360/509]  eta: 0:00:55    time: 0.3669  data: 0.0037  max mem: 57344
+eval (train): [20]  [380/509]  eta: 0:00:47    time: 0.3675  data: 0.0035  max mem: 57344
+eval (train): [20]  [400/509]  eta: 0:00:40    time: 0.3669  data: 0.0036  max mem: 57344
+eval (train): [20]  [420/509]  eta: 0:00:32    time: 0.3670  data: 0.0036  max mem: 57344
+eval (train): [20]  [440/509]  eta: 0:00:25    time: 0.3666  data: 0.0036  max mem: 57344
+eval (train): [20]  [460/509]  eta: 0:00:18    time: 0.3663  data: 0.0034  max mem: 57344
+eval (train): [20]  [480/509]  eta: 0:00:10    time: 0.3667  data: 0.0034  max mem: 57344
+eval (train): [20]  [500/509]  eta: 0:00:03    time: 0.3663  data: 0.0034  max mem: 57344
+eval (train): [20]  [508/509]  eta: 0:00:00    time: 0.3564  data: 0.0034  max mem: 57344
+eval (train): [20] Total time: 0:03:07 (0.3684 s / it)
+eval (validation): [20]  [ 0/85]  eta: 0:01:29    time: 1.0489  data: 0.6913  max mem: 57344
+eval (validation): [20]  [20/85]  eta: 0:00:25    time: 0.3669  data: 0.0028  max mem: 57344
+eval (validation): [20]  [40/85]  eta: 0:00:17    time: 0.3678  data: 0.0035  max mem: 57344
+eval (validation): [20]  [60/85]  eta: 0:00:09    time: 0.3679  data: 0.0035  max mem: 57344
+eval (validation): [20]  [80/85]  eta: 0:00:01    time: 0.3676  data: 0.0035  max mem: 57344
+eval (validation): [20]  [84/85]  eta: 0:00:00    time: 0.3615  data: 0.0035  max mem: 57344
+eval (validation): [20] Total time: 0:00:31 (0.3755 s / it)
+eval (test): [20]  [ 0/85]  eta: 0:01:27    time: 1.0303  data: 0.6724  max mem: 57344
+eval (test): [20]  [20/85]  eta: 0:00:25    time: 0.3652  data: 0.0026  max mem: 57344
+eval (test): [20]  [40/85]  eta: 0:00:17    time: 0.3660  data: 0.0035  max mem: 57344
+eval (test): [20]  [60/85]  eta: 0:00:09    time: 0.3660  data: 0.0034  max mem: 57344
+eval (test): [20]  [80/85]  eta: 0:00:01    time: 0.3663  data: 0.0034  max mem: 57344
+eval (test): [20]  [84/85]  eta: 0:00:00    time: 0.3527  data: 0.0033  max mem: 57344
+eval (test): [20] Total time: 0:00:31 (0.3718 s / it)
+eval (testid): [20]  [ 0/82]  eta: 0:01:22    time: 1.0000  data: 0.6434  max mem: 57344
+eval (testid): [20]  [20/82]  eta: 0:00:24    time: 0.3680  data: 0.0031  max mem: 57344
+eval (testid): [20]  [40/82]  eta: 0:00:16    time: 0.3676  data: 0.0034  max mem: 57344
+eval (testid): [20]  [60/82]  eta: 0:00:08    time: 0.3679  data: 0.0034  max mem: 57344
+eval (testid): [20]  [80/82]  eta: 0:00:00    time: 0.3683  data: 0.0033  max mem: 57344
+eval (testid): [20]  [81/82]  eta: 0:00:00    time: 0.3512  data: 0.0033  max mem: 57344
+eval (testid): [20] Total time: 0:00:30 (0.3729 s / it)
+evaluating best checkpoint: experiments/schaefer1000/output/schaefer1000/schaefer1000_lr3e-4_3/eval_v2/nsd_cococlip__patch__attn/checkpoint-best.pth
+eval model info:
+{"score": 0.29549649317091176, "hparam": [1.2, 1.0], "hparam_id": 25, "epoch": 17, "is_best": true, "best_score": 0.29549649317091176}
+eval (train): [20]  [  0/509]  eta: 0:07:25    time: 0.8744  data: 0.5184  max mem: 57344
+eval (train): [20]  [ 20/509]  eta: 0:03:10    time: 0.3653  data: 0.0033  max mem: 57344
+eval (train): [20]  [ 40/509]  eta: 0:02:57    time: 0.3655  data: 0.0033  max mem: 57344
+eval (train): [20]  [ 60/509]  eta: 0:02:47    time: 0.3662  data: 0.0033  max mem: 57344
+eval (train): [20]  [ 80/509]  eta: 0:02:39    time: 0.3660  data: 0.0033  max mem: 57344
+eval (train): [20]  [100/509]  eta: 0:02:31    time: 0.3651  data: 0.0033  max mem: 57344
+eval (train): [20]  [120/509]  eta: 0:02:23    time: 0.3665  data: 0.0033  max mem: 57344
+eval (train): [20]  [140/509]  eta: 0:02:16    time: 0.3659  data: 0.0034  max mem: 57344
+eval (train): [20]  [160/509]  eta: 0:02:08    time: 0.3661  data: 0.0033  max mem: 57344
+eval (train): [20]  [180/509]  eta: 0:02:01    time: 0.3668  data: 0.0033  max mem: 57344
+eval (train): [20]  [200/509]  eta: 0:01:53    time: 0.3673  data: 0.0034  max mem: 57344
+eval (train): [20]  [220/509]  eta: 0:01:46    time: 0.3666  data: 0.0034  max mem: 57344
+eval (train): [20]  [240/509]  eta: 0:01:39    time: 0.3660  data: 0.0032  max mem: 57344
+eval (train): [20]  [260/509]  eta: 0:01:31    time: 0.3666  data: 0.0033  max mem: 57344
+eval (train): [20]  [280/509]  eta: 0:01:24    time: 0.3659  data: 0.0031  max mem: 57344
+eval (train): [20]  [300/509]  eta: 0:01:16    time: 0.3657  data: 0.0033  max mem: 57344
+eval (train): [20]  [320/509]  eta: 0:01:09    time: 0.3658  data: 0.0033  max mem: 57344
+eval (train): [20]  [340/509]  eta: 0:01:02    time: 0.3662  data: 0.0033  max mem: 57344
+eval (train): [20]  [360/509]  eta: 0:00:54    time: 0.3662  data: 0.0034  max mem: 57344
+eval (train): [20]  [380/509]  eta: 0:00:47    time: 0.3670  data: 0.0034  max mem: 57344
+eval (train): [20]  [400/509]  eta: 0:00:40    time: 0.3662  data: 0.0034  max mem: 57344
+eval (train): [20]  [420/509]  eta: 0:00:32    time: 0.3669  data: 0.0036  max mem: 57344
+eval (train): [20]  [440/509]  eta: 0:00:25    time: 0.3670  data: 0.0037  max mem: 57344
+eval (train): [20]  [460/509]  eta: 0:00:18    time: 0.3674  data: 0.0038  max mem: 57344
+eval (train): [20]  [480/509]  eta: 0:00:10    time: 0.3672  data: 0.0037  max mem: 57344
+eval (train): [20]  [500/509]  eta: 0:00:03    time: 0.3670  data: 0.0037  max mem: 57344
+eval (train): [20]  [508/509]  eta: 0:00:00    time: 0.3565  data: 0.0036  max mem: 57344
+eval (train): [20] Total time: 0:03:06 (0.3674 s / it)
+eval (validation): [20]  [ 0/85]  eta: 0:01:27    time: 1.0271  data: 0.6692  max mem: 57344
+eval (validation): [20]  [20/85]  eta: 0:00:25    time: 0.3661  data: 0.0028  max mem: 57344
+eval (validation): [20]  [40/85]  eta: 0:00:17    time: 0.3661  data: 0.0034  max mem: 57344
+eval (validation): [20]  [60/85]  eta: 0:00:09    time: 0.3659  data: 0.0035  max mem: 57344
+eval (validation): [20]  [80/85]  eta: 0:00:01    time: 0.3661  data: 0.0034  max mem: 57344
+eval (validation): [20]  [84/85]  eta: 0:00:00    time: 0.3596  data: 0.0034  max mem: 57344
+eval (validation): [20] Total time: 0:00:31 (0.3737 s / it)
+eval (test): [20]  [ 0/85]  eta: 0:01:25    time: 1.0070  data: 0.6487  max mem: 57344
+eval (test): [20]  [20/85]  eta: 0:00:25    time: 0.3653  data: 0.0031  max mem: 57344
+eval (test): [20]  [40/85]  eta: 0:00:17    time: 0.3656  data: 0.0035  max mem: 57344
+eval (test): [20]  [60/85]  eta: 0:00:09    time: 0.3657  data: 0.0035  max mem: 57344
+eval (test): [20]  [80/85]  eta: 0:00:01    time: 0.3669  data: 0.0034  max mem: 57344
+eval (test): [20]  [84/85]  eta: 0:00:00    time: 0.3530  data: 0.0034  max mem: 57344
+eval (test): [20] Total time: 0:00:31 (0.3716 s / it)
+eval (testid): [20]  [ 0/82]  eta: 0:01:27    time: 1.0696  data: 0.7103  max mem: 57344
+eval (testid): [20]  [20/82]  eta: 0:00:24    time: 0.3649  data: 0.0026  max mem: 57344
+eval (testid): [20]  [40/82]  eta: 0:00:16    time: 0.3657  data: 0.0035  max mem: 57344
+eval (testid): [20]  [60/82]  eta: 0:00:08    time: 0.3659  data: 0.0034  max mem: 57344
+eval (testid): [20]  [80/82]  eta: 0:00:00    time: 0.3664  data: 0.0034  max mem: 57344
+eval (testid): [20]  [81/82]  eta: 0:00:00    time: 0.3496  data: 0.0034  max mem: 57344
+eval (testid): [20] Total time: 0:00:30 (0.3717 s / it)
+eval results:
+
+| model            | repr   | clf   | dataset      | ckpt   |   epoch |      lr |   wd |   hparam_id | hparam     | split      |   loss |     acc |   acc_std |      f1 |    f1_std |
+|:-----------------|:-------|:------|:-------------|:-------|--------:|--------:|-----:|------------:|:-----------|:-----------|-------:|--------:|----------:|--------:|----------:|
+| schaefer1000_mae | patch  | attn  | nsd_cococlip | best   |      17 | 0.00036 | 0.05 |          25 | [1.2, 1.0] | train      | 2.0313 | 0.3852  | 0.0023882 | 0.33978 | 0.0025475 |
+| schaefer1000_mae | patch  | attn  | nsd_cococlip | best   |      17 | 0.00036 | 0.05 |          25 | [1.2, 1.0] | validation | 2.3556 | 0.2955  | 0.0056563 | 0.23706 | 0.0053114 |
+| schaefer1000_mae | patch  | attn  | nsd_cococlip | best   |      17 | 0.00036 | 0.05 |          25 | [1.2, 1.0] | test       | 2.2628 | 0.31076 | 0.005407  | 0.23729 | 0.0051753 |
+| schaefer1000_mae | patch  | attn  | nsd_cococlip | best   |      17 | 0.00036 | 0.05 |          25 | [1.2, 1.0] | testid     | 2.3136 | 0.29863 | 0.005567  | 0.25309 | 0.0054782 |
+
+
+done! total time: 1:43:23

schaefer1000/schaefer1000_lr3e-4_3/pretrain/config.yaml

@@ -0,0 +1,102 @@
+name: schaefer1000/schaefer1000_lr3e-4_3/pretrain
+notes: schaefer1000 ablation schaefer1000_lr3e-4_3 (input_space=schaefer1000 base_lr=3e-4
+  seed=5403)
+output_dir: experiments/schaefer1000/output/schaefer1000/schaefer1000_lr3e-4_3/pretrain
+input_space: schaefer1000
+patch_size: 1
+num_frames: 16
+t_patch_size: 4
+mask_ratio: 0.9
+pred_mask_ratio: null
+masking: tube
+masking_kwargs: {}
+mask_patch_size: null
+model: mae_vit_base
+model_kwargs:
+  decoding: attn
+  pos_embed: sep
+  target_norm: null
+  pca_norm_nc: 2
+  t_pred_stride: 2
+  no_decode_pos: true
+  mask_drop_scale: false
+  pred_edge_pad: 0
+  gauss_sigma: null
+  class_token: true
+  reg_tokens: 0
+  no_embed_class: true
+  head_init_scale: 0.0
+  decoder_depth: 4
+  drop_path_rate: 0.0
+datasets:
+  hcp-train:
+    type: wds
+    url: /data/fmri-datasets/pretrain/hcpya-all.${input_space}.wds/hcpya-all-${input_space}-{00000..01799}.tar
+    clipping: random
+    clipping_kwargs:
+      oversample: 4.0
+    shuffle: true
+    buffer_size: 2000
+    samples_per_epoch: 200000
+  hcp-train-subset:
+    type: arrow
+    root: s3://medarc/fmri-datasets/eval/hcpya-clips.${input_space}.arrow/validation
+    split_range:
+    - 0
+    - 2000
+    shuffle: false
+  hcp-val:
+    type: arrow
+    root: s3://medarc/fmri-datasets/eval/hcpya-clips.${input_space}.arrow/test
+    split_range:
+    - 0
+    - 2000
+    shuffle: false
+train_dataset: hcp-train
+eval_datasets:
+- hcp-train-subset
+- hcp-val
+val_dataset: null
+clip_vmax: 3.0
+normalize: frame
+tr_scale: null
+crop_scale: null
+crop_aspect: null
+gray_jitter: null
+num_workers: 16
+epochs: 100
+batch_size: 32
+accum_iter: 1
+base_lr: 0.0003
+min_lr: 0.0
+warmup_epochs: 5
+weight_decay: 0.05
+betas:
+- 0.9
+- 0.95
+clip_grad: 1.0
+amp: true
+amp_dtype: float16
+ckpt: null
+resume: true
+auto_resume: true
+start_epoch: 0
+max_checkpoints: 0
+checkpoint_period: null
+plot_period: 5
+device: cuda
+presend_cuda: false
+seed: 5403
+debug: false
+wandb: true
+wandb_entity: null
+wandb_project: fMRI-foundation-model
+rank: 0
+world_size: 1
+gpu: 0
+distributed: true
+dist_backend: nccl
+in_chans: 1
+img_size:
+- 1000
+- 1

schaefer1000/schaefer1000_lr3e-4_3/pretrain/log.json

@@ -0,0 +1,100 @@
+{"epoch": 0, "train/lr": 3.7507200230407366e-06, "train/grad": 1.5692052158799363, "train/loss": 0.9857268715190888, "eval/hcp-train-subset/loss": 0.9680695293411132, "eval/hcp-val/loss": 0.9636633713399211}
+{"epoch": 1, "train/lr": 1.125096003072098e-05, "train/grad": 3.2624280522856997, "train/loss": 0.9147129586124421, "eval/hcp-train-subset/loss": 0.8749773146644715, "eval/hcp-val/loss": 0.8708968422105235}
+{"epoch": 2, "train/lr": 1.875120003840123e-05, "train/grad": 3.0644060760145, "train/loss": 0.8059692593669892, "eval/hcp-train-subset/loss": 0.785597647390058, "eval/hcp-val/loss": 0.7760233946384922}
+{"epoch": 3, "train/lr": 2.625144004608147e-05, "train/grad": 1.9221768834545885, "train/loss": 0.7688120330429077, "eval/hcp-train-subset/loss": 0.7674491991919856, "eval/hcp-val/loss": 0.7573790357958886}
+{"epoch": 4, "train/lr": 3.375167986175557e-05, "train/grad": 1.2283830323733766, "train/loss": 0.755620023317337, "eval/hcp-train-subset/loss": 0.7563188585542864, "eval/hcp-val/loss": 0.7477518243174399}
+{"epoch": 5, "train/lr": 3.749658191365373e-05, "train/grad": 0.8344827419320997, "train/loss": 0.748303571805954, "eval/hcp-train-subset/loss": 0.7493925661809983, "eval/hcp-val/loss": 0.7394362803428404}
+{"epoch": 6, "train/lr": 3.74760811613377e-05, "train/grad": 0.6443673204654463, "train/loss": 0.738844185142517, "eval/hcp-train-subset/loss": 0.7415499187284901, "eval/hcp-val/loss": 0.7329423206467782}
+{"epoch": 7, "train/lr": 3.743510371420547e-05, "train/grad": 0.5348618692572792, "train/loss": 0.7309427853584289, "eval/hcp-train-subset/loss": 0.736884355545044, "eval/hcp-val/loss": 0.7288225466205228}
+{"epoch": 8, "train/lr": 3.73736943804934e-05, "train/grad": 0.4745288983533645, "train/loss": 0.7279760493946076, "eval/hcp-train-subset/loss": 0.7340370743505417, "eval/hcp-val/loss": 0.7255179016820846}
+{"epoch": 9, "train/lr": 3.7291920310406644e-05, "train/grad": 0.43001580518380117, "train/loss": 0.7250286063861847, "eval/hcp-train-subset/loss": 0.7317022758145486, "eval/hcp-val/loss": 0.7220333535825053}
+{"epoch": 10, "train/lr": 3.718987092269037e-05, "train/grad": 0.40384940853381474, "train/loss": 0.724578535528183, "eval/hcp-train-subset/loss": 0.7304689451571433, "eval/hcp-val/loss": 0.7199358296009802}
+{"epoch": 11, "train/lr": 3.706765780685143e-05, "train/grad": 0.3839408332383287, "train/loss": 0.7209072523593902, "eval/hcp-train-subset/loss": 0.7272679440436824, "eval/hcp-val/loss": 0.7185802132852616}
+{"epoch": 12, "train/lr": 3.692541460113792e-05, "train/grad": 0.372209113484747, "train/loss": 0.7164478220748901, "eval/hcp-train-subset/loss": 0.7253863330810301, "eval/hcp-val/loss": 0.7160299516493275}
+{"epoch": 13, "train/lr": 3.6763296846406675e-05, "train/grad": 0.35629202856509157, "train/loss": 0.7160235848426819, "eval/hcp-train-subset/loss": 0.7239290725800299, "eval/hcp-val/loss": 0.7153966349940146}
+{"epoch": 14, "train/lr": 3.658148181604263e-05, "train/grad": 0.35097664373053045, "train/loss": 0.7143508050060272, "eval/hcp-train-subset/loss": 0.7200255653550548, "eval/hcp-val/loss": 0.7132342323180167}
+{"epoch": 15, "train/lr": 3.6380168322111824e-05, "train/grad": 0.3453185911204923, "train/loss": 0.7097558912658691, "eval/hcp-train-subset/loss": 0.7195313611338215, "eval/hcp-val/loss": 0.7113439209999577}
+{"epoch": 16, "train/lr": 3.615957649796421e-05, "train/grad": 0.3388091952648662, "train/loss": 0.7071359890460968, "eval/hcp-train-subset/loss": 0.7193457551540867, "eval/hcp-val/loss": 0.7111214572383512}
+{"epoch": 17, "train/lr": 3.591994755752113e-05, "train/grad": 0.3409102491632507, "train/loss": 0.7057786175632477, "eval/hcp-train-subset/loss": 0.7179183959960938, "eval/hcp-val/loss": 0.710104807730644}
+{"epoch": 18, "train/lr": 3.5661543531510486e-05, "train/grad": 0.33757689253881645, "train/loss": 0.7050803663730622, "eval/hcp-train-subset/loss": 0.7165835999673412, "eval/hcp-val/loss": 0.7095545539932866}
+{"epoch": 19, "train/lr": 3.538464698094067e-05, "train/grad": 0.32954235030008694, "train/loss": 0.7049054228401184, "eval/hcp-train-subset/loss": 0.7162883945049778, "eval/hcp-val/loss": 0.7084918156746896}
+{"epoch": 20, "train/lr": 3.508956068812486e-05, "train/grad": 0.3295209528885785, "train/loss": 0.7023358335208892, "eval/hcp-train-subset/loss": 0.715675990427694, "eval/hcp-val/loss": 0.7078745576643175}
+{"epoch": 21, "train/lr": 3.4776607325591504e-05, "train/grad": 0.33164318751741145, "train/loss": 0.7019726961803436, "eval/hcp-train-subset/loss": 0.7165801169410828, "eval/hcp-val/loss": 0.7079531194702271}
+{"epoch": 22, "train/lr": 3.4446129103247903e-05, "train/grad": 0.3278170729854192, "train/loss": 0.699008160943985, "eval/hcp-train-subset/loss": 0.7123150652454745, "eval/hcp-val/loss": 0.7069503424629089}
+{"epoch": 23, "train/lr": 3.4098487394178203e-05, "train/grad": 0.32773768210317566, "train/loss": 0.6980064904880524, "eval/hcp-train-subset/loss": 0.7136735320091248, "eval/hcp-val/loss": 0.7059025331850974}
+{"epoch": 24, "train/lr": 3.37340623394871e-05, "train/grad": 0.3252800912504599, "train/loss": 0.6974921784496307, "eval/hcp-train-subset/loss": 0.7139810477533648, "eval/hcp-val/loss": 0.7065992605301642}
+{"epoch": 25, "train/lr": 3.335325243262167e-05, "train/grad": 0.32889571637320714, "train/loss": 0.6961316944789886, "eval/hcp-train-subset/loss": 0.7098356370002993, "eval/hcp-val/loss": 0.7052332326289146}
+{"epoch": 26, "train/lr": 3.295647408362393e-05, "train/grad": 0.33646529040534273, "train/loss": 0.6948735297203064, "eval/hcp-train-subset/loss": 0.7113901674747467, "eval/hcp-val/loss": 0.7039076397495885}
+{"epoch": 27, "train/lr": 3.254416116379294e-05, "train/grad": 0.3372331983912978, "train/loss": 0.6946172658729554, "eval/hcp-train-subset/loss": 0.7115432994980966, "eval/hcp-val/loss": 0.7053387597683938}
+{"epoch": 28, "train/lr": 3.211676453125335e-05, "train/grad": 0.33348862192087675, "train/loss": 0.6910485474967957, "eval/hcp-train-subset/loss": 0.7125448055805699, "eval/hcp-val/loss": 0.7052655537282267}
+{"epoch": 29, "train/lr": 3.1674751537947714e-05, "train/grad": 0.3396271452539078, "train/loss": 0.6884204471969605, "eval/hcp-train-subset/loss": 0.7094625061558139, "eval/hcp-val/loss": 0.7040185784139941}
+{"epoch": 30, "train/lr": 3.1218605518594614e-05, "train/grad": 0.3320936166287919, "train/loss": 0.6931153695678711, "eval/hcp-train-subset/loss": 0.7105431873952189, "eval/hcp-val/loss": 0.7031033029479365}
+{"epoch": 31, "train/lr": 3.074882526216893e-05, "train/grad": 0.3387619171840812, "train/loss": 0.6883212745285034, "eval/hcp-train-subset/loss": 0.7083456256697255, "eval/hcp-val/loss": 0.703125697951163}
+{"epoch": 32, "train/lr": 3.0265924466484304e-05, "train/grad": 0.33726599691809894, "train/loss": 0.6868814957427979, "eval/hcp-train-subset/loss": 0.7079083640729228, "eval/hcp-val/loss": 0.7026382752003208}
+{"epoch": 33, "train/lr": 2.977043117647054e-05, "train/grad": 0.33027436258755855, "train/loss": 0.6892728199386596, "eval/hcp-train-subset/loss": 0.7086907634812016, "eval/hcp-val/loss": 0.7031399963363525}
+{"epoch": 34, "train/lr": 2.9262887206765643e-05, "train/grad": 0.33681419492960546, "train/loss": 0.6870174325847626, "eval/hcp-train-subset/loss": 0.7071865816270152, "eval/hcp-val/loss": 0.7012880867527377}
+{"epoch": 35, "train/lr": 2.8743847549248834e-05, "train/grad": 0.33963567170215947, "train/loss": 0.6870327368736268, "eval/hcp-train-subset/loss": 0.7060080708995942, "eval/hcp-val/loss": 0.7015848342449434}
+{"epoch": 36, "train/lr": 2.821387976616516e-05, "train/grad": 0.3451497349354369, "train/loss": 0.6835688859939575, "eval/hcp-train-subset/loss": 0.7078843347487911, "eval/hcp-val/loss": 0.7030668220212383}
+{"epoch": 37, "train/lr": 2.7673563369504416e-05, "train/grad": 0.345504092610657, "train/loss": 0.6834061105632782, "eval/hcp-train-subset/loss": 0.7079177383453615, "eval/hcp-val/loss": 0.7027584593142232}
+{"epoch": 38, "train/lr": 2.7123489187313515e-05, "train/grad": 0.347737196158443, "train/loss": 0.6822440135383606, "eval/hcp-train-subset/loss": 0.704343436225768, "eval/hcp-val/loss": 0.7014743454994694}
+{"epoch": 39, "train/lr": 2.6564258717634662e-05, "train/grad": 0.3562175120632992, "train/loss": 0.6822932584714889, "eval/hcp-train-subset/loss": 0.704769492149353, "eval/hcp-val/loss": 0.7003187767920955}
+{"epoch": 40, "train/lr": 2.5996483470775956e-05, "train/grad": 0.3543595119115733, "train/loss": 0.6797226435470581, "eval/hcp-train-subset/loss": 0.7038520555342397, "eval/hcp-val/loss": 0.7007266590672154}
+{"epoch": 41, "train/lr": 2.5420784300633598e-05, "train/grad": 0.3493961739689923, "train/loss": 0.6842184673976898, "eval/hcp-train-subset/loss": 0.705357369876677, "eval/hcp-val/loss": 0.7005555802775968}
+{"epoch": 42, "train/lr": 2.4837790725798037e-05, "train/grad": 0.35674744963684096, "train/loss": 0.6803887548923493, "eval/hcp-train-subset/loss": 0.7030450086439809, "eval/hcp-val/loss": 0.7011541743432322}
+{"epoch": 43, "train/lr": 2.4248140241183872e-05, "train/grad": 0.350894656475399, "train/loss": 0.6795104444217682, "eval/hcp-train-subset/loss": 0.7044258886767972, "eval/hcp-val/loss": 0.6986707160549779}
+{"epoch": 44, "train/lr": 2.36524776209388e-05, "train/grad": 0.35485822762635033, "train/loss": 0.6812120409297943, "eval/hcp-train-subset/loss": 0.7049746715253399, "eval/hcp-val/loss": 0.7004548061278558}
+{"epoch": 45, "train/lr": 2.305145421339275e-05, "train/grad": 0.3632956369840681, "train/loss": 0.6785168221759796, "eval/hcp-train-subset/loss": 0.7058336196407196, "eval/hcp-val/loss": 0.7006226681893871}
+{"epoch": 46, "train/lr": 2.2445727228817648e-05, "train/grad": 0.36778920585725217, "train/loss": 0.6762780577850341, "eval/hcp-train-subset/loss": 0.7003843380558875, "eval/hcp-val/loss": 0.6997875869274139}
+{"epoch": 47, "train/lr": 2.1835959020777867e-05, "train/grad": 0.370750277146771, "train/loss": 0.6775857842445373, "eval/hcp-train-subset/loss": 0.7001353656091998, "eval/hcp-val/loss": 0.7010902425935192}
+{"epoch": 48, "train/lr": 2.1222816361856524e-05, "train/grad": 0.3719557352726576, "train/loss": 0.6761284592247009, "eval/hcp-train-subset/loss": 0.7016976341124503, "eval/hcp-val/loss": 0.7010709201135943}
+{"epoch": 49, "train/lr": 2.0606969714549076e-05, "train/grad": 0.37747293027072826, "train/loss": 0.6742544750213623, "eval/hcp-train-subset/loss": 0.6990556101645192, "eval/hcp-val/loss": 0.6986670667125333}
+{"epoch": 50, "train/lr": 1.9989092498122458e-05, "train/grad": 0.37805133970457017, "train/loss": 0.6741572340393066, "eval/hcp-train-subset/loss": 0.6982421346249119, "eval/hcp-val/loss": 0.698340343852197}
+{"epoch": 51, "train/lr": 1.936986035224125e-05, "train/grad": 0.3818382188374851, "train/loss": 0.6773040638351441, "eval/hcp-train-subset/loss": 0.6961049533659412, "eval/hcp-val/loss": 0.6982582922904722}
+{"epoch": 52, "train/lr": 1.8749950398165728e-05, "train/grad": 0.38511339629318536, "train/loss": 0.6715853364944459, "eval/hcp-train-subset/loss": 0.6965270542329357, "eval/hcp-val/loss": 0.6976662528130316}
+{"epoch": 53, "train/lr": 1.8130040498328958e-05, "train/grad": 0.3813028623194471, "train/loss": 0.6754535886859894, "eval/hcp-train-subset/loss": 0.6994368770430165, "eval/hcp-val/loss": 0.6981539533984277}
+{"epoch": 54, "train/lr": 1.751080851510517e-05, "train/grad": 0.3959278035577973, "train/loss": 0.6696969284248352, "eval/hcp-train-subset/loss": 0.6987666554989354, "eval/hcp-val/loss": 0.697522993049314}
+{"epoch": 55, "train/lr": 1.689293156957646e-05, "train/grad": 0.4035230275033645, "train/loss": 0.6682268149280548, "eval/hcp-train-subset/loss": 0.6954169629081604, "eval/hcp-val/loss": 0.6984351056237375}
+{"epoch": 56, "train/lr": 1.627708530111125e-05, "train/grad": 0.3976198381469863, "train/loss": 0.6730791558361053, "eval/hcp-train-subset/loss": 0.6981263420274181, "eval/hcp-val/loss": 0.6979274201777673}
+{"epoch": 57, "train/lr": 1.566394312856223e-05, "train/grad": 0.40140138883265913, "train/loss": 0.6742932442378998, "eval/hcp-train-subset/loss": 0.6983444056203288, "eval/hcp-val/loss": 0.6966883126766451}
+{"epoch": 58, "train/lr": 1.5054175513893184e-05, "train/grad": 0.4068282823076396, "train/loss": 0.6693542087936402, "eval/hcp-train-subset/loss": 0.6953647396256847, "eval/hcp-val/loss": 0.6957321243901406}
+{"epoch": 59, "train/lr": 1.444844922903809e-05, "train/grad": 0.4022570568543505, "train/loss": 0.6700057065296173, "eval/hcp-train-subset/loss": 0.6923402547836304, "eval/hcp-val/loss": 0.6945483799903623}
+{"epoch": 60, "train/lr": 1.3847426626796371e-05, "train/grad": 0.40555765827767143, "train/loss": 0.6712343048858642, "eval/hcp-train-subset/loss": 0.6917753959855726, "eval/hcp-val/loss": 0.6944833320956076}
+{"epoch": 61, "train/lr": 1.3251764916559323e-05, "train/grad": 0.4093855313465438, "train/loss": 0.6733830636310577, "eval/hcp-train-subset/loss": 0.6927937846029958, "eval/hcp-val/loss": 0.6940909941350261}
+{"epoch": 62, "train/lr": 1.2662115445661782e-05, "train/grad": 0.41183811696615935, "train/loss": 0.6671849621677398, "eval/hcp-train-subset/loss": 0.6923541993864121, "eval/hcp-val/loss": 0.6943240656006721}
+{"epoch": 63, "train/lr": 1.2079122987143025e-05, "train/grad": 0.4270764069379149, "train/loss": 0.6664408423709869, "eval/hcp-train-subset/loss": 0.6924344762679069, "eval/hcp-val/loss": 0.6955072485631512}
+{"epoch": 64, "train/lr": 1.1503425034696866e-05, "train/grad": 0.41739537209935507, "train/loss": 0.6714184842109681, "eval/hcp-train-subset/loss": 0.6913651574042535, "eval/hcp-val/loss": 0.6949986988498319}
+{"epoch": 65, "train/lr": 1.0935651105582284e-05, "train/grad": 0.42379808755370235, "train/loss": 0.6704631824874878, "eval/hcp-train-subset/loss": 0.6928060362415929, "eval/hcp-val/loss": 0.6965652531193148}
+{"epoch": 66, "train/lr": 1.0376422052254546e-05, "train/grad": 0.42755572111482965, "train/loss": 0.6680676095962524, "eval/hcp-train-subset/loss": 0.6911212090523012, "eval/hcp-val/loss": 0.6941236940122419}
+{"epoch": 67, "train/lr": 9.82634938347307e-06, "train/grad": 0.4346660729888269, "train/loss": 0.6702318864536285, "eval/hcp-train-subset/loss": 0.6882516184160786, "eval/hcp-val/loss": 0.6941954445454382}
+{"epoch": 68, "train/lr": 9.286034595625e-06, "train/grad": 0.43992226756781183, "train/loss": 0.6671486667919159, "eval/hcp-train-subset/loss": 0.6863247896394422, "eval/hcp-val/loss": 0.6946903159541469}
+{"epoch": 69, "train/lr": 8.756068514998114e-06, "train/grad": 0.44327839248826806, "train/loss": 0.6690675859069825, "eval/hcp-train-subset/loss": 0.6848775138778072, "eval/hcp-val/loss": 0.6923904861173322}
+{"epoch": 70, "train/lr": 8.237030651720665e-06, "train/grad": 0.4451539502043883, "train/loss": 0.6680399763393402, "eval/hcp-train-subset/loss": 0.6872604547008392, "eval/hcp-val/loss": 0.6924347185319469}
+{"epoch": 71, "train/lr": 7.729488566075681e-06, "train/grad": 0.4417733099197727, "train/loss": 0.6695161906051635, "eval/hcp-train-subset/loss": 0.6864452833129514, "eval/hcp-val/loss": 0.6935344761417758}
+{"epoch": 72, "train/lr": 7.233997247882178e-06, "train/grad": 0.44688647258770303, "train/loss": 0.6668959783267975, "eval/hcp-train-subset/loss": 0.6863080580388347, "eval/hcp-val/loss": 0.6929781956057395}
+{"epoch": 73, "train/lr": 6.751098509621876e-06, "train/grad": 0.47144054506980504, "train/loss": 0.6640548050117493, "eval/hcp-train-subset/loss": 0.6866888269301383, "eval/hcp-val/loss": 0.6919965974746212}
+{"epoch": 74, "train/lr": 6.281320393975191e-06, "train/grad": 0.4646602188083182, "train/loss": 0.6658676354789734, "eval/hcp-train-subset/loss": 0.6861167271291057, "eval/hcp-val/loss": 0.692183643579483}
+{"epoch": 75, "train/lr": 5.825176596414363e-06, "train/grad": 0.46845923881540685, "train/loss": 0.6647836077785492, "eval/hcp-train-subset/loss": 0.6877285059421293, "eval/hcp-val/loss": 0.6934423465882579}
+{"epoch": 76, "train/lr": 5.3831659034850535e-06, "train/grad": 0.4706737583267017, "train/loss": 0.6649509129428863, "eval/hcp-train-subset/loss": 0.6851985175763408, "eval/hcp-val/loss": 0.6909103556986778}
+{"epoch": 77, "train/lr": 4.955771647390373e-06, "train/grad": 0.4696506527318081, "train/loss": 0.6642163661766052, "eval/hcp-train-subset/loss": 0.6853945582143722, "eval/hcp-val/loss": 0.6910395160798104}
+{"epoch": 78, "train/lr": 4.543461177474475e-06, "train/grad": 0.4770270008379152, "train/loss": 0.6670944433307647, "eval/hcp-train-subset/loss": 0.6839212294547788, "eval/hcp-val/loss": 0.6916454040235088}
+{"epoch": 79, "train/lr": 4.146685349182691e-06, "train/grad": 0.48552370255111005, "train/loss": 0.6683693843317032, "eval/hcp-train-subset/loss": 0.682494705723178, "eval/hcp-val/loss": 0.690316621334322}
+{"epoch": 80, "train/lr": 3.7658780310578768e-06, "train/grad": 0.4754748591510253, "train/loss": 0.6646157481384277, "eval/hcp-train-subset/loss": 0.6806786983243881, "eval/hcp-val/loss": 0.6909599035016952}
+{"epoch": 81, "train/lr": 3.401455630311446e-06, "train/grad": 0.47882948339949116, "train/loss": 0.6670280052375793, "eval/hcp-train-subset/loss": 0.6805518692539584, "eval/hcp-val/loss": 0.6899988401320672}
+{"epoch": 82, "train/lr": 3.0538166374881993e-06, "train/grad": 0.4848450655534997, "train/loss": 0.6650464585208893, "eval/hcp-train-subset/loss": 0.6791223652901188, "eval/hcp-val/loss": 0.6902142965024517}
+{"epoch": 83, "train/lr": 2.723341190722833e-06, "train/grad": 0.48015087908288884, "train/loss": 0.6668450689315796, "eval/hcp-train-subset/loss": 0.6787165384138784, "eval/hcp-val/loss": 0.6892832344578158}
+{"epoch": 84, "train/lr": 2.410390660064364e-06, "train/grad": 0.48152043543536593, "train/loss": 0.6698820513057708, "eval/hcp-train-subset/loss": 0.6775700728739461, "eval/hcp-val/loss": 0.6889576854244355}
+{"epoch": 85, "train/lr": 2.115307252323394e-06, "train/grad": 0.4917814660837663, "train/loss": 0.6665777829837799, "eval/hcp-train-subset/loss": 0.6767233081402317, "eval/hcp-val/loss": 0.6879915806554979}
+{"epoch": 86, "train/lr": 1.838413636873912e-06, "train/grad": 0.5026889388433246, "train/loss": 0.6676180237197876, "eval/hcp-train-subset/loss": 0.6763049796704323, "eval/hcp-val/loss": 0.6889908438728701}
+{"epoch": 87, "train/lr": 1.5800125928190943e-06, "train/grad": 0.4847615266229814, "train/loss": 0.6671797801876068, "eval/hcp-train-subset/loss": 0.677047521837296, "eval/hcp-val/loss": 0.6886952029120538}
+{"epoch": 88, "train/lr": 1.3403866779068222e-06, "train/grad": 0.49768354199473747, "train/loss": 0.6721156395530701, "eval/hcp-train-subset/loss": 0.6766369054394383, "eval/hcp-val/loss": 0.6881197239122083}
+{"epoch": 89, "train/lr": 1.1197979195568888e-06, "train/grad": 0.4945417618501543, "train/loss": 0.6685085526323319, "eval/hcp-train-subset/loss": 0.6764899867196237, "eval/hcp-val/loss": 0.6885435167820223}
+{"epoch": 90, "train/lr": 9.184875283379039e-07, "train/grad": 0.5027800506784577, "train/loss": 0.6669693539142608, "eval/hcp-train-subset/loss": 0.6754624170641745, "eval/hcp-val/loss": 0.6876922903522369}
+{"epoch": 91, "train/lr": 7.366756342070463e-07, "train/grad": 0.5098568844818127, "train/loss": 0.6693566687202454, "eval/hcp-train-subset/loss": 0.6749280220077883, "eval/hcp-val/loss": 0.6883444430366639}
+{"epoch": 92, "train/lr": 5.745610458012273e-07, "train/grad": 0.5214255051961629, "train/loss": 0.6692327187252045, "eval/hcp-train-subset/loss": 0.6751869811165717, "eval/hcp-val/loss": 0.6869627112342466}
+{"epoch": 93, "train/lr": 4.323210330427178e-07, "train/grad": 0.5138676824838004, "train/loss": 0.6720273920631409, "eval/hcp-train-subset/loss": 0.6749346756166027, "eval/hcp-val/loss": 0.6869472342152749}
+{"epoch": 94, "train/lr": 3.1011113329712343e-07, "train/grad": 0.5229863038432049, "train/loss": 0.6686516582107543, "eval/hcp-train-subset/loss": 0.6732702812840862, "eval/hcp-val/loss": 0.6865402921553581}
+{"epoch": 95, "train/lr": 2.080649812955481e-07, "train/grad": 0.5085970294678818, "train/loss": 0.6721922983932496, "eval/hcp-train-subset/loss": 0.6735593397771159, "eval/hcp-val/loss": 0.6865103177485927}
+{"epoch": 96, "train/lr": 1.2629416300698208e-07, "train/grad": 0.5223715941966648, "train/loss": 0.6697415300273896, "eval/hcp-train-subset/loss": 0.6737183505488981, "eval/hcp-val/loss": 0.6871635154370339}
+{"epoch": 97, "train/lr": 6.488809362067338e-08, "train/grad": 0.5258212484390726, "train/loss": 0.6728149759101868, "eval/hcp-train-subset/loss": 0.6728160246726005, "eval/hcp-val/loss": 0.6865606481029142}
+{"epoch": 98, "train/lr": 2.391391977194211e-08, "train/grad": 0.5104606752110245, "train/loss": 0.6714796848392487, "eval/hcp-train-subset/loss": 0.6732331004834944, "eval/hcp-val/loss": 0.687980329798114}
+{"epoch": 99, "train/lr": 3.4164461183156008e-09, "train/grad": 0.5246001164760898, "train/loss": 0.67414359957695, "eval/hcp-train-subset/loss": 0.6729534153015383, "eval/hcp-val/loss": 0.6866767012303875}

schaefer1000/schaefer1000_lr3e-4_4/eval_v2/nsd_cococlip__patch__attn/config.yaml

@@ -0,0 +1,96 @@
+output_root: experiments/schaefer1000/output
+name_prefix: eval_probe
+remote_root: null
+notes: schaefer1000 ablation schaefer1000_lr3e-4_4; eval v2 (nsd_cococlip patch attn)
+model_kwargs:
+  ckpt_path: experiments/schaefer1000/output/schaefer1000/schaefer1000_lr3e-4_4/pretrain/checkpoint-last.pth
+dataset_kwargs: {}
+classifier_kwargs:
+  embed_dim: null
+  dropout: 0.0
+  xavier_init: true
+  norm: true
+lr_scale_grid:
+- 0.02
+- 0.023
+- 0.028
+- 0.033
+- 0.038
+- 0.045
+- 0.053
+- 0.062
+- 0.074
+- 0.087
+- 0.1
+- 0.12
+- 0.14
+- 0.17
+- 0.2
+- 0.23
+- 0.27
+- 0.32
+- 0.38
+- 0.44
+- 0.52
+- 0.61
+- 0.72
+- 0.85
+- 1
+- 1.2
+- 1.4
+- 1.6
+- 1.9
+- 2.3
+- 2.7
+- 3.1
+- 3.7
+- 4.3
+- 5.1
+- 6
+- 7.1
+- 8.3
+- 9.8
+- 12
+- 14
+- 16
+- 19
+- 22
+- 26
+- 31
+- 36
+- 43
+- 50
+wd_scale_grid:
+- 1.0
+num_workers: 8
+prefetch_factor: null
+balanced_sampling: false
+epochs: 20
+steps_per_epoch: 200
+batch_size: 64
+accum_iter: 2
+lr: 0.0003
+warmup_epochs: 5
+no_decay: false
+weight_decay: 0.05
+clip_grad: 1.0
+metrics:
+- acc
+- f1
+cv_metric: acc
+early_stopping: true
+amp: true
+device: cuda
+seed: 4466
+debug: false
+wandb: false
+wandb_entity: null
+wandb_project: fMRI-fm-eval
+name: schaefer1000/schaefer1000_lr3e-4_4/eval_v2/nsd_cococlip__patch__attn
+model: schaefer1000_mae
+representation: patch
+classifier: attn
+dataset: nsd_cococlip
+distributed: false
+output_dir: experiments/schaefer1000/output/schaefer1000/schaefer1000_lr3e-4_4/eval_v2/nsd_cococlip__patch__attn
+remote_dir: null

schaefer1000/schaefer1000_lr3e-4_4/eval_v2/nsd_cococlip__patch__attn/eval_log.json

@@ -0,0 +1 @@
+{"eval/epoch": 15, "eval/id_best": 25, "eval/lr_best": 0.00035999999999999997, "eval/wd_best": 0.05, "eval/train/loss": 1.9453297853469849, "eval/train/acc": 0.4113832631611297, "eval/train/acc_std": 0.0024169233639588553, "eval/train/f1": 0.3681901173082503, "eval/train/f1_std": 0.0026382665994097027, "eval/validation/loss": 2.3597986698150635, "eval/validation/acc": 0.29088224437061644, "eval/validation/acc_std": 0.00554328802582658, "eval/validation/f1": 0.242732353251757, "eval/validation/f1_std": 0.005606659484929468, "eval/test/loss": 2.276444435119629, "eval/test/acc": 0.30500927643784786, "eval/test/acc_std": 0.005442291980285984, "eval/test/f1": 0.24335883215883589, "eval/test/f1_std": 0.005426956635984132, "eval/testid/loss": 2.277097702026367, "eval/testid/acc": 0.3001735106998265, "eval/testid/acc_std": 0.005744578391488162, "eval/testid/f1": 0.25679578695480937, "eval/testid/f1_std": 0.005662045643834307}

schaefer1000/schaefer1000_lr3e-4_4/eval_v2/nsd_cococlip__patch__attn/eval_log_best.json

@@ -0,0 +1 @@
+{"eval/best/epoch": 15, "eval/best/id_best": 25, "eval/best/lr_best": 0.00035999999999999997, "eval/best/wd_best": 0.05, "eval/best/train/loss": 1.9453297853469849, "eval/best/train/acc": 0.4113832631611297, "eval/best/train/acc_std": 0.0024169233639588553, "eval/best/train/f1": 0.3681901173082503, "eval/best/train/f1_std": 0.0026382665994097027, "eval/best/validation/loss": 2.3597986698150635, "eval/best/validation/acc": 0.29088224437061644, "eval/best/validation/acc_std": 0.00554328802582658, "eval/best/validation/f1": 0.242732353251757, "eval/best/validation/f1_std": 0.005606659484929468, "eval/best/test/loss": 2.276444435119629, "eval/best/test/acc": 0.30500927643784786, "eval/best/test/acc_std": 0.005442291980285984, "eval/best/test/f1": 0.24335883215883589, "eval/best/test/f1_std": 0.005426956635984132, "eval/best/testid/loss": 2.277097702026367, "eval/best/testid/acc": 0.3001735106998265, "eval/best/testid/acc_std": 0.005744578391488162, "eval/best/testid/f1": 0.25679578695480937, "eval/best/testid/f1_std": 0.005662045643834307}

schaefer1000/schaefer1000_lr3e-4_4/eval_v2/nsd_cococlip__patch__attn/eval_log_last.json

@@ -0,0 +1 @@
+{"eval/last/epoch": 19, "eval/last/id_best": 26, "eval/last/lr_best": 0.00041999999999999996, "eval/last/wd_best": 0.05, "eval/last/train/loss": 1.890472650527954, "eval/last/train/acc": 0.4276406773410369, "eval/last/train/acc_std": 0.0025199914121819753, "eval/last/train/f1": 0.3863264107776285, "eval/last/train/f1_std": 0.0027383497081882495, "eval/last/validation/loss": 2.388059139251709, "eval/last/validation/acc": 0.2888519748984865, "eval/last/validation/acc_std": 0.005405298560493853, "eval/last/validation/f1": 0.23857168930622308, "eval/last/validation/f1_std": 0.005474306214480714, "eval/last/test/loss": 2.280224084854126, "eval/last/test/acc": 0.3079777365491651, "eval/last/test/acc_std": 0.005351248091270318, "eval/last/test/f1": 0.24533979172401646, "eval/last/test/f1_std": 0.0052202269246829905, "eval/last/testid/loss": 2.275458335876465, "eval/last/testid/acc": 0.30364372469635625, "eval/last/testid/acc_std": 0.005625782688142505, "eval/last/testid/f1": 0.25852394802470546, "eval/last/testid/f1_std": 0.005644025399682938}

schaefer1000/schaefer1000_lr3e-4_4/eval_v2/nsd_cococlip__patch__attn/eval_table.csv

@@ -0,0 +1,5 @@
+model,repr,clf,dataset,ckpt,epoch,lr,wd,hparam_id,hparam,split,loss,acc,acc_std,f1,f1_std
+schaefer1000_mae,patch,attn,nsd_cococlip,best,15,0.00035999999999999997,0.05,25,"[1.2, 1.0]",train,1.9453297853469849,0.4113832631611297,0.0024169233639588553,0.3681901173082503,0.0026382665994097027
+schaefer1000_mae,patch,attn,nsd_cococlip,best,15,0.00035999999999999997,0.05,25,"[1.2, 1.0]",validation,2.3597986698150635,0.29088224437061644,0.00554328802582658,0.242732353251757,0.005606659484929468
+schaefer1000_mae,patch,attn,nsd_cococlip,best,15,0.00035999999999999997,0.05,25,"[1.2, 1.0]",test,2.276444435119629,0.30500927643784786,0.005442291980285984,0.24335883215883589,0.005426956635984132
+schaefer1000_mae,patch,attn,nsd_cococlip,best,15,0.00035999999999999997,0.05,25,"[1.2, 1.0]",testid,2.277097702026367,0.3001735106998265,0.005744578391488162,0.25679578695480937,0.005662045643834307

schaefer1000/schaefer1000_lr3e-4_4/eval_v2/nsd_cococlip__patch__attn/eval_table_best.csv

@@ -0,0 +1,5 @@
+model,repr,clf,dataset,ckpt,epoch,lr,wd,hparam_id,hparam,split,loss,acc,acc_std,f1,f1_std
+schaefer1000_mae,patch,attn,nsd_cococlip,best,15,0.00035999999999999997,0.05,25,"[1.2, 1.0]",train,1.9453297853469849,0.4113832631611297,0.0024169233639588553,0.3681901173082503,0.0026382665994097027
+schaefer1000_mae,patch,attn,nsd_cococlip,best,15,0.00035999999999999997,0.05,25,"[1.2, 1.0]",validation,2.3597986698150635,0.29088224437061644,0.00554328802582658,0.242732353251757,0.005606659484929468
+schaefer1000_mae,patch,attn,nsd_cococlip,best,15,0.00035999999999999997,0.05,25,"[1.2, 1.0]",test,2.276444435119629,0.30500927643784786,0.005442291980285984,0.24335883215883589,0.005426956635984132
+schaefer1000_mae,patch,attn,nsd_cococlip,best,15,0.00035999999999999997,0.05,25,"[1.2, 1.0]",testid,2.277097702026367,0.3001735106998265,0.005744578391488162,0.25679578695480937,0.005662045643834307

schaefer1000/schaefer1000_lr3e-4_4/eval_v2/nsd_cococlip__patch__attn/eval_table_last.csv

@@ -0,0 +1,5 @@
+model,repr,clf,dataset,ckpt,epoch,lr,wd,hparam_id,hparam,split,loss,acc,acc_std,f1,f1_std
+schaefer1000_mae,patch,attn,nsd_cococlip,last,19,0.00041999999999999996,0.05,26,"[1.4, 1.0]",train,1.890472650527954,0.4276406773410369,0.0025199914121819753,0.3863264107776285,0.0027383497081882495
+schaefer1000_mae,patch,attn,nsd_cococlip,last,19,0.00041999999999999996,0.05,26,"[1.4, 1.0]",validation,2.388059139251709,0.2888519748984865,0.005405298560493853,0.23857168930622308,0.005474306214480714
+schaefer1000_mae,patch,attn,nsd_cococlip,last,19,0.00041999999999999996,0.05,26,"[1.4, 1.0]",test,2.280224084854126,0.3079777365491651,0.005351248091270318,0.24533979172401646,0.0052202269246829905
+schaefer1000_mae,patch,attn,nsd_cococlip,last,19,0.00041999999999999996,0.05,26,"[1.4, 1.0]",testid,2.275458335876465,0.30364372469635625,0.005625782688142505,0.25852394802470546,0.005644025399682938

schaefer1000/schaefer1000_lr3e-4_4/eval_v2/nsd_cococlip__patch__attn/log.txt

@@ -0,0 +1,969 @@
+fMRI foundation model probe eval
+version: 0.1.dev95+g65be98d36
+sha: 87e31aaa465443ed5f0da58176ac8395447cdbd0, status: has uncommitted changes, branch: dev/clane9
+cwd: /data/connor/fmri-fm
+start: 2026-05-12 20:54:49
+config:
+output_root: experiments/schaefer1000/output
+name_prefix: eval_probe
+remote_root: null
+notes: schaefer1000 ablation schaefer1000_lr3e-4_4; eval v2 (nsd_cococlip patch attn)
+model_kwargs:
+  ckpt_path: experiments/schaefer1000/output/schaefer1000/schaefer1000_lr3e-4_4/pretrain/checkpoint-last.pth
+dataset_kwargs: {}
+classifier_kwargs:
+  embed_dim: null
+  dropout: 0.0
+  xavier_init: true
+  norm: true
+lr_scale_grid:
+- 0.02
+- 0.023
+- 0.028
+- 0.033
+- 0.038
+- 0.045
+- 0.053
+- 0.062
+- 0.074
+- 0.087
+- 0.1
+- 0.12
+- 0.14
+- 0.17
+- 0.2
+- 0.23
+- 0.27
+- 0.32
+- 0.38
+- 0.44
+- 0.52
+- 0.61
+- 0.72
+- 0.85
+- 1
+- 1.2
+- 1.4
+- 1.6
+- 1.9
+- 2.3
+- 2.7
+- 3.1
+- 3.7
+- 4.3
+- 5.1
+- 6
+- 7.1
+- 8.3
+- 9.8
+- 12
+- 14
+- 16
+- 19
+- 22
+- 26
+- 31
+- 36
+- 43
+- 50
+wd_scale_grid:
+- 1.0
+num_workers: 8
+prefetch_factor: null
+balanced_sampling: false
+epochs: 20
+steps_per_epoch: 200
+batch_size: 64
+accum_iter: 2
+lr: 0.0003
+warmup_epochs: 5
+no_decay: false
+weight_decay: 0.05
+clip_grad: 1.0
+metrics:
+- acc
+- f1
+cv_metric: acc
+early_stopping: true
+amp: true
+device: cuda
+seed: 4466
+debug: false
+wandb: false
+wandb_entity: null
+wandb_project: fMRI-fm-eval
+name: schaefer1000/schaefer1000_lr3e-4_4/eval_v2/nsd_cococlip__patch__attn
+model: schaefer1000_mae
+representation: patch
+classifier: attn
+dataset: nsd_cococlip
+distributed: false
+output_dir: experiments/schaefer1000/output/schaefer1000/schaefer1000_lr3e-4_4/eval_v2/nsd_cococlip__patch__attn
+remote_dir: null
+
+creating frozen backbone model: schaefer1000_mae
+backbone:
+MaskedEncoderWrapper(
+  (model): MaskedEncoder(
+    class_token=True, reg_tokens=0, no_embed_class=True, mask_drop_scale=False
+    (patchify): Patchify3D((16, 1000, 1), (4, 1, 1), in_chans=1)
+    (patch_embed): Linear(in_features=4, out_features=768, bias=True)
+    (pos_embed): SeparablePosEmbed(768, (4, 1000, 1))
+    (blocks): ModuleList(
+      (0-11): 12 x Block(
+        (norm1): LayerNorm((768,), eps=1e-06, elementwise_affine=True)
+        (attn): Attention(
+          num_heads=12
+          (q): Linear(in_features=768, out_features=768, bias=True)
+          (k): Linear(in_features=768, out_features=768, bias=True)
+          (v): Linear(in_features=768, out_features=768, bias=True)
+          (proj): Linear(in_features=768, out_features=768, bias=True)
+        )
+        (drop_path1): Identity()
+        (norm2): LayerNorm((768,), eps=1e-06, elementwise_affine=True)
+        (mlp): Mlp(
+          (fc1): Linear(in_features=768, out_features=3072, bias=True)
+          (act): GELU(approximate='none')
+          (fc2): Linear(in_features=3072, out_features=768, bias=True)
+        )
+        (drop_path2): Identity()
+      )
+    )
+    (norm): LayerNorm((768,), eps=1e-06, elementwise_affine=True)
+  )
+)
+creating dataset: nsd_cococlip (schaefer1000)
+train (n=32539):
+HFDataset(
+    dataset=Dataset({
+    features: ['sub', 'ses', 'run', 'trial_id', 'nsd_id', 'category_id', 'path', 'start', 'end', 'n_frames', 'tr', 'bold', 'mean', 'std'],
+    num_rows: 32539
+}),
+    labels=[ 3  4  5  6 10 11 17 18 19 20 22 23 25 30 31 33 36 37 38 53 55 59 61 74],
+    counts=[1286 1180 1639 1868  834  824 1026 1042  913 1853 1503 2092 1001 1410
+  794 1241 1904 1872 2267 1428  889  904 1447 1322]
+)
+
+validation (n=5418):
+HFDataset(
+    dataset=Dataset({
+    features: ['sub', 'ses', 'run', 'trial_id', 'nsd_id', 'category_id', 'path', 'start', 'end', 'n_frames', 'tr', 'bold', 'mean', 'std'],
+    num_rows: 5418
+}),
+    labels=[ 3  4  5  6 10 11 17 18 19 20 22 23 25 30 31 33 36 37 38 53 55 59 61 74],
+    counts=[197 161 276 345 126 142 143 185 112 295 285 387 169 250 159 193 316 334
+ 343 215 172 141 226 246]
+)
+
+test (n=5390):
+HFDataset(
+    dataset=Dataset({
+    features: ['sub', 'ses', 'run', 'trial_id', 'nsd_id', 'category_id', 'path', 'start', 'end', 'n_frames', 'tr', 'bold', 'mean', 'std'],
+    num_rows: 5390
+}),
+    labels=[ 3  4  5  6 10 11 17 18 19 20 22 23 25 30 31 33 36 37 38 53 55 59 61 74],
+    counts=[202 172 274 298 144 180 134 182 186 293 218 343 165 185 140 177 346 333
+ 345 271 165 140 251 246]
+)
+
+testid (n=5187):
+HFDataset(
+    dataset=Dataset({
+    features: ['sub', 'ses', 'run', 'trial_id', 'nsd_id', 'category_id', 'path', 'start', 'end', 'n_frames', 'tr', 'bold', 'mean', 'std'],
+    num_rows: 5187
+}),
+    labels=[ 3  4  5  6 10 11 17 18 19 20 22 23 25 30 31 33 36 37 38 53 55 59 61 74],
+    counts=[197 159 267 273 123 153 175 184 139 310 215 386 153 230 118 192 330 306
+ 349 223 143 127 249 186]
+)
+
+running backbone on example batch to get embedding dim
+embedding feature dim (patch): 768
+initializing sweep of classifier heads
+classifiers:
+ModuleList(
+  (0-48): 49 x AttnPoolClassifier(
+    (kv): Linear(in_features=768, out_features=1536, bias=True)
+    (linear): Linear(in_features=768, out_features=24, bias=True)
+  )
+)
+classifier params (train): 58.8M (58.8M)
+setting up optimizer
+total batch size: 128 = 64 bs per gpu x 2 accum
+lr: 3.00e-04
+full schedule: epochs = 20 (steps = 4000) (decay = True)
+warmup: epochs = 5 (steps = 1000)
+start training for 20 epochs
+train: [0]  [  0/400]  eta: 0:10:54  lr: nan  time: 1.6350  data: 0.8118  max mem: 56639
+train: [0]  [ 20/400]  eta: 0:04:27  lr: 0.000003  loss: 3.2013 (3.2077)  grad: 0.1921 (0.1978)  time: 0.6572  data: 0.0032  max mem: 57344
+train: [0]  [ 40/400]  eta: 0:04:03  lr: 0.000006  loss: 3.1875 (3.1922)  grad: 0.1912 (0.1889)  time: 0.6493  data: 0.0037  max mem: 57344
+train: [0]  [ 60/400]  eta: 0:03:47  lr: 0.000009  loss: 3.1743 (3.1881)  grad: 0.1871 (0.1910)  time: 0.6484  data: 0.0033  max mem: 57344
+train: [0]  [ 80/400]  eta: 0:03:32  lr: 0.000012  loss: 3.1630 (3.1802)  grad: 0.1868 (0.1888)  time: 0.6503  data: 0.0040  max mem: 57344
+train: [0]  [100/400]  eta: 0:03:18  lr: 0.000015  loss: 3.1592 (3.1788)  grad: 0.1798 (0.1887)  time: 0.6500  data: 0.0039  max mem: 57344
+train: [0]  [120/400]  eta: 0:03:04  lr: 0.000018  loss: 3.1679 (3.1757)  grad: 0.1753 (0.1865)  time: 0.6494  data: 0.0037  max mem: 57344
+train: [0]  [140/400]  eta: 0:02:50  lr: 0.000021  loss: 3.1475 (3.1722)  grad: 0.1796 (0.1859)  time: 0.6498  data: 0.0037  max mem: 57344
+train: [0]  [160/400]  eta: 0:02:37  lr: 0.000024  loss: 3.1586 (3.1716)  grad: 0.1774 (0.1845)  time: 0.6500  data: 0.0037  max mem: 57344
+train: [0]  [180/400]  eta: 0:02:24  lr: 0.000027  loss: 3.1658 (3.1705)  grad: 0.1659 (0.1828)  time: 0.6501  data: 0.0037  max mem: 57344
+train: [0]  [200/400]  eta: 0:02:11  lr: 0.000030  loss: 3.1452 (3.1673)  grad: 0.1654 (0.1820)  time: 0.6509  data: 0.0037  max mem: 57344
+train: [0]  [220/400]  eta: 0:01:57  lr: 0.000033  loss: 3.1360 (3.1655)  grad: 0.1792 (0.1825)  time: 0.6507  data: 0.0037  max mem: 57344
+train: [0]  [240/400]  eta: 0:01:44  lr: 0.000036  loss: 3.1394 (3.1645)  grad: 0.1792 (0.1819)  time: 0.6508  data: 0.0037  max mem: 57344
+train: [0]  [260/400]  eta: 0:01:31  lr: 0.000039  loss: 3.1394 (3.1637)  grad: 0.1700 (0.1807)  time: 0.6509  data: 0.0037  max mem: 57344
+train: [0]  [280/400]  eta: 0:01:18  lr: 0.000042  loss: 3.1329 (3.1615)  grad: 0.1698 (0.1799)  time: 0.6504  data: 0.0037  max mem: 57344
+train: [0]  [300/400]  eta: 0:01:05  lr: 0.000045  loss: 3.1432 (3.1608)  grad: 0.1712 (0.1794)  time: 0.6497  data: 0.0036  max mem: 57344
+train: [0]  [320/400]  eta: 0:00:52  lr: 0.000048  loss: 3.1312 (3.1586)  grad: 0.1772 (0.1797)  time: 0.6499  data: 0.0037  max mem: 57344
+train: [0]  [340/400]  eta: 0:00:39  lr: 0.000051  loss: 3.1090 (3.1567)  grad: 0.1788 (0.1792)  time: 0.6493  data: 0.0037  max mem: 57344
+train: [0]  [360/400]  eta: 0:00:26  lr: 0.000054  loss: 3.1221 (3.1555)  grad: 0.1709 (0.1790)  time: 0.6500  data: 0.0039  max mem: 57344
+train: [0]  [380/400]  eta: 0:00:13  lr: 0.000057  loss: 3.1196 (3.1529)  grad: 0.1685 (0.1787)  time: 0.6499  data: 0.0037  max mem: 57344
+train: [0]  [399/400]  eta: 0:00:00  lr: 0.000060  loss: 3.1127 (3.1506)  grad: 0.1685 (0.1784)  time: 0.6497  data: 0.0037  max mem: 57344
+train: [0] Total time: 0:04:21 (0.6531 s / it)
+train: [0] Summary: lr: 0.000060  loss: 3.1127 (3.1506)  grad: 0.1685 (0.1784)
+eval (validation): [0]  [ 0/85]  eta: 0:01:33    time: 1.0964  data: 0.7304  max mem: 57344
+eval (validation): [0]  [20/85]  eta: 0:00:26    time: 0.3722  data: 0.0026  max mem: 57344
+eval (validation): [0]  [40/85]  eta: 0:00:17    time: 0.3731  data: 0.0038  max mem: 57344
+eval (validation): [0]  [60/85]  eta: 0:00:09    time: 0.3733  data: 0.0036  max mem: 57344
+eval (validation): [0]  [80/85]  eta: 0:00:01    time: 0.3721  data: 0.0034  max mem: 57344
+eval (validation): [0]  [84/85]  eta: 0:00:00    time: 0.3659  data: 0.0034  max mem: 57344
+eval (validation): [0] Total time: 0:00:32 (0.3809 s / it)
+cv: [0]  best hparam: (50, 1.0) (048) ('048_lr5.0e+01_wd1.0e+00')  loss: 2.679  acc: 0.192  f1: 0.128
+saving checkpoint experiments/schaefer1000/output/schaefer1000/schaefer1000_lr3e-4_4/eval_v2/nsd_cococlip__patch__attn/checkpoint-last.pth
+saving best checkpoint experiments/schaefer1000/output/schaefer1000/schaefer1000_lr3e-4_4/eval_v2/nsd_cococlip__patch__attn/checkpoint-best.pth
+train: [1]  [  0/400]  eta: 0:07:44  lr: nan  time: 1.1616  data: 0.5254  max mem: 57344
+train: [1]  [ 20/400]  eta: 0:04:15  lr: 0.000063  loss: 3.0269 (3.0638)  grad: 0.1706 (0.1770)  time: 0.6485  data: 0.0031  max mem: 57344
+train: [1]  [ 40/400]  eta: 0:03:58  lr: 0.000066  loss: 3.0486 (3.0683)  grad: 0.1698 (0.1726)  time: 0.6508  data: 0.0038  max mem: 57344
+train: [1]  [ 60/400]  eta: 0:03:44  lr: 0.000069  loss: 3.0607 (3.0711)  grad: 0.1724 (0.1780)  time: 0.6522  data: 0.0039  max mem: 57344
+train: [1]  [ 80/400]  eta: 0:03:30  lr: 0.000072  loss: 3.0792 (3.0781)  grad: 0.1864 (0.1816)  time: 0.6491  data: 0.0034  max mem: 57344
+train: [1]  [100/400]  eta: 0:03:16  lr: 0.000075  loss: 3.0663 (3.0741)  grad: 0.1903 (0.1834)  time: 0.6504  data: 0.0036  max mem: 57344
+train: [1]  [120/400]  eta: 0:03:03  lr: 0.000078  loss: 3.0443 (3.0696)  grad: 0.1880 (0.1835)  time: 0.6501  data: 0.0038  max mem: 57344
+train: [1]  [140/400]  eta: 0:02:50  lr: 0.000081  loss: 3.0230 (3.0621)  grad: 0.1803 (0.1845)  time: 0.6506  data: 0.0036  max mem: 57344
+train: [1]  [160/400]  eta: 0:02:36  lr: 0.000084  loss: 3.0337 (3.0592)  grad: 0.1798 (0.1844)  time: 0.6500  data: 0.0037  max mem: 57344
+train: [1]  [180/400]  eta: 0:02:23  lr: 0.000087  loss: 3.0485 (3.0593)  grad: 0.1889 (0.1851)  time: 0.6504  data: 0.0037  max mem: 57344
+train: [1]  [200/400]  eta: 0:02:10  lr: 0.000090  loss: 3.0523 (3.0579)  grad: 0.1954 (0.1868)  time: 0.6508  data: 0.0037  max mem: 57344
+train: [1]  [220/400]  eta: 0:01:57  lr: 0.000093  loss: 3.0337 (3.0557)  grad: 0.1995 (0.1876)  time: 0.6499  data: 0.0036  max mem: 57344
+train: [1]  [240/400]  eta: 0:01:44  lr: 0.000096  loss: 3.0090 (3.0514)  grad: 0.1880 (0.1879)  time: 0.6506  data: 0.0036  max mem: 57344
+train: [1]  [260/400]  eta: 0:01:31  lr: 0.000099  loss: 2.9946 (3.0493)  grad: 0.2027 (0.1898)  time: 0.6498  data: 0.0036  max mem: 57344
+train: [1]  [280/400]  eta: 0:01:18  lr: 0.000102  loss: 3.0201 (3.0477)  grad: 0.2177 (0.1930)  time: 0.6496  data: 0.0035  max mem: 57344
+train: [1]  [300/400]  eta: 0:01:05  lr: 0.000105  loss: 3.0102 (3.0461)  grad: 0.2602 (0.2025)  time: 0.6500  data: 0.0037  max mem: 57344
+train: [1]  [320/400]  eta: 0:00:52  lr: 0.000108  loss: 3.1198 (3.0621)  grad: 0.4035 (0.2505)  time: 0.6504  data: 0.0036  max mem: 57344
+train: [1]  [340/400]  eta: 0:00:39  lr: 0.000111  loss: 3.4640 (3.0987)  grad: 1.2664 (0.3213)  time: 0.6510  data: 0.0037  max mem: 57344
+WARNING: classifier 48 (50, 1.0) diverged (loss=63.69 > 63.56) at step 371. Freezing.
+train: [1]  [360/400]  eta: 0:00:26  lr: 0.000114  loss: 3.4628 (3.1008)  grad: 1.1506 (0.3250)  time: 0.6450  data: 0.0037  max mem: 57344
+train: [1]  [380/400]  eta: 0:00:13  lr: 0.000117  loss: 3.0149 (3.0958)  grad: 0.2018 (0.3188)  time: 0.6443  data: 0.0036  max mem: 57344
+train: [1]  [399/400]  eta: 0:00:00  lr: 0.000120  loss: 3.0157 (3.0921)  grad: 0.2018 (0.3130)  time: 0.6451  data: 0.0037  max mem: 57344
+train: [1] Total time: 0:04:20 (0.6510 s / it)
+train: [1] Summary: lr: 0.000120  loss: 3.0157 (3.0921)  grad: 0.2018 (0.3130)
+eval (validation): [1]  [ 0/85]  eta: 0:01:24    time: 0.9989  data: 0.6375  max mem: 57344
+eval (validation): [1]  [20/85]  eta: 0:00:26    time: 0.3719  data: 0.0037  max mem: 57344
+eval (validation): [1]  [40/85]  eta: 0:00:17    time: 0.3731  data: 0.0034  max mem: 57344
+eval (validation): [1]  [60/85]  eta: 0:00:09    time: 0.3731  data: 0.0036  max mem: 57344
+eval (validation): [1]  [80/85]  eta: 0:00:01    time: 0.3724  data: 0.0034  max mem: 57344
+eval (validation): [1]  [84/85]  eta: 0:00:00    time: 0.3660  data: 0.0034  max mem: 57344
+eval (validation): [1] Total time: 0:00:32 (0.3799 s / it)
+cv: [1]  best hparam: (16, 1.0) (041) ('041_lr1.6e+01_wd1.0e+00')  loss: 2.546  acc: 0.229  f1: 0.161
+saving checkpoint experiments/schaefer1000/output/schaefer1000/schaefer1000_lr3e-4_4/eval_v2/nsd_cococlip__patch__attn/checkpoint-last.pth
+saving best checkpoint experiments/schaefer1000/output/schaefer1000/schaefer1000_lr3e-4_4/eval_v2/nsd_cococlip__patch__attn/checkpoint-best.pth
+train: [2]  [  0/400]  eta: 0:07:37  lr: nan  time: 1.1445  data: 0.5144  max mem: 57344
+train: [2]  [ 20/400]  eta: 0:04:13  lr: 0.000123  loss: 2.9171 (2.9554)  grad: 0.1845 (0.1896)  time: 0.6437  data: 0.0033  max mem: 57344
+train: [2]  [ 40/400]  eta: 0:03:56  lr: 0.000126  loss: 2.9311 (2.9432)  grad: 0.1892 (0.1939)  time: 0.6446  data: 0.0037  max mem: 57344
+train: [2]  [ 60/400]  eta: 0:03:41  lr: 0.000129  loss: 2.9311 (2.9441)  grad: 0.2009 (0.2013)  time: 0.6433  data: 0.0035  max mem: 57344
+train: [2]  [ 80/400]  eta: 0:03:28  lr: 0.000132  loss: 2.9738 (2.9543)  grad: 0.2199 (0.2120)  time: 0.6448  data: 0.0036  max mem: 57344
+train: [2]  [100/400]  eta: 0:03:14  lr: 0.000135  loss: 2.9761 (2.9603)  grad: 0.2505 (0.2195)  time: 0.6443  data: 0.0036  max mem: 57344
+train: [2]  [120/400]  eta: 0:03:01  lr: 0.000138  loss: 3.0313 (2.9961)  grad: 0.2851 (0.3178)  time: 0.6445  data: 0.0037  max mem: 57344
+WARNING: classifier 47 (43, 1.0) diverged (loss=72.55 > 63.56) at step 467. Freezing.
+train: [2]  [140/400]  eta: 0:02:48  lr: 0.000141  loss: 3.2434 (3.0785)  grad: 1.0274 (0.4842)  time: 0.6432  data: 0.0036  max mem: 57344
+train: [2]  [160/400]  eta: 0:02:35  lr: 0.000144  loss: 2.9834 (3.0615)  grad: 0.2087 (0.4487)  time: 0.6391  data: 0.0035  max mem: 57344
+train: [2]  [180/400]  eta: 0:02:22  lr: 0.000147  loss: 2.9371 (3.0475)  grad: 0.2055 (0.4237)  time: 0.6394  data: 0.0037  max mem: 57344
+train: [2]  [200/400]  eta: 0:02:09  lr: 0.000150  loss: 2.9231 (3.0349)  grad: 0.2184 (0.4032)  time: 0.6393  data: 0.0037  max mem: 57344
+train: [2]  [220/400]  eta: 0:01:56  lr: 0.000153  loss: 2.9243 (3.0250)  grad: 0.2135 (0.3858)  time: 0.6393  data: 0.0037  max mem: 57344
+train: [2]  [240/400]  eta: 0:01:43  lr: 0.000156  loss: 2.9363 (3.0186)  grad: 0.2154 (0.3719)  time: 0.6391  data: 0.0037  max mem: 57344
+train: [2]  [260/400]  eta: 0:01:30  lr: 0.000159  loss: 2.9420 (3.0130)  grad: 0.2187 (0.3609)  time: 0.6394  data: 0.0039  max mem: 57344
+train: [2]  [280/400]  eta: 0:01:17  lr: 0.000162  loss: 2.9420 (3.0100)  grad: 0.2355 (0.3520)  time: 0.6398  data: 0.0038  max mem: 57344
+train: [2]  [300/400]  eta: 0:01:04  lr: 0.000165  loss: 2.9656 (3.0077)  grad: 0.2509 (0.3475)  time: 0.6393  data: 0.0037  max mem: 57344
+train: [2]  [320/400]  eta: 0:00:51  lr: 0.000168  loss: 3.0748 (3.0355)  grad: 0.3825 (0.4211)  time: 0.6391  data: 0.0037  max mem: 57344
+WARNING: classifier 46 (36, 1.0) diverged (loss=74.23 > 63.56) at step 561. Freezing.
+train: [2]  [340/400]  eta: 0:00:38  lr: 0.000171  loss: 3.1002 (3.0393)  grad: 0.5135 (0.4292)  time: 0.6339  data: 0.0037  max mem: 57344
+train: [2]  [360/400]  eta: 0:00:25  lr: 0.000174  loss: 2.9508 (3.0352)  grad: 0.2296 (0.4183)  time: 0.6344  data: 0.0037  max mem: 57344
+train: [2]  [380/400]  eta: 0:00:12  lr: 0.000177  loss: 2.9513 (3.0295)  grad: 0.2347 (0.4089)  time: 0.6338  data: 0.0037  max mem: 57344
+train: [2]  [399/400]  eta: 0:00:00  lr: 0.000180  loss: 2.9394 (3.0252)  grad: 0.2461 (0.4048)  time: 0.6335  data: 0.0038  max mem: 57344
+train: [2] Total time: 0:04:16 (0.6414 s / it)
+train: [2] Summary: lr: 0.000180  loss: 2.9394 (3.0252)  grad: 0.2461 (0.4048)
+eval (validation): [2]  [ 0/85]  eta: 0:01:25    time: 1.0083  data: 0.6486  max mem: 57344
+eval (validation): [2]  [20/85]  eta: 0:00:26    time: 0.3708  data: 0.0022  max mem: 57344
+eval (validation): [2]  [40/85]  eta: 0:00:17    time: 0.3723  data: 0.0036  max mem: 57344
+eval (validation): [2]  [60/85]  eta: 0:00:09    time: 0.3718  data: 0.0034  max mem: 57344
+eval (validation): [2]  [80/85]  eta: 0:00:01    time: 0.3717  data: 0.0035  max mem: 57344
+eval (validation): [2]  [84/85]  eta: 0:00:00    time: 0.3652  data: 0.0034  max mem: 57344
+eval (validation): [2] Total time: 0:00:32 (0.3788 s / it)
+cv: [2]  best hparam: (8.3, 1.0) (037) ('037_lr8.3e+00_wd1.0e+00')  loss: 2.481  acc: 0.243  f1: 0.189
+saving checkpoint experiments/schaefer1000/output/schaefer1000/schaefer1000_lr3e-4_4/eval_v2/nsd_cococlip__patch__attn/checkpoint-last.pth
+saving best checkpoint experiments/schaefer1000/output/schaefer1000/schaefer1000_lr3e-4_4/eval_v2/nsd_cococlip__patch__attn/checkpoint-best.pth
+train: [3]  [  0/400]  eta: 0:07:21  lr: nan  time: 1.1033  data: 0.4852  max mem: 57344
+train: [3]  [ 20/400]  eta: 0:04:08  lr: 0.000183  loss: 3.1719 (3.2783)  grad: 0.8354 (1.0384)  time: 0.6322  data: 0.0031  max mem: 57344
+WARNING: classifier 45 (31, 1.0) diverged (loss=73.62 > 63.56) at step 617. Freezing.
+train: [3]  [ 40/400]  eta: 0:03:51  lr: 0.000186  loss: 3.3767 (3.4747)  grad: 1.1361 (1.2126)  time: 0.6310  data: 0.0035  max mem: 57344
+train: [3]  [ 60/400]  eta: 0:03:36  lr: 0.000189  loss: 2.9865 (3.2877)  grad: 0.2221 (0.8782)  time: 0.6272  data: 0.0038  max mem: 57344
+train: [3]  [ 80/400]  eta: 0:03:23  lr: 0.000192  loss: 2.8933 (3.1859)  grad: 0.2136 (0.7135)  time: 0.6261  data: 0.0035  max mem: 57344
+train: [3]  [100/400]  eta: 0:03:10  lr: 0.000195  loss: 2.9099 (3.1336)  grad: 0.2147 (0.6157)  time: 0.6280  data: 0.0038  max mem: 57344
+train: [3]  [120/400]  eta: 0:02:57  lr: 0.000198  loss: 2.9171 (3.0964)  grad: 0.2224 (0.5504)  time: 0.6282  data: 0.0038  max mem: 57344
+train: [3]  [140/400]  eta: 0:02:44  lr: 0.000201  loss: 2.9163 (3.0691)  grad: 0.2146 (0.5021)  time: 0.6281  data: 0.0038  max mem: 57344
+train: [3]  [160/400]  eta: 0:02:31  lr: 0.000204  loss: 2.8473 (3.0395)  grad: 0.2112 (0.4676)  time: 0.6277  data: 0.0038  max mem: 57344
+train: [3]  [180/400]  eta: 0:02:18  lr: 0.000207  loss: 2.9071 (3.0310)  grad: 0.2606 (0.4470)  time: 0.6274  data: 0.0037  max mem: 57344
+train: [3]  [200/400]  eta: 0:02:06  lr: 0.000210  loss: 3.0273 (3.0418)  grad: 0.3522 (0.4821)  time: 0.6281  data: 0.0038  max mem: 57344
+WARNING: classifier 44 (26, 1.0) diverged (loss=69.16 > 63.56) at step 705. Freezing.
+train: [3]  [220/400]  eta: 0:01:53  lr: 0.000213  loss: 3.0835 (3.0691)  grad: 0.6844 (0.5416)  time: 0.6250  data: 0.0038  max mem: 57344
+train: [3]  [240/400]  eta: 0:01:40  lr: 0.000216  loss: 2.9286 (3.0542)  grad: 0.2218 (0.5150)  time: 0.6213  data: 0.0038  max mem: 57344
+train: [3]  [260/400]  eta: 0:01:28  lr: 0.000219  loss: 2.8700 (3.0419)  grad: 0.2141 (0.4917)  time: 0.6229  data: 0.0039  max mem: 57344
+train: [3]  [280/400]  eta: 0:01:15  lr: 0.000222  loss: 2.8680 (3.0294)  grad: 0.2072 (0.4715)  time: 0.6222  data: 0.0038  max mem: 57344
+train: [3]  [300/400]  eta: 0:01:02  lr: 0.000225  loss: 2.8690 (3.0188)  grad: 0.2130 (0.4547)  time: 0.6216  data: 0.0038  max mem: 57344
+train: [3]  [320/400]  eta: 0:00:50  lr: 0.000228  loss: 2.8808 (3.0114)  grad: 0.2180 (0.4398)  time: 0.6214  data: 0.0038  max mem: 57344
+train: [3]  [340/400]  eta: 0:00:37  lr: 0.000231  loss: 2.8788 (3.0030)  grad: 0.2212 (0.4265)  time: 0.6223  data: 0.0038  max mem: 57344
+train: [3]  [360/400]  eta: 0:00:25  lr: 0.000234  loss: 2.8788 (2.9963)  grad: 0.2229 (0.4156)  time: 0.6229  data: 0.0039  max mem: 57344
+train: [3]  [380/400]  eta: 0:00:12  lr: 0.000237  loss: 2.8896 (2.9907)  grad: 0.2271 (0.4058)  time: 0.6221  data: 0.0038  max mem: 57344
+train: [3]  [399/400]  eta: 0:00:00  lr: 0.000240  loss: 2.9018 (2.9860)  grad: 0.2265 (0.3970)  time: 0.6215  data: 0.0038  max mem: 57344
+train: [3] Total time: 0:04:10 (0.6268 s / it)
+train: [3] Summary: lr: 0.000240  loss: 2.9018 (2.9860)  grad: 0.2265 (0.3970)
+eval (validation): [3]  [ 0/85]  eta: 0:01:23    time: 0.9846  data: 0.6258  max mem: 57344
+eval (validation): [3]  [20/85]  eta: 0:00:25    time: 0.3706  data: 0.0026  max mem: 57344
+eval (validation): [3]  [40/85]  eta: 0:00:17    time: 0.3713  data: 0.0036  max mem: 57344
+eval (validation): [3]  [60/85]  eta: 0:00:09    time: 0.3721  data: 0.0036  max mem: 57344
+eval (validation): [3]  [80/85]  eta: 0:00:01    time: 0.3722  data: 0.0036  max mem: 57344
+eval (validation): [3]  [84/85]  eta: 0:00:00    time: 0.3660  data: 0.0036  max mem: 57344
+eval (validation): [3] Total time: 0:00:32 (0.3786 s / it)
+cv: [3]  best hparam: (3.1, 1.0) (031) ('031_lr3.1e+00_wd1.0e+00')  loss: 2.538  acc: 0.241  f1: 0.186
+saving checkpoint experiments/schaefer1000/output/schaefer1000/schaefer1000_lr3e-4_4/eval_v2/nsd_cococlip__patch__attn/checkpoint-last.pth
+train: [4]  [  0/400]  eta: 0:08:10  lr: nan  time: 1.2269  data: 0.6186  max mem: 57344
+train: [4]  [ 20/400]  eta: 0:04:06  lr: 0.000243  loss: 2.8730 (2.8802)  grad: 0.2246 (0.2261)  time: 0.6204  data: 0.0032  max mem: 57344
+train: [4]  [ 40/400]  eta: 0:03:48  lr: 0.000246  loss: 2.8677 (2.8619)  grad: 0.2132 (0.2176)  time: 0.6192  data: 0.0034  max mem: 57344
+train: [4]  [ 60/400]  eta: 0:03:34  lr: 0.000249  loss: 2.8598 (2.8611)  grad: 0.2144 (0.2218)  time: 0.6225  data: 0.0039  max mem: 57344
+train: [4]  [ 80/400]  eta: 0:03:21  lr: 0.000252  loss: 2.8577 (2.8586)  grad: 0.2223 (0.2215)  time: 0.6224  data: 0.0039  max mem: 57344
+train: [4]  [100/400]  eta: 0:03:08  lr: 0.000255  loss: 2.8371 (2.8550)  grad: 0.2260 (0.2234)  time: 0.6214  data: 0.0037  max mem: 57344
+train: [4]  [120/400]  eta: 0:02:55  lr: 0.000258  loss: 2.8386 (2.8524)  grad: 0.2204 (0.2209)  time: 0.6206  data: 0.0037  max mem: 57344
+train: [4]  [140/400]  eta: 0:02:42  lr: 0.000261  loss: 2.8290 (2.8470)  grad: 0.2156 (0.2215)  time: 0.6220  data: 0.0038  max mem: 57344
+train: [4]  [160/400]  eta: 0:02:30  lr: 0.000264  loss: 2.8030 (2.8437)  grad: 0.2259 (0.2223)  time: 0.6218  data: 0.0038  max mem: 57344
+train: [4]  [180/400]  eta: 0:02:17  lr: 0.000267  loss: 2.8498 (2.8438)  grad: 0.2267 (0.2243)  time: 0.6222  data: 0.0038  max mem: 57344
+train: [4]  [200/400]  eta: 0:02:04  lr: 0.000270  loss: 2.8560 (2.8459)  grad: 0.2279 (0.2256)  time: 0.6227  data: 0.0039  max mem: 57344
+train: [4]  [220/400]  eta: 0:01:52  lr: 0.000273  loss: 2.8615 (2.8484)  grad: 0.2522 (0.2297)  time: 0.6215  data: 0.0039  max mem: 57344
+train: [4]  [240/400]  eta: 0:01:39  lr: 0.000276  loss: 2.8448 (2.8487)  grad: 0.2755 (0.2346)  time: 0.6220  data: 0.0040  max mem: 57344
+train: [4]  [260/400]  eta: 0:01:27  lr: 0.000279  loss: 2.9275 (2.8690)  grad: 0.3400 (0.2777)  time: 0.6214  data: 0.0037  max mem: 57344
+WARNING: classifier 43 (22, 1.0) diverged (loss=77.46 > 63.56) at step 933. Freezing.
+train: [4]  [280/400]  eta: 0:01:14  lr: 0.000282  loss: 2.9518 (2.8907)  grad: 0.3995 (0.3090)  time: 0.6182  data: 0.0040  max mem: 57344
+train: [4]  [300/400]  eta: 0:01:02  lr: 0.000285  loss: 2.8755 (2.8887)  grad: 0.2221 (0.3027)  time: 0.6162  data: 0.0039  max mem: 57344
+train: [4]  [320/400]  eta: 0:00:49  lr: 0.000288  loss: 2.8502 (2.8862)  grad: 0.2221 (0.2985)  time: 0.6159  data: 0.0039  max mem: 57344
+train: [4]  [340/400]  eta: 0:00:37  lr: 0.000291  loss: 2.8502 (2.8860)  grad: 0.2393 (0.2954)  time: 0.6158  data: 0.0039  max mem: 57344
+train: [4]  [360/400]  eta: 0:00:24  lr: 0.000294  loss: 2.8724 (2.8876)  grad: 0.2538 (0.3033)  time: 0.6168  data: 0.0039  max mem: 57344
+train: [4]  [380/400]  eta: 0:00:12  lr: 0.000297  loss: 3.0781 (2.9208)  grad: 0.7847 (0.3792)  time: 0.6159  data: 0.0039  max mem: 57344
+WARNING: classifier 42 (19, 1.0) diverged (loss=98.48 > 63.56) at step 991. Freezing.
+train: [4]  [399/400]  eta: 0:00:00  lr: 0.000300  loss: 3.0781 (2.9266)  grad: 0.9276 (0.3832)  time: 0.6114  data: 0.0038  max mem: 57344
+train: [4] Total time: 0:04:08 (0.6213 s / it)
+train: [4] Summary: lr: 0.000300  loss: 3.0781 (2.9266)  grad: 0.9276 (0.3832)
+eval (validation): [4]  [ 0/85]  eta: 0:01:31    time: 1.0780  data: 0.7152  max mem: 57344
+eval (validation): [4]  [20/85]  eta: 0:00:26    time: 0.3705  data: 0.0023  max mem: 57344
+eval (validation): [4]  [40/85]  eta: 0:00:17    time: 0.3713  data: 0.0037  max mem: 57344
+eval (validation): [4]  [60/85]  eta: 0:00:09    time: 0.3718  data: 0.0038  max mem: 57344
+eval (validation): [4]  [80/85]  eta: 0:00:01    time: 0.3713  data: 0.0038  max mem: 57344
+eval (validation): [4]  [84/85]  eta: 0:00:00    time: 0.3650  data: 0.0038  max mem: 57344
+eval (validation): [4] Total time: 0:00:32 (0.3793 s / it)
+cv: [4]  best hparam: (1.6, 1.0) (027) ('027_lr1.6e+00_wd1.0e+00')  loss: 2.476  acc: 0.257  f1: 0.193
+saving checkpoint experiments/schaefer1000/output/schaefer1000/schaefer1000_lr3e-4_4/eval_v2/nsd_cococlip__patch__attn/checkpoint-last.pth
+saving best checkpoint experiments/schaefer1000/output/schaefer1000/schaefer1000_lr3e-4_4/eval_v2/nsd_cococlip__patch__attn/checkpoint-best.pth
+train: [5]  [  0/400]  eta: 0:08:35  lr: nan  time: 1.2892  data: 0.6929  max mem: 57344
+train: [5]  [ 20/400]  eta: 0:04:04  lr: 0.000300  loss: 2.8414 (2.8498)  grad: 0.2202 (0.2234)  time: 0.6104  data: 0.0028  max mem: 57344
+train: [5]  [ 40/400]  eta: 0:03:45  lr: 0.000300  loss: 2.8235 (2.8358)  grad: 0.2169 (0.2218)  time: 0.6118  data: 0.0039  max mem: 57344
+train: [5]  [ 60/400]  eta: 0:03:31  lr: 0.000300  loss: 2.8412 (2.8377)  grad: 0.2250 (0.2272)  time: 0.6105  data: 0.0038  max mem: 57344
+train: [5]  [ 80/400]  eta: 0:03:18  lr: 0.000300  loss: 2.8476 (2.8366)  grad: 0.2316 (0.2268)  time: 0.6107  data: 0.0039  max mem: 57344
+train: [5]  [100/400]  eta: 0:03:05  lr: 0.000300  loss: 2.8354 (2.8375)  grad: 0.2311 (0.2270)  time: 0.6096  data: 0.0038  max mem: 57344
+train: [5]  [120/400]  eta: 0:02:52  lr: 0.000300  loss: 2.8308 (2.8371)  grad: 0.2311 (0.2288)  time: 0.6087  data: 0.0034  max mem: 57344
+train: [5]  [140/400]  eta: 0:02:39  lr: 0.000300  loss: 2.8131 (2.8357)  grad: 0.2340 (0.2301)  time: 0.6105  data: 0.0037  max mem: 57344
+train: [5]  [160/400]  eta: 0:02:27  lr: 0.000299  loss: 2.7975 (2.8296)  grad: 0.2264 (0.2294)  time: 0.6099  data: 0.0038  max mem: 57344
+train: [5]  [180/400]  eta: 0:02:15  lr: 0.000299  loss: 2.8039 (2.8295)  grad: 0.2241 (0.2293)  time: 0.6091  data: 0.0036  max mem: 57344
+train: [5]  [200/400]  eta: 0:02:02  lr: 0.000299  loss: 2.8136 (2.8257)  grad: 0.2252 (0.2293)  time: 0.6089  data: 0.0036  max mem: 57344
+train: [5]  [220/400]  eta: 0:01:50  lr: 0.000299  loss: 2.7876 (2.8220)  grad: 0.2348 (0.2298)  time: 0.6103  data: 0.0038  max mem: 57344
+train: [5]  [240/400]  eta: 0:01:38  lr: 0.000299  loss: 2.7718 (2.8168)  grad: 0.2295 (0.2286)  time: 0.6107  data: 0.0039  max mem: 57344
+train: [5]  [260/400]  eta: 0:01:25  lr: 0.000299  loss: 2.8093 (2.8181)  grad: 0.2126 (0.2280)  time: 0.6102  data: 0.0040  max mem: 57344
+train: [5]  [280/400]  eta: 0:01:13  lr: 0.000298  loss: 2.8040 (2.8160)  grad: 0.2184 (0.2276)  time: 0.6098  data: 0.0039  max mem: 57344
+train: [5]  [300/400]  eta: 0:01:01  lr: 0.000298  loss: 2.7764 (2.8141)  grad: 0.2205 (0.2269)  time: 0.6107  data: 0.0039  max mem: 57344
+train: [5]  [320/400]  eta: 0:00:48  lr: 0.000298  loss: 2.8313 (2.8161)  grad: 0.2100 (0.2260)  time: 0.6099  data: 0.0039  max mem: 57344
+train: [5]  [340/400]  eta: 0:00:36  lr: 0.000298  loss: 2.8321 (2.8150)  grad: 0.2111 (0.2262)  time: 0.6098  data: 0.0038  max mem: 57344
+train: [5]  [360/400]  eta: 0:00:24  lr: 0.000297  loss: 2.8277 (2.8154)  grad: 0.2342 (0.2265)  time: 0.6096  data: 0.0038  max mem: 57344
+train: [5]  [380/400]  eta: 0:00:12  lr: 0.000297  loss: 2.8265 (2.8148)  grad: 0.2258 (0.2262)  time: 0.6097  data: 0.0037  max mem: 57344
+train: [5]  [399/400]  eta: 0:00:00  lr: 0.000297  loss: 2.7900 (2.8137)  grad: 0.2086 (0.2251)  time: 0.6109  data: 0.0038  max mem: 57344
+train: [5] Total time: 0:04:04 (0.6121 s / it)
+train: [5] Summary: lr: 0.000297  loss: 2.7900 (2.8137)  grad: 0.2086 (0.2251)
+eval (validation): [5]  [ 0/85]  eta: 0:01:25    time: 1.0008  data: 0.6370  max mem: 57344
+eval (validation): [5]  [20/85]  eta: 0:00:26    time: 0.3718  data: 0.0038  max mem: 57344
+eval (validation): [5]  [40/85]  eta: 0:00:17    time: 0.3716  data: 0.0037  max mem: 57344
+eval (validation): [5]  [60/85]  eta: 0:00:09    time: 0.3716  data: 0.0039  max mem: 57344
+eval (validation): [5]  [80/85]  eta: 0:00:01    time: 0.3718  data: 0.0039  max mem: 57344
+eval (validation): [5]  [84/85]  eta: 0:00:00    time: 0.3654  data: 0.0038  max mem: 57344
+eval (validation): [5] Total time: 0:00:32 (0.3789 s / it)
+cv: [5]  best hparam: (1.9, 1.0) (028) ('028_lr1.9e+00_wd1.0e+00')  loss: 2.405  acc: 0.268  f1: 0.208
+saving checkpoint experiments/schaefer1000/output/schaefer1000/schaefer1000_lr3e-4_4/eval_v2/nsd_cococlip__patch__attn/checkpoint-last.pth
+saving best checkpoint experiments/schaefer1000/output/schaefer1000/schaefer1000_lr3e-4_4/eval_v2/nsd_cococlip__patch__attn/checkpoint-best.pth
+train: [6]  [  0/400]  eta: 0:08:21  lr: nan  time: 1.2546  data: 0.6561  max mem: 57344
+train: [6]  [ 20/400]  eta: 0:04:03  lr: 0.000296  loss: 2.7806 (2.7939)  grad: 0.2140 (0.2168)  time: 0.6089  data: 0.0032  max mem: 57344
+train: [6]  [ 40/400]  eta: 0:03:45  lr: 0.000296  loss: 2.7674 (2.7713)  grad: 0.2160 (0.2199)  time: 0.6098  data: 0.0039  max mem: 57344
+train: [6]  [ 60/400]  eta: 0:03:30  lr: 0.000296  loss: 2.7602 (2.7710)  grad: 0.2133 (0.2190)  time: 0.6101  data: 0.0039  max mem: 57344
+train: [6]  [ 80/400]  eta: 0:03:17  lr: 0.000295  loss: 2.7487 (2.7624)  grad: 0.2098 (0.2178)  time: 0.6108  data: 0.0039  max mem: 57344
+train: [6]  [100/400]  eta: 0:03:04  lr: 0.000295  loss: 2.7201 (2.7584)  grad: 0.2123 (0.2168)  time: 0.6105  data: 0.0039  max mem: 57344
+train: [6]  [120/400]  eta: 0:02:52  lr: 0.000295  loss: 2.7608 (2.7611)  grad: 0.2123 (0.2165)  time: 0.6106  data: 0.0038  max mem: 57344
+train: [6]  [140/400]  eta: 0:02:39  lr: 0.000294  loss: 2.7572 (2.7533)  grad: 0.2069 (0.2143)  time: 0.6091  data: 0.0036  max mem: 57344
+train: [6]  [160/400]  eta: 0:02:27  lr: 0.000294  loss: 2.6980 (2.7468)  grad: 0.2081 (0.2150)  time: 0.6080  data: 0.0034  max mem: 57344
+train: [6]  [180/400]  eta: 0:02:14  lr: 0.000293  loss: 2.7025 (2.7483)  grad: 0.2168 (0.2157)  time: 0.6098  data: 0.0037  max mem: 57344
+train: [6]  [200/400]  eta: 0:02:02  lr: 0.000293  loss: 2.7185 (2.7459)  grad: 0.2199 (0.2172)  time: 0.6100  data: 0.0039  max mem: 57344
+train: [6]  [220/400]  eta: 0:01:50  lr: 0.000292  loss: 2.7309 (2.7459)  grad: 0.2162 (0.2166)  time: 0.6100  data: 0.0037  max mem: 57344
+train: [6]  [240/400]  eta: 0:01:37  lr: 0.000292  loss: 2.7491 (2.7463)  grad: 0.2118 (0.2166)  time: 0.6102  data: 0.0039  max mem: 57344
+train: [6]  [260/400]  eta: 0:01:25  lr: 0.000291  loss: 2.7674 (2.7481)  grad: 0.2111 (0.2165)  time: 0.6105  data: 0.0038  max mem: 57344
+train: [6]  [280/400]  eta: 0:01:13  lr: 0.000291  loss: 2.7647 (2.7482)  grad: 0.2127 (0.2159)  time: 0.6100  data: 0.0038  max mem: 57344
+train: [6]  [300/400]  eta: 0:01:01  lr: 0.000290  loss: 2.7239 (2.7492)  grad: 0.2135 (0.2164)  time: 0.6103  data: 0.0038  max mem: 57344
+train: [6]  [320/400]  eta: 0:00:48  lr: 0.000290  loss: 2.7621 (2.7511)  grad: 0.2135 (0.2165)  time: 0.6110  data: 0.0040  max mem: 57344
+train: [6]  [340/400]  eta: 0:00:36  lr: 0.000289  loss: 2.7621 (2.7503)  grad: 0.2211 (0.2171)  time: 0.6112  data: 0.0041  max mem: 57344
+train: [6]  [360/400]  eta: 0:00:24  lr: 0.000288  loss: 2.7221 (2.7491)  grad: 0.2252 (0.2175)  time: 0.6103  data: 0.0039  max mem: 57344
+train: [6]  [380/400]  eta: 0:00:12  lr: 0.000288  loss: 2.7221 (2.7504)  grad: 0.2227 (0.2178)  time: 0.6099  data: 0.0038  max mem: 57344
+train: [6]  [399/400]  eta: 0:00:00  lr: 0.000287  loss: 2.7313 (2.7489)  grad: 0.2209 (0.2178)  time: 0.6104  data: 0.0039  max mem: 57344
+train: [6] Total time: 0:04:04 (0.6120 s / it)
+train: [6] Summary: lr: 0.000287  loss: 2.7313 (2.7489)  grad: 0.2209 (0.2178)
+eval (validation): [6]  [ 0/85]  eta: 0:01:17    time: 0.9119  data: 0.5520  max mem: 57344
+eval (validation): [6]  [20/85]  eta: 0:00:25    time: 0.3707  data: 0.0035  max mem: 57344
+eval (validation): [6]  [40/85]  eta: 0:00:17    time: 0.3715  data: 0.0036  max mem: 57344
+eval (validation): [6]  [60/85]  eta: 0:00:09    time: 0.3718  data: 0.0037  max mem: 57344
+eval (validation): [6]  [80/85]  eta: 0:00:01    time: 0.3714  data: 0.0037  max mem: 57344
+eval (validation): [6]  [84/85]  eta: 0:00:00    time: 0.3654  data: 0.0036  max mem: 57344
+eval (validation): [6] Total time: 0:00:32 (0.3775 s / it)
+cv: [6]  best hparam: (1.2, 1.0) (025) ('025_lr1.2e+00_wd1.0e+00')  loss: 2.415  acc: 0.265  f1: 0.207
+saving checkpoint experiments/schaefer1000/output/schaefer1000/schaefer1000_lr3e-4_4/eval_v2/nsd_cococlip__patch__attn/checkpoint-last.pth
+train: [7]  [  0/400]  eta: 0:08:26  lr: nan  time: 1.2657  data: 0.6666  max mem: 57344
+train: [7]  [ 20/400]  eta: 0:04:03  lr: 0.000286  loss: 2.6793 (2.6768)  grad: 0.2149 (0.2158)  time: 0.6108  data: 0.0036  max mem: 57344
+train: [7]  [ 40/400]  eta: 0:03:45  lr: 0.000286  loss: 2.6859 (2.6990)  grad: 0.2186 (0.2186)  time: 0.6106  data: 0.0039  max mem: 57344
+train: [7]  [ 60/400]  eta: 0:03:31  lr: 0.000285  loss: 2.6879 (2.6977)  grad: 0.2208 (0.2203)  time: 0.6107  data: 0.0039  max mem: 57344
+train: [7]  [ 80/400]  eta: 0:03:18  lr: 0.000284  loss: 2.6977 (2.6947)  grad: 0.2180 (0.2206)  time: 0.6112  data: 0.0042  max mem: 57344
+train: [7]  [100/400]  eta: 0:03:05  lr: 0.000284  loss: 2.6977 (2.6959)  grad: 0.2185 (0.2204)  time: 0.6111  data: 0.0039  max mem: 57344
+train: [7]  [120/400]  eta: 0:02:52  lr: 0.000283  loss: 2.6811 (2.6919)  grad: 0.2228 (0.2226)  time: 0.6106  data: 0.0039  max mem: 57344
+train: [7]  [140/400]  eta: 0:02:40  lr: 0.000282  loss: 2.7067 (2.6964)  grad: 0.2293 (0.2240)  time: 0.6110  data: 0.0039  max mem: 57344
+train: [7]  [160/400]  eta: 0:02:27  lr: 0.000282  loss: 2.7112 (2.6984)  grad: 0.2161 (0.2230)  time: 0.6106  data: 0.0039  max mem: 57344
+train: [7]  [180/400]  eta: 0:02:15  lr: 0.000281  loss: 2.6947 (2.6950)  grad: 0.2107 (0.2218)  time: 0.6104  data: 0.0038  max mem: 57344
+train: [7]  [200/400]  eta: 0:02:02  lr: 0.000280  loss: 2.6542 (2.6916)  grad: 0.2073 (0.2209)  time: 0.6095  data: 0.0037  max mem: 57344
+train: [7]  [220/400]  eta: 0:01:50  lr: 0.000279  loss: 2.6664 (2.6918)  grad: 0.2156 (0.2213)  time: 0.6085  data: 0.0035  max mem: 57344
+train: [7]  [240/400]  eta: 0:01:38  lr: 0.000278  loss: 2.6720 (2.6892)  grad: 0.2182 (0.2209)  time: 0.6088  data: 0.0035  max mem: 57344
+train: [7]  [260/400]  eta: 0:01:25  lr: 0.000278  loss: 2.6899 (2.6931)  grad: 0.2232 (0.2218)  time: 0.6104  data: 0.0038  max mem: 57344
+train: [7]  [280/400]  eta: 0:01:13  lr: 0.000277  loss: 2.7408 (2.6976)  grad: 0.2286 (0.2225)  time: 0.6101  data: 0.0038  max mem: 57344
+train: [7]  [300/400]  eta: 0:01:01  lr: 0.000276  loss: 2.7147 (2.6977)  grad: 0.2388 (0.2237)  time: 0.6087  data: 0.0036  max mem: 57344
+train: [7]  [320/400]  eta: 0:00:48  lr: 0.000275  loss: 2.6860 (2.6983)  grad: 0.2438 (0.2250)  time: 0.6101  data: 0.0036  max mem: 57344
+train: [7]  [340/400]  eta: 0:00:36  lr: 0.000274  loss: 2.7272 (2.7018)  grad: 0.2438 (0.2260)  time: 0.6097  data: 0.0036  max mem: 57344
+train: [7]  [360/400]  eta: 0:00:24  lr: 0.000273  loss: 2.7331 (2.7030)  grad: 0.2329 (0.2260)  time: 0.6097  data: 0.0037  max mem: 57344
+train: [7]  [380/400]  eta: 0:00:12  lr: 0.000272  loss: 2.7547 (2.7058)  grad: 0.2222 (0.2260)  time: 0.6100  data: 0.0037  max mem: 57344
+train: [7]  [399/400]  eta: 0:00:00  lr: 0.000271  loss: 2.7880 (2.7105)  grad: 0.2220 (0.2261)  time: 0.6230  data: 0.0037  max mem: 57344
+train: [7] Total time: 0:04:05 (0.6127 s / it)
+train: [7] Summary: lr: 0.000271  loss: 2.7880 (2.7105)  grad: 0.2220 (0.2261)
+eval (validation): [7]  [ 0/85]  eta: 0:01:16    time: 0.8956  data: 0.5334  max mem: 57344
+eval (validation): [7]  [20/85]  eta: 0:00:25    time: 0.3708  data: 0.0030  max mem: 57344
+eval (validation): [7]  [40/85]  eta: 0:00:17    time: 0.3716  data: 0.0037  max mem: 57344
+eval (validation): [7]  [60/85]  eta: 0:00:09    time: 0.3712  data: 0.0035  max mem: 57344
+eval (validation): [7]  [80/85]  eta: 0:00:01    time: 0.3715  data: 0.0037  max mem: 57344
+eval (validation): [7]  [84/85]  eta: 0:00:00    time: 0.3649  data: 0.0036  max mem: 57344
+eval (validation): [7] Total time: 0:00:32 (0.3772 s / it)
+cv: [7]  best hparam: (1.9, 1.0) (028) ('028_lr1.9e+00_wd1.0e+00')  loss: 2.439  acc: 0.272  f1: 0.208
+saving checkpoint experiments/schaefer1000/output/schaefer1000/schaefer1000_lr3e-4_4/eval_v2/nsd_cococlip__patch__attn/checkpoint-last.pth
+saving best checkpoint experiments/schaefer1000/output/schaefer1000/schaefer1000_lr3e-4_4/eval_v2/nsd_cococlip__patch__attn/checkpoint-best.pth
+train: [8]  [  0/400]  eta: 0:08:18  lr: nan  time: 1.2473  data: 0.6493  max mem: 57344
+train: [8]  [ 20/400]  eta: 0:04:02  lr: 0.000270  loss: 2.6559 (2.6607)  grad: 0.2170 (0.2192)  time: 0.6084  data: 0.0030  max mem: 57344
+train: [8]  [ 40/400]  eta: 0:03:45  lr: 0.000270  loss: 2.6570 (2.6518)  grad: 0.2168 (0.2177)  time: 0.6108  data: 0.0039  max mem: 57344
+train: [8]  [ 60/400]  eta: 0:03:30  lr: 0.000269  loss: 2.6834 (2.6607)  grad: 0.2188 (0.2210)  time: 0.6097  data: 0.0038  max mem: 57344
+train: [8]  [ 80/400]  eta: 0:03:17  lr: 0.000268  loss: 2.6523 (2.6574)  grad: 0.2257 (0.2232)  time: 0.6098  data: 0.0038  max mem: 57344
+train: [8]  [100/400]  eta: 0:03:04  lr: 0.000267  loss: 2.6523 (2.6602)  grad: 0.2180 (0.2220)  time: 0.6095  data: 0.0037  max mem: 57344
+train: [8]  [120/400]  eta: 0:02:52  lr: 0.000266  loss: 2.6711 (2.6647)  grad: 0.2159 (0.2218)  time: 0.6110  data: 0.0038  max mem: 57344
+train: [8]  [140/400]  eta: 0:02:39  lr: 0.000265  loss: 2.6759 (2.6641)  grad: 0.2219 (0.2230)  time: 0.6096  data: 0.0038  max mem: 57344
+train: [8]  [160/400]  eta: 0:02:27  lr: 0.000264  loss: 2.6499 (2.6652)  grad: 0.2319 (0.2245)  time: 0.6101  data: 0.0037  max mem: 57344
+train: [8]  [180/400]  eta: 0:02:14  lr: 0.000263  loss: 2.6495 (2.6666)  grad: 0.2265 (0.2248)  time: 0.6097  data: 0.0038  max mem: 57344
+train: [8]  [200/400]  eta: 0:02:02  lr: 0.000262  loss: 2.6535 (2.6677)  grad: 0.2274 (0.2251)  time: 0.6098  data: 0.0037  max mem: 57344
+train: [8]  [220/400]  eta: 0:01:50  lr: 0.000260  loss: 2.6436 (2.6657)  grad: 0.2290 (0.2250)  time: 0.6098  data: 0.0037  max mem: 57344
+train: [8]  [240/400]  eta: 0:01:37  lr: 0.000259  loss: 2.6567 (2.6671)  grad: 0.2201 (0.2249)  time: 0.6098  data: 0.0038  max mem: 57344
+train: [8]  [260/400]  eta: 0:01:25  lr: 0.000258  loss: 2.6893 (2.6678)  grad: 0.2219 (0.2254)  time: 0.6098  data: 0.0037  max mem: 57344
+train: [8]  [280/400]  eta: 0:01:13  lr: 0.000257  loss: 2.6840 (2.6685)  grad: 0.2213 (0.2251)  time: 0.6096  data: 0.0038  max mem: 57344
+train: [8]  [300/400]  eta: 0:01:01  lr: 0.000256  loss: 2.6794 (2.6671)  grad: 0.2224 (0.2256)  time: 0.6100  data: 0.0038  max mem: 57344
+train: [8]  [320/400]  eta: 0:00:48  lr: 0.000255  loss: 2.6852 (2.6681)  grad: 0.2303 (0.2260)  time: 0.6097  data: 0.0037  max mem: 57344
+train: [8]  [340/400]  eta: 0:00:36  lr: 0.000254  loss: 2.6905 (2.6693)  grad: 0.2292 (0.2262)  time: 0.6097  data: 0.0037  max mem: 57344
+train: [8]  [360/400]  eta: 0:00:24  lr: 0.000253  loss: 2.6521 (2.6689)  grad: 0.2254 (0.2259)  time: 0.6091  data: 0.0036  max mem: 57344
+train: [8]  [380/400]  eta: 0:00:12  lr: 0.000252  loss: 2.6381 (2.6683)  grad: 0.2183 (0.2255)  time: 0.6083  data: 0.0034  max mem: 57344
+train: [8]  [399/400]  eta: 0:00:00  lr: 0.000250  loss: 2.6381 (2.6675)  grad: 0.2241 (0.2257)  time: 0.6089  data: 0.0034  max mem: 57344
+train: [8] Total time: 0:04:04 (0.6115 s / it)
+train: [8] Summary: lr: 0.000250  loss: 2.6381 (2.6675)  grad: 0.2241 (0.2257)
+eval (validation): [8]  [ 0/85]  eta: 0:01:26    time: 1.0179  data: 0.6584  max mem: 57344
+eval (validation): [8]  [20/85]  eta: 0:00:26    time: 0.3707  data: 0.0031  max mem: 57344
+eval (validation): [8]  [40/85]  eta: 0:00:17    time: 0.3723  data: 0.0038  max mem: 57344
+eval (validation): [8]  [60/85]  eta: 0:00:09    time: 0.3728  data: 0.0040  max mem: 57344
+eval (validation): [8]  [80/85]  eta: 0:00:01    time: 0.3728  data: 0.0037  max mem: 57344
+eval (validation): [8]  [84/85]  eta: 0:00:00    time: 0.3662  data: 0.0036  max mem: 57344
+eval (validation): [8] Total time: 0:00:32 (0.3796 s / it)
+cv: [8]  best hparam: (1.9, 1.0) (028) ('028_lr1.9e+00_wd1.0e+00')  loss: 2.426  acc: 0.279  f1: 0.213
+saving checkpoint experiments/schaefer1000/output/schaefer1000/schaefer1000_lr3e-4_4/eval_v2/nsd_cococlip__patch__attn/checkpoint-last.pth
+saving best checkpoint experiments/schaefer1000/output/schaefer1000/schaefer1000_lr3e-4_4/eval_v2/nsd_cococlip__patch__attn/checkpoint-best.pth
+train: [9]  [  0/400]  eta: 0:08:10  lr: nan  time: 1.2256  data: 0.6292  max mem: 57344
+train: [9]  [ 20/400]  eta: 0:04:01  lr: 0.000249  loss: 2.6476 (2.6434)  grad: 0.2114 (0.2149)  time: 0.6064  data: 0.0022  max mem: 57344
+train: [9]  [ 40/400]  eta: 0:03:44  lr: 0.000248  loss: 2.5957 (2.6220)  grad: 0.2177 (0.2172)  time: 0.6095  data: 0.0039  max mem: 57344
+train: [9]  [ 60/400]  eta: 0:03:30  lr: 0.000247  loss: 2.6168 (2.6364)  grad: 0.2200 (0.2176)  time: 0.6102  data: 0.0040  max mem: 57344
+train: [9]  [ 80/400]  eta: 0:03:17  lr: 0.000246  loss: 2.6864 (2.6467)  grad: 0.2220 (0.2195)  time: 0.6100  data: 0.0039  max mem: 57344
+train: [9]  [100/400]  eta: 0:03:04  lr: 0.000244  loss: 2.6563 (2.6453)  grad: 0.2249 (0.2208)  time: 0.6096  data: 0.0038  max mem: 57344
+train: [9]  [120/400]  eta: 0:02:52  lr: 0.000243  loss: 2.6323 (2.6422)  grad: 0.2201 (0.2203)  time: 0.6099  data: 0.0038  max mem: 57344
+train: [9]  [140/400]  eta: 0:02:39  lr: 0.000242  loss: 2.5768 (2.6327)  grad: 0.2204 (0.2210)  time: 0.6098  data: 0.0038  max mem: 57344
+train: [9]  [160/400]  eta: 0:02:27  lr: 0.000241  loss: 2.6420 (2.6389)  grad: 0.2246 (0.2212)  time: 0.6092  data: 0.0037  max mem: 57344
+train: [9]  [180/400]  eta: 0:02:14  lr: 0.000240  loss: 2.6747 (2.6439)  grad: 0.2175 (0.2211)  time: 0.6071  data: 0.0032  max mem: 57344
+train: [9]  [200/400]  eta: 0:02:02  lr: 0.000238  loss: 2.6379 (2.6421)  grad: 0.2172 (0.2207)  time: 0.6072  data: 0.0032  max mem: 57344
+train: [9]  [220/400]  eta: 0:01:50  lr: 0.000237  loss: 2.6379 (2.6457)  grad: 0.2172 (0.2208)  time: 0.6071  data: 0.0032  max mem: 57344
+train: [9]  [240/400]  eta: 0:01:37  lr: 0.000236  loss: 2.6510 (2.6466)  grad: 0.2225 (0.2214)  time: 0.6068  data: 0.0032  max mem: 57344
+train: [9]  [260/400]  eta: 0:01:25  lr: 0.000234  loss: 2.6713 (2.6470)  grad: 0.2213 (0.2212)  time: 0.6068  data: 0.0032  max mem: 57344
+train: [9]  [280/400]  eta: 0:01:13  lr: 0.000233  loss: 2.5929 (2.6445)  grad: 0.2099 (0.2200)  time: 0.6070  data: 0.0032  max mem: 57344
+train: [9]  [300/400]  eta: 0:01:01  lr: 0.000232  loss: 2.5929 (2.6437)  grad: 0.2144 (0.2206)  time: 0.6072  data: 0.0032  max mem: 57344
+train: [9]  [320/400]  eta: 0:00:48  lr: 0.000230  loss: 2.6403 (2.6439)  grad: 0.2257 (0.2209)  time: 0.6071  data: 0.0032  max mem: 57344
+train: [9]  [340/400]  eta: 0:00:36  lr: 0.000229  loss: 2.6559 (2.6459)  grad: 0.2227 (0.2211)  time: 0.6069  data: 0.0032  max mem: 57344
+train: [9]  [360/400]  eta: 0:00:24  lr: 0.000228  loss: 2.6439 (2.6427)  grad: 0.2173 (0.2211)  time: 0.6072  data: 0.0031  max mem: 57344
+train: [9]  [380/400]  eta: 0:00:12  lr: 0.000226  loss: 2.5990 (2.6403)  grad: 0.2169 (0.2209)  time: 0.6067  data: 0.0032  max mem: 57344
+train: [9]  [399/400]  eta: 0:00:00  lr: 0.000225  loss: 2.5998 (2.6402)  grad: 0.2208 (0.2210)  time: 0.6068  data: 0.0032  max mem: 57344
+train: [9] Total time: 0:04:03 (0.6097 s / it)
+train: [9] Summary: lr: 0.000225  loss: 2.5998 (2.6402)  grad: 0.2208 (0.2210)
+eval (validation): [9]  [ 0/85]  eta: 0:01:01    time: 0.7266  data: 0.3661  max mem: 57344
+eval (validation): [9]  [20/85]  eta: 0:00:25    time: 0.3700  data: 0.0030  max mem: 57344
+eval (validation): [9]  [40/85]  eta: 0:00:17    time: 0.3704  data: 0.0033  max mem: 57344
+eval (validation): [9]  [60/85]  eta: 0:00:09    time: 0.3712  data: 0.0035  max mem: 57344
+eval (validation): [9]  [80/85]  eta: 0:00:01    time: 0.3700  data: 0.0033  max mem: 57344
+eval (validation): [9]  [84/85]  eta: 0:00:00    time: 0.3638  data: 0.0032  max mem: 57344
+eval (validation): [9] Total time: 0:00:31 (0.3743 s / it)
+cv: [9]  best hparam: (5.1, 1.0) (034) ('034_lr5.1e+00_wd1.0e+00')  loss: 2.498  acc: 0.274  f1: 0.215
+saving checkpoint experiments/schaefer1000/output/schaefer1000/schaefer1000_lr3e-4_4/eval_v2/nsd_cococlip__patch__attn/checkpoint-last.pth
+train: [10]  [  0/400]  eta: 0:07:02  lr: nan  time: 1.0562  data: 0.4600  max mem: 57344
+train: [10]  [ 20/400]  eta: 0:03:58  lr: 0.000224  loss: 2.6105 (2.5993)  grad: 0.2158 (0.2140)  time: 0.6073  data: 0.0030  max mem: 57344
+train: [10]  [ 40/400]  eta: 0:03:43  lr: 0.000222  loss: 2.5548 (2.5821)  grad: 0.2204 (0.2222)  time: 0.6102  data: 0.0036  max mem: 57344
+train: [10]  [ 60/400]  eta: 0:03:29  lr: 0.000221  loss: 2.5810 (2.5841)  grad: 0.2281 (0.2248)  time: 0.6106  data: 0.0040  max mem: 57344
+train: [10]  [ 80/400]  eta: 0:03:17  lr: 0.000220  loss: 2.5897 (2.5924)  grad: 0.2252 (0.2236)  time: 0.6135  data: 0.0042  max mem: 57344
+train: [10]  [100/400]  eta: 0:03:04  lr: 0.000218  loss: 2.5835 (2.5894)  grad: 0.2235 (0.2229)  time: 0.6134  data: 0.0043  max mem: 57344
+train: [10]  [120/400]  eta: 0:02:52  lr: 0.000217  loss: 2.5840 (2.5992)  grad: 0.2137 (0.2219)  time: 0.6185  data: 0.0046  max mem: 57344
+train: [10]  [140/400]  eta: 0:02:39  lr: 0.000215  loss: 2.6152 (2.5940)  grad: 0.2164 (0.2226)  time: 0.6074  data: 0.0033  max mem: 57344
+train: [10]  [160/400]  eta: 0:02:27  lr: 0.000214  loss: 2.5692 (2.5945)  grad: 0.2310 (0.2238)  time: 0.6091  data: 0.0036  max mem: 57344
+train: [10]  [180/400]  eta: 0:02:14  lr: 0.000213  loss: 2.5788 (2.5960)  grad: 0.2340 (0.2244)  time: 0.6072  data: 0.0033  max mem: 57344
+train: [10]  [200/400]  eta: 0:02:02  lr: 0.000211  loss: 2.5809 (2.5929)  grad: 0.2271 (0.2248)  time: 0.6082  data: 0.0035  max mem: 57344
+train: [10]  [220/400]  eta: 0:01:50  lr: 0.000210  loss: 2.5836 (2.5935)  grad: 0.2239 (0.2249)  time: 0.6080  data: 0.0036  max mem: 57344
+train: [10]  [240/400]  eta: 0:01:37  lr: 0.000208  loss: 2.5732 (2.5922)  grad: 0.2222 (0.2247)  time: 0.6069  data: 0.0033  max mem: 57344
+train: [10]  [260/400]  eta: 0:01:25  lr: 0.000207  loss: 2.5958 (2.5945)  grad: 0.2222 (0.2249)  time: 0.6071  data: 0.0033  max mem: 57344
+train: [10]  [280/400]  eta: 0:01:13  lr: 0.000205  loss: 2.6068 (2.5981)  grad: 0.2194 (0.2243)  time: 0.6069  data: 0.0032  max mem: 57344
+train: [10]  [300/400]  eta: 0:01:01  lr: 0.000204  loss: 2.6354 (2.6002)  grad: 0.2122 (0.2239)  time: 0.6068  data: 0.0033  max mem: 57344
+train: [10]  [320/400]  eta: 0:00:48  lr: 0.000202  loss: 2.6216 (2.6011)  grad: 0.2121 (0.2234)  time: 0.6094  data: 0.0036  max mem: 57344
+train: [10]  [340/400]  eta: 0:00:36  lr: 0.000201  loss: 2.6101 (2.6022)  grad: 0.2150 (0.2231)  time: 0.6110  data: 0.0040  max mem: 57344
+train: [10]  [360/400]  eta: 0:00:24  lr: 0.000199  loss: 2.6132 (2.6008)  grad: 0.2129 (0.2225)  time: 0.6090  data: 0.0037  max mem: 57344
+train: [10]  [380/400]  eta: 0:00:12  lr: 0.000198  loss: 2.6132 (2.6019)  grad: 0.2170 (0.2226)  time: 0.6106  data: 0.0040  max mem: 57344
+train: [10]  [399/400]  eta: 0:00:00  lr: 0.000196  loss: 2.5991 (2.6015)  grad: 0.2191 (0.2224)  time: 0.6112  data: 0.0037  max mem: 57344
+train: [10] Total time: 0:04:04 (0.6110 s / it)
+train: [10] Summary: lr: 0.000196  loss: 2.5991 (2.6015)  grad: 0.2191 (0.2224)
+eval (validation): [10]  [ 0/85]  eta: 0:01:25    time: 1.0062  data: 0.6442  max mem: 57344
+eval (validation): [10]  [20/85]  eta: 0:00:26    time: 0.3708  data: 0.0038  max mem: 57344
+eval (validation): [10]  [40/85]  eta: 0:00:17    time: 0.3718  data: 0.0038  max mem: 57344
+eval (validation): [10]  [60/85]  eta: 0:00:09    time: 0.3719  data: 0.0037  max mem: 57344
+eval (validation): [10]  [80/85]  eta: 0:00:01    time: 0.3713  data: 0.0036  max mem: 57344
+eval (validation): [10]  [84/85]  eta: 0:00:00    time: 0.3648  data: 0.0036  max mem: 57344
+eval (validation): [10] Total time: 0:00:32 (0.3786 s / it)
+cv: [10]  best hparam: (2.7, 1.0) (030) ('030_lr2.7e+00_wd1.0e+00')  loss: 2.434  acc: 0.277  f1: 0.232
+saving checkpoint experiments/schaefer1000/output/schaefer1000/schaefer1000_lr3e-4_4/eval_v2/nsd_cococlip__patch__attn/checkpoint-last.pth
+train: [11]  [  0/400]  eta: 0:08:53  lr: nan  time: 1.3335  data: 0.7359  max mem: 57344
+train: [11]  [ 20/400]  eta: 0:04:04  lr: 0.000195  loss: 2.5596 (2.5514)  grad: 0.2189 (0.2298)  time: 0.6096  data: 0.0030  max mem: 57344
+train: [11]  [ 40/400]  eta: 0:03:46  lr: 0.000193  loss: 2.5490 (2.5499)  grad: 0.2162 (0.2244)  time: 0.6116  data: 0.0044  max mem: 57344
+train: [11]  [ 60/400]  eta: 0:03:31  lr: 0.000192  loss: 2.5743 (2.5652)  grad: 0.2152 (0.2212)  time: 0.6113  data: 0.0043  max mem: 57344
+train: [11]  [ 80/400]  eta: 0:03:18  lr: 0.000190  loss: 2.5849 (2.5700)  grad: 0.2133 (0.2200)  time: 0.6103  data: 0.0039  max mem: 57344
+train: [11]  [100/400]  eta: 0:03:05  lr: 0.000189  loss: 2.6016 (2.5848)  grad: 0.2156 (0.2208)  time: 0.6096  data: 0.0038  max mem: 57344
+train: [11]  [120/400]  eta: 0:02:52  lr: 0.000187  loss: 2.6166 (2.5913)  grad: 0.2213 (0.2217)  time: 0.6095  data: 0.0037  max mem: 57344
+train: [11]  [140/400]  eta: 0:02:39  lr: 0.000186  loss: 2.6109 (2.5872)  grad: 0.2120 (0.2196)  time: 0.6088  data: 0.0036  max mem: 57344
+train: [11]  [160/400]  eta: 0:02:27  lr: 0.000184  loss: 2.5720 (2.5834)  grad: 0.2047 (0.2187)  time: 0.6084  data: 0.0036  max mem: 57344
+train: [11]  [180/400]  eta: 0:02:15  lr: 0.000183  loss: 2.5273 (2.5787)  grad: 0.2168 (0.2190)  time: 0.6086  data: 0.0036  max mem: 57344
+train: [11]  [200/400]  eta: 0:02:02  lr: 0.000181  loss: 2.5265 (2.5791)  grad: 0.2222 (0.2196)  time: 0.6091  data: 0.0037  max mem: 57344
+train: [11]  [220/400]  eta: 0:01:50  lr: 0.000180  loss: 2.6080 (2.5801)  grad: 0.2164 (0.2194)  time: 0.6085  data: 0.0036  max mem: 57344
+train: [11]  [240/400]  eta: 0:01:37  lr: 0.000178  loss: 2.6088 (2.5818)  grad: 0.2186 (0.2206)  time: 0.6079  data: 0.0035  max mem: 57344
+train: [11]  [260/400]  eta: 0:01:25  lr: 0.000177  loss: 2.6088 (2.5813)  grad: 0.2256 (0.2210)  time: 0.6079  data: 0.0035  max mem: 57344
+train: [11]  [280/400]  eta: 0:01:13  lr: 0.000175  loss: 2.5699 (2.5780)  grad: 0.2256 (0.2210)  time: 0.6075  data: 0.0035  max mem: 57344
+train: [11]  [300/400]  eta: 0:01:01  lr: 0.000174  loss: 2.5775 (2.5776)  grad: 0.2255 (0.2212)  time: 0.6076  data: 0.0034  max mem: 57344
+train: [11]  [320/400]  eta: 0:00:48  lr: 0.000172  loss: 2.5775 (2.5761)  grad: 0.2187 (0.2210)  time: 0.6074  data: 0.0034  max mem: 57344
+train: [11]  [340/400]  eta: 0:00:36  lr: 0.000170  loss: 2.5450 (2.5783)  grad: 0.2187 (0.2208)  time: 0.6075  data: 0.0034  max mem: 57344
+train: [11]  [360/400]  eta: 0:00:24  lr: 0.000169  loss: 2.5517 (2.5766)  grad: 0.2201 (0.2209)  time: 0.6068  data: 0.0034  max mem: 57344
+train: [11]  [380/400]  eta: 0:00:12  lr: 0.000167  loss: 2.5611 (2.5775)  grad: 0.2262 (0.2209)  time: 0.6078  data: 0.0034  max mem: 57344
+train: [11]  [399/400]  eta: 0:00:00  lr: 0.000166  loss: 2.6081 (2.5789)  grad: 0.2234 (0.2210)  time: 0.6094  data: 0.0038  max mem: 57344
+train: [11] Total time: 0:04:04 (0.6108 s / it)
+train: [11] Summary: lr: 0.000166  loss: 2.6081 (2.5789)  grad: 0.2234 (0.2210)
+eval (validation): [11]  [ 0/85]  eta: 0:01:20    time: 0.9439  data: 0.5804  max mem: 57344
+eval (validation): [11]  [20/85]  eta: 0:00:25    time: 0.3712  data: 0.0038  max mem: 57344
+eval (validation): [11]  [40/85]  eta: 0:00:17    time: 0.3716  data: 0.0035  max mem: 57344
+eval (validation): [11]  [60/85]  eta: 0:00:09    time: 0.3715  data: 0.0037  max mem: 57344
+eval (validation): [11]  [80/85]  eta: 0:00:01    time: 0.3707  data: 0.0034  max mem: 57344
+eval (validation): [11]  [84/85]  eta: 0:00:00    time: 0.3644  data: 0.0034  max mem: 57344
+eval (validation): [11] Total time: 0:00:32 (0.3776 s / it)
+cv: [11]  best hparam: (0.85, 1.0) (023) ('023_lr8.5e-01_wd1.0e+00')  loss: 2.364  acc: 0.281  f1: 0.226
+saving checkpoint experiments/schaefer1000/output/schaefer1000/schaefer1000_lr3e-4_4/eval_v2/nsd_cococlip__patch__attn/checkpoint-last.pth
+saving best checkpoint experiments/schaefer1000/output/schaefer1000/schaefer1000_lr3e-4_4/eval_v2/nsd_cococlip__patch__attn/checkpoint-best.pth
+train: [12]  [  0/400]  eta: 0:07:58  lr: nan  time: 1.1952  data: 0.5974  max mem: 57344
+train: [12]  [ 20/400]  eta: 0:04:01  lr: 0.000164  loss: 2.5516 (2.5913)  grad: 0.2161 (0.2180)  time: 0.6065  data: 0.0025  max mem: 57344
+train: [12]  [ 40/400]  eta: 0:03:43  lr: 0.000163  loss: 2.5516 (2.5693)  grad: 0.2145 (0.2158)  time: 0.6076  data: 0.0035  max mem: 57344
+train: [12]  [ 60/400]  eta: 0:03:29  lr: 0.000161  loss: 2.5654 (2.5649)  grad: 0.2122 (0.2144)  time: 0.6082  data: 0.0035  max mem: 57344
+train: [12]  [ 80/400]  eta: 0:03:16  lr: 0.000160  loss: 2.5572 (2.5575)  grad: 0.2129 (0.2156)  time: 0.6098  data: 0.0038  max mem: 57344
+train: [12]  [100/400]  eta: 0:03:04  lr: 0.000158  loss: 2.5306 (2.5558)  grad: 0.2182 (0.2171)  time: 0.6100  data: 0.0038  max mem: 57344
+train: [12]  [120/400]  eta: 0:02:51  lr: 0.000156  loss: 2.5844 (2.5644)  grad: 0.2277 (0.2199)  time: 0.6086  data: 0.0037  max mem: 57344
+train: [12]  [140/400]  eta: 0:02:39  lr: 0.000155  loss: 2.5851 (2.5644)  grad: 0.2238 (0.2205)  time: 0.6098  data: 0.0040  max mem: 57344
+train: [12]  [160/400]  eta: 0:02:26  lr: 0.000153  loss: 2.5260 (2.5587)  grad: 0.2194 (0.2206)  time: 0.6103  data: 0.0040  max mem: 57344
+train: [12]  [180/400]  eta: 0:02:14  lr: 0.000152  loss: 2.5193 (2.5566)  grad: 0.2256 (0.2216)  time: 0.6104  data: 0.0039  max mem: 57344
+train: [12]  [200/400]  eta: 0:02:02  lr: 0.000150  loss: 2.5271 (2.5586)  grad: 0.2316 (0.2229)  time: 0.6095  data: 0.0037  max mem: 57344
+train: [12]  [220/400]  eta: 0:01:50  lr: 0.000149  loss: 2.5569 (2.5542)  grad: 0.2290 (0.2231)  time: 0.6086  data: 0.0036  max mem: 57344
+train: [12]  [240/400]  eta: 0:01:37  lr: 0.000147  loss: 2.4961 (2.5531)  grad: 0.2158 (0.2221)  time: 0.6089  data: 0.0036  max mem: 57344
+train: [12]  [260/400]  eta: 0:01:25  lr: 0.000145  loss: 2.5078 (2.5512)  grad: 0.2102 (0.2211)  time: 0.6087  data: 0.0036  max mem: 57344
+train: [12]  [280/400]  eta: 0:01:13  lr: 0.000144  loss: 2.5228 (2.5521)  grad: 0.2104 (0.2211)  time: 0.6088  data: 0.0036  max mem: 57344
+train: [12]  [300/400]  eta: 0:01:01  lr: 0.000142  loss: 2.5543 (2.5538)  grad: 0.2189 (0.2210)  time: 0.6095  data: 0.0036  max mem: 57344
+train: [12]  [320/400]  eta: 0:00:48  lr: 0.000141  loss: 2.5473 (2.5525)  grad: 0.2200 (0.2209)  time: 0.6088  data: 0.0036  max mem: 57344
+train: [12]  [340/400]  eta: 0:00:36  lr: 0.000139  loss: 2.5370 (2.5532)  grad: 0.2200 (0.2208)  time: 0.6085  data: 0.0035  max mem: 57344
+train: [12]  [360/400]  eta: 0:00:24  lr: 0.000138  loss: 2.5296 (2.5509)  grad: 0.2209 (0.2210)  time: 0.6075  data: 0.0033  max mem: 57344
+train: [12]  [380/400]  eta: 0:00:12  lr: 0.000136  loss: 2.5165 (2.5520)  grad: 0.2210 (0.2211)  time: 0.6079  data: 0.0033  max mem: 57344
+train: [12]  [399/400]  eta: 0:00:00  lr: 0.000134  loss: 2.5276 (2.5506)  grad: 0.2168 (0.2207)  time: 0.6078  data: 0.0034  max mem: 57344
+train: [12] Total time: 0:04:04 (0.6106 s / it)
+train: [12] Summary: lr: 0.000134  loss: 2.5276 (2.5506)  grad: 0.2168 (0.2207)
+eval (validation): [12]  [ 0/85]  eta: 0:01:13    time: 0.8658  data: 0.5056  max mem: 57344
+eval (validation): [12]  [20/85]  eta: 0:00:25    time: 0.3690  data: 0.0024  max mem: 57344
+eval (validation): [12]  [40/85]  eta: 0:00:17    time: 0.3699  data: 0.0031  max mem: 57344
+eval (validation): [12]  [60/85]  eta: 0:00:09    time: 0.3705  data: 0.0031  max mem: 57344
+eval (validation): [12]  [80/85]  eta: 0:00:01    time: 0.3704  data: 0.0031  max mem: 57344
+eval (validation): [12]  [84/85]  eta: 0:00:00    time: 0.3638  data: 0.0031  max mem: 57344
+eval (validation): [12] Total time: 0:00:31 (0.3755 s / it)
+cv: [12]  best hparam: (3.1, 1.0) (031) ('031_lr3.1e+00_wd1.0e+00')  loss: 2.478  acc: 0.277  f1: 0.221
+saving checkpoint experiments/schaefer1000/output/schaefer1000/schaefer1000_lr3e-4_4/eval_v2/nsd_cococlip__patch__attn/checkpoint-last.pth
+train: [13]  [  0/400]  eta: 0:08:58  lr: nan  time: 1.3453  data: 0.7480  max mem: 57344
+train: [13]  [ 20/400]  eta: 0:04:04  lr: 0.000133  loss: 2.4752 (2.5091)  grad: 0.2224 (0.2222)  time: 0.6087  data: 0.0029  max mem: 57344
+train: [13]  [ 40/400]  eta: 0:03:45  lr: 0.000131  loss: 2.4991 (2.5041)  grad: 0.2213 (0.2194)  time: 0.6097  data: 0.0038  max mem: 57344
+train: [13]  [ 60/400]  eta: 0:03:31  lr: 0.000130  loss: 2.4991 (2.5039)  grad: 0.2132 (0.2175)  time: 0.6084  data: 0.0035  max mem: 57344
+train: [13]  [ 80/400]  eta: 0:03:17  lr: 0.000128  loss: 2.4566 (2.5094)  grad: 0.2109 (0.2167)  time: 0.6090  data: 0.0035  max mem: 57344
+train: [13]  [100/400]  eta: 0:03:04  lr: 0.000127  loss: 2.5312 (2.5150)  grad: 0.2101 (0.2160)  time: 0.6090  data: 0.0035  max mem: 57344
+train: [13]  [120/400]  eta: 0:02:52  lr: 0.000125  loss: 2.5563 (2.5267)  grad: 0.2132 (0.2179)  time: 0.6089  data: 0.0036  max mem: 57344
+train: [13]  [140/400]  eta: 0:02:39  lr: 0.000124  loss: 2.5518 (2.5287)  grad: 0.2132 (0.2174)  time: 0.6097  data: 0.0037  max mem: 57344
+train: [13]  [160/400]  eta: 0:02:27  lr: 0.000122  loss: 2.5158 (2.5342)  grad: 0.2149 (0.2177)  time: 0.6099  data: 0.0039  max mem: 57344
+train: [13]  [180/400]  eta: 0:02:14  lr: 0.000120  loss: 2.5406 (2.5340)  grad: 0.2180 (0.2180)  time: 0.6096  data: 0.0039  max mem: 57344
+train: [13]  [200/400]  eta: 0:02:02  lr: 0.000119  loss: 2.5361 (2.5302)  grad: 0.2192 (0.2185)  time: 0.6104  data: 0.0039  max mem: 57344
+train: [13]  [220/400]  eta: 0:01:50  lr: 0.000117  loss: 2.5056 (2.5275)  grad: 0.2176 (0.2187)  time: 0.6090  data: 0.0038  max mem: 57344
+train: [13]  [240/400]  eta: 0:01:37  lr: 0.000116  loss: 2.4796 (2.5237)  grad: 0.2176 (0.2182)  time: 0.6088  data: 0.0035  max mem: 57344
+train: [13]  [260/400]  eta: 0:01:25  lr: 0.000114  loss: 2.4818 (2.5225)  grad: 0.2105 (0.2178)  time: 0.6084  data: 0.0035  max mem: 57344
+train: [13]  [280/400]  eta: 0:01:13  lr: 0.000113  loss: 2.4877 (2.5213)  grad: 0.2202 (0.2184)  time: 0.6084  data: 0.0035  max mem: 57344
+train: [13]  [300/400]  eta: 0:01:01  lr: 0.000111  loss: 2.5011 (2.5221)  grad: 0.2208 (0.2186)  time: 0.6085  data: 0.0036  max mem: 57344
+train: [13]  [320/400]  eta: 0:00:48  lr: 0.000110  loss: 2.5215 (2.5211)  grad: 0.2164 (0.2190)  time: 0.6088  data: 0.0035  max mem: 57344
+train: [13]  [340/400]  eta: 0:00:36  lr: 0.000108  loss: 2.5102 (2.5230)  grad: 0.2214 (0.2193)  time: 0.6086  data: 0.0035  max mem: 57344
+train: [13]  [360/400]  eta: 0:00:24  lr: 0.000107  loss: 2.5623 (2.5247)  grad: 0.2201 (0.2198)  time: 0.6085  data: 0.0035  max mem: 57344
+train: [13]  [380/400]  eta: 0:00:12  lr: 0.000105  loss: 2.5315 (2.5247)  grad: 0.2218 (0.2201)  time: 0.6072  data: 0.0033  max mem: 57344
+train: [13]  [399/400]  eta: 0:00:00  lr: 0.000104  loss: 2.5210 (2.5254)  grad: 0.2233 (0.2202)  time: 0.6072  data: 0.0033  max mem: 57344
+train: [13] Total time: 0:04:04 (0.6109 s / it)
+train: [13] Summary: lr: 0.000104  loss: 2.5210 (2.5254)  grad: 0.2233 (0.2202)
+eval (validation): [13]  [ 0/85]  eta: 0:01:14    time: 0.8713  data: 0.5106  max mem: 57344
+eval (validation): [13]  [20/85]  eta: 0:00:25    time: 0.3697  data: 0.0027  max mem: 57344
+eval (validation): [13]  [40/85]  eta: 0:00:17    time: 0.3709  data: 0.0034  max mem: 57344
+eval (validation): [13]  [60/85]  eta: 0:00:09    time: 0.3707  data: 0.0033  max mem: 57344
+eval (validation): [13]  [80/85]  eta: 0:00:01    time: 0.3708  data: 0.0034  max mem: 57344
+eval (validation): [13]  [84/85]  eta: 0:00:00    time: 0.3645  data: 0.0033  max mem: 57344
+eval (validation): [13] Total time: 0:00:31 (0.3762 s / it)
+cv: [13]  best hparam: (1.4, 1.0) (026) ('026_lr1.4e+00_wd1.0e+00')  loss: 2.414  acc: 0.282  f1: 0.232
+saving checkpoint experiments/schaefer1000/output/schaefer1000/schaefer1000_lr3e-4_4/eval_v2/nsd_cococlip__patch__attn/checkpoint-last.pth
+saving best checkpoint experiments/schaefer1000/output/schaefer1000/schaefer1000_lr3e-4_4/eval_v2/nsd_cococlip__patch__attn/checkpoint-best.pth
+train: [14]  [  0/400]  eta: 0:08:28  lr: nan  time: 1.2719  data: 0.6753  max mem: 57344
+train: [14]  [ 20/400]  eta: 0:04:03  lr: 0.000102  loss: 2.4809 (2.4819)  grad: 0.2195 (0.2171)  time: 0.6083  data: 0.0025  max mem: 57344
+train: [14]  [ 40/400]  eta: 0:03:44  lr: 0.000101  loss: 2.4986 (2.5076)  grad: 0.2173 (0.2175)  time: 0.6071  data: 0.0032  max mem: 57344
+train: [14]  [ 60/400]  eta: 0:03:30  lr: 0.000099  loss: 2.5328 (2.5024)  grad: 0.2089 (0.2134)  time: 0.6087  data: 0.0037  max mem: 57344
+train: [14]  [ 80/400]  eta: 0:03:17  lr: 0.000098  loss: 2.4595 (2.5024)  grad: 0.2061 (0.2149)  time: 0.6102  data: 0.0039  max mem: 57344
+train: [14]  [100/400]  eta: 0:03:04  lr: 0.000096  loss: 2.4820 (2.4990)  grad: 0.2198 (0.2165)  time: 0.6088  data: 0.0037  max mem: 57344
+train: [14]  [120/400]  eta: 0:02:51  lr: 0.000095  loss: 2.4758 (2.4931)  grad: 0.2181 (0.2154)  time: 0.6079  data: 0.0035  max mem: 57344
+train: [14]  [140/400]  eta: 0:02:39  lr: 0.000093  loss: 2.4746 (2.4927)  grad: 0.2144 (0.2162)  time: 0.6074  data: 0.0034  max mem: 57344
+train: [14]  [160/400]  eta: 0:02:26  lr: 0.000092  loss: 2.5613 (2.5021)  grad: 0.2171 (0.2163)  time: 0.6082  data: 0.0036  max mem: 57344
+train: [14]  [180/400]  eta: 0:02:14  lr: 0.000090  loss: 2.5115 (2.4986)  grad: 0.2191 (0.2168)  time: 0.6083  data: 0.0036  max mem: 57344
+train: [14]  [200/400]  eta: 0:02:02  lr: 0.000089  loss: 2.4735 (2.5005)  grad: 0.2165 (0.2165)  time: 0.6086  data: 0.0035  max mem: 57344
+train: [14]  [220/400]  eta: 0:01:50  lr: 0.000088  loss: 2.4735 (2.5016)  grad: 0.2107 (0.2160)  time: 0.6108  data: 0.0040  max mem: 57344
+train: [14]  [240/400]  eta: 0:01:37  lr: 0.000086  loss: 2.5305 (2.5049)  grad: 0.2122 (0.2163)  time: 0.6105  data: 0.0040  max mem: 57344
+train: [14]  [260/400]  eta: 0:01:25  lr: 0.000085  loss: 2.5002 (2.5043)  grad: 0.2171 (0.2167)  time: 0.6105  data: 0.0040  max mem: 57344
+train: [14]  [280/400]  eta: 0:01:13  lr: 0.000083  loss: 2.5002 (2.5057)  grad: 0.2171 (0.2169)  time: 0.6098  data: 0.0038  max mem: 57344
+train: [14]  [300/400]  eta: 0:01:01  lr: 0.000082  loss: 2.5163 (2.5067)  grad: 0.2164 (0.2172)  time: 0.6097  data: 0.0038  max mem: 57344
+train: [14]  [320/400]  eta: 0:00:48  lr: 0.000081  loss: 2.4989 (2.5052)  grad: 0.2139 (0.2170)  time: 0.6094  data: 0.0035  max mem: 57344
+train: [14]  [340/400]  eta: 0:00:36  lr: 0.000079  loss: 2.4772 (2.5037)  grad: 0.2099 (0.2164)  time: 0.6082  data: 0.0035  max mem: 57344
+train: [14]  [360/400]  eta: 0:00:24  lr: 0.000078  loss: 2.4863 (2.5038)  grad: 0.2071 (0.2164)  time: 0.6088  data: 0.0035  max mem: 57344
+train: [14]  [380/400]  eta: 0:00:12  lr: 0.000076  loss: 2.5025 (2.5029)  grad: 0.2099 (0.2165)  time: 0.6087  data: 0.0034  max mem: 57344
+train: [14]  [399/400]  eta: 0:00:00  lr: 0.000075  loss: 2.4663 (2.5020)  grad: 0.2136 (0.2164)  time: 0.6087  data: 0.0035  max mem: 57344
+train: [14] Total time: 0:04:04 (0.6109 s / it)
+train: [14] Summary: lr: 0.000075  loss: 2.4663 (2.5020)  grad: 0.2136 (0.2164)
+eval (validation): [14]  [ 0/85]  eta: 0:01:21    time: 0.9632  data: 0.6051  max mem: 57344
+eval (validation): [14]  [20/85]  eta: 0:00:25    time: 0.3704  data: 0.0031  max mem: 57344
+eval (validation): [14]  [40/85]  eta: 0:00:17    time: 0.3702  data: 0.0030  max mem: 57344
+eval (validation): [14]  [60/85]  eta: 0:00:09    time: 0.3706  data: 0.0033  max mem: 57344
+eval (validation): [14]  [80/85]  eta: 0:00:01    time: 0.3700  data: 0.0031  max mem: 57344
+eval (validation): [14]  [84/85]  eta: 0:00:00    time: 0.3638  data: 0.0031  max mem: 57344
+eval (validation): [14] Total time: 0:00:32 (0.3770 s / it)
+cv: [14]  best hparam: (1.4, 1.0) (026) ('026_lr1.4e+00_wd1.0e+00')  loss: 2.392  acc: 0.285  f1: 0.237
+saving checkpoint experiments/schaefer1000/output/schaefer1000/schaefer1000_lr3e-4_4/eval_v2/nsd_cococlip__patch__attn/checkpoint-last.pth
+saving best checkpoint experiments/schaefer1000/output/schaefer1000/schaefer1000_lr3e-4_4/eval_v2/nsd_cococlip__patch__attn/checkpoint-best.pth
+train: [15]  [  0/400]  eta: 0:07:13  lr: nan  time: 1.0844  data: 0.4886  max mem: 57344
+train: [15]  [ 20/400]  eta: 0:03:59  lr: 0.000074  loss: 2.4689 (2.4764)  grad: 0.2153 (0.2208)  time: 0.6067  data: 0.0030  max mem: 57344
+train: [15]  [ 40/400]  eta: 0:03:42  lr: 0.000072  loss: 2.4527 (2.4547)  grad: 0.2158 (0.2195)  time: 0.6064  data: 0.0033  max mem: 57344
+train: [15]  [ 60/400]  eta: 0:03:28  lr: 0.000071  loss: 2.4527 (2.4673)  grad: 0.2209 (0.2205)  time: 0.6066  data: 0.0033  max mem: 57344
+train: [15]  [ 80/400]  eta: 0:03:15  lr: 0.000070  loss: 2.5012 (2.4731)  grad: 0.2223 (0.2207)  time: 0.6066  data: 0.0033  max mem: 57344
+train: [15]  [100/400]  eta: 0:03:03  lr: 0.000068  loss: 2.4723 (2.4738)  grad: 0.2201 (0.2214)  time: 0.6068  data: 0.0033  max mem: 57344
+train: [15]  [120/400]  eta: 0:02:50  lr: 0.000067  loss: 2.4799 (2.4792)  grad: 0.2191 (0.2214)  time: 0.6065  data: 0.0033  max mem: 57344
+train: [15]  [140/400]  eta: 0:02:38  lr: 0.000066  loss: 2.5055 (2.4833)  grad: 0.2246 (0.2223)  time: 0.6091  data: 0.0037  max mem: 57344
+train: [15]  [160/400]  eta: 0:02:26  lr: 0.000064  loss: 2.4786 (2.4814)  grad: 0.2235 (0.2218)  time: 0.6092  data: 0.0037  max mem: 57344
+train: [15]  [180/400]  eta: 0:02:14  lr: 0.000063  loss: 2.4748 (2.4799)  grad: 0.2101 (0.2217)  time: 0.6076  data: 0.0034  max mem: 57344
+train: [15]  [200/400]  eta: 0:02:01  lr: 0.000062  loss: 2.4803 (2.4816)  grad: 0.2235 (0.2222)  time: 0.6078  data: 0.0034  max mem: 57344
+train: [15]  [220/400]  eta: 0:01:49  lr: 0.000061  loss: 2.5121 (2.4831)  grad: 0.2235 (0.2222)  time: 0.6079  data: 0.0035  max mem: 57344
+train: [15]  [240/400]  eta: 0:01:37  lr: 0.000059  loss: 2.4876 (2.4805)  grad: 0.2225 (0.2219)  time: 0.6080  data: 0.0034  max mem: 57344
+train: [15]  [260/400]  eta: 0:01:25  lr: 0.000058  loss: 2.4717 (2.4824)  grad: 0.2151 (0.2216)  time: 0.6079  data: 0.0033  max mem: 57344
+train: [15]  [280/400]  eta: 0:01:13  lr: 0.000057  loss: 2.4818 (2.4817)  grad: 0.2144 (0.2213)  time: 0.6103  data: 0.0040  max mem: 57344
+train: [15]  [300/400]  eta: 0:01:00  lr: 0.000056  loss: 2.4502 (2.4779)  grad: 0.2115 (0.2205)  time: 0.6105  data: 0.0042  max mem: 57344
+train: [15]  [320/400]  eta: 0:00:48  lr: 0.000054  loss: 2.4595 (2.4808)  grad: 0.2115 (0.2203)  time: 0.6095  data: 0.0039  max mem: 57344
+train: [15]  [340/400]  eta: 0:00:36  lr: 0.000053  loss: 2.4709 (2.4783)  grad: 0.2170 (0.2199)  time: 0.6097  data: 0.0037  max mem: 57344
+train: [15]  [360/400]  eta: 0:00:24  lr: 0.000052  loss: 2.4557 (2.4776)  grad: 0.2140 (0.2197)  time: 0.6097  data: 0.0037  max mem: 57344
+train: [15]  [380/400]  eta: 0:00:12  lr: 0.000051  loss: 2.4717 (2.4780)  grad: 0.2140 (0.2194)  time: 0.6088  data: 0.0035  max mem: 57344
+train: [15]  [399/400]  eta: 0:00:00  lr: 0.000050  loss: 2.4808 (2.4789)  grad: 0.2190 (0.2196)  time: 0.6092  data: 0.0035  max mem: 57344
+train: [15] Total time: 0:04:03 (0.6097 s / it)
+train: [15] Summary: lr: 0.000050  loss: 2.4808 (2.4789)  grad: 0.2190 (0.2196)
+eval (validation): [15]  [ 0/85]  eta: 0:01:24    time: 0.9913  data: 0.6287  max mem: 57344
+eval (validation): [15]  [20/85]  eta: 0:00:26    time: 0.3706  data: 0.0029  max mem: 57344
+eval (validation): [15]  [40/85]  eta: 0:00:17    time: 0.3706  data: 0.0032  max mem: 57344
+eval (validation): [15]  [60/85]  eta: 0:00:09    time: 0.3706  data: 0.0035  max mem: 57344
+eval (validation): [15]  [80/85]  eta: 0:00:01    time: 0.3709  data: 0.0033  max mem: 57344
+eval (validation): [15]  [84/85]  eta: 0:00:00    time: 0.3643  data: 0.0033  max mem: 57344
+eval (validation): [15] Total time: 0:00:32 (0.3776 s / it)
+cv: [15]  best hparam: (1.2, 1.0) (025) ('025_lr1.2e+00_wd1.0e+00')  loss: 2.360  acc: 0.291  f1: 0.243
+saving checkpoint experiments/schaefer1000/output/schaefer1000/schaefer1000_lr3e-4_4/eval_v2/nsd_cococlip__patch__attn/checkpoint-last.pth
+saving best checkpoint experiments/schaefer1000/output/schaefer1000/schaefer1000_lr3e-4_4/eval_v2/nsd_cococlip__patch__attn/checkpoint-best.pth
+train: [16]  [  0/400]  eta: 0:08:16  lr: nan  time: 1.2410  data: 0.6460  max mem: 57344
+train: [16]  [ 20/400]  eta: 0:04:02  lr: 0.000048  loss: 2.4360 (2.4863)  grad: 0.2155 (0.2175)  time: 0.6069  data: 0.0024  max mem: 57344
+train: [16]  [ 40/400]  eta: 0:03:44  lr: 0.000047  loss: 2.4360 (2.4772)  grad: 0.2119 (0.2154)  time: 0.6076  data: 0.0034  max mem: 57344
+train: [16]  [ 60/400]  eta: 0:03:29  lr: 0.000046  loss: 2.4436 (2.4599)  grad: 0.2053 (0.2133)  time: 0.6070  data: 0.0033  max mem: 57344
+train: [16]  [ 80/400]  eta: 0:03:16  lr: 0.000045  loss: 2.4542 (2.4597)  grad: 0.2068 (0.2138)  time: 0.6073  data: 0.0033  max mem: 57344
+train: [16]  [100/400]  eta: 0:03:04  lr: 0.000044  loss: 2.4323 (2.4506)  grad: 0.2122 (0.2135)  time: 0.6074  data: 0.0033  max mem: 57344
+train: [16]  [120/400]  eta: 0:02:51  lr: 0.000043  loss: 2.4668 (2.4603)  grad: 0.2104 (0.2139)  time: 0.6070  data: 0.0032  max mem: 57344
+train: [16]  [140/400]  eta: 0:02:39  lr: 0.000042  loss: 2.4795 (2.4578)  grad: 0.2068 (0.2130)  time: 0.6078  data: 0.0034  max mem: 57344
+train: [16]  [160/400]  eta: 0:02:26  lr: 0.000041  loss: 2.4271 (2.4567)  grad: 0.2065 (0.2128)  time: 0.6071  data: 0.0032  max mem: 57344
+train: [16]  [180/400]  eta: 0:02:14  lr: 0.000040  loss: 2.4644 (2.4597)  grad: 0.2065 (0.2127)  time: 0.6072  data: 0.0032  max mem: 57344
+train: [16]  [200/400]  eta: 0:02:02  lr: 0.000039  loss: 2.4679 (2.4609)  grad: 0.2045 (0.2124)  time: 0.6095  data: 0.0037  max mem: 57344
+train: [16]  [220/400]  eta: 0:01:49  lr: 0.000038  loss: 2.4605 (2.4618)  grad: 0.2170 (0.2134)  time: 0.6102  data: 0.0039  max mem: 57344
+train: [16]  [240/400]  eta: 0:01:37  lr: 0.000036  loss: 2.4721 (2.4643)  grad: 0.2161 (0.2130)  time: 0.6083  data: 0.0036  max mem: 57344
+train: [16]  [260/400]  eta: 0:01:25  lr: 0.000035  loss: 2.4751 (2.4640)  grad: 0.2123 (0.2134)  time: 0.6077  data: 0.0034  max mem: 57344
+train: [16]  [280/400]  eta: 0:01:13  lr: 0.000034  loss: 2.4686 (2.4632)  grad: 0.2229 (0.2143)  time: 0.6072  data: 0.0034  max mem: 57344
+train: [16]  [300/400]  eta: 0:01:00  lr: 0.000033  loss: 2.4332 (2.4614)  grad: 0.2243 (0.2147)  time: 0.6082  data: 0.0035  max mem: 57344
+train: [16]  [320/400]  eta: 0:00:48  lr: 0.000032  loss: 2.4716 (2.4638)  grad: 0.2140 (0.2144)  time: 0.6076  data: 0.0035  max mem: 57344
+train: [16]  [340/400]  eta: 0:00:36  lr: 0.000031  loss: 2.4979 (2.4651)  grad: 0.2140 (0.2149)  time: 0.6085  data: 0.0035  max mem: 57344
+train: [16]  [360/400]  eta: 0:00:24  lr: 0.000031  loss: 2.4776 (2.4666)  grad: 0.2158 (0.2150)  time: 0.6106  data: 0.0040  max mem: 57344
+train: [16]  [380/400]  eta: 0:00:12  lr: 0.000030  loss: 2.4859 (2.4673)  grad: 0.2066 (0.2148)  time: 0.6110  data: 0.0040  max mem: 57344
+train: [16]  [399/400]  eta: 0:00:00  lr: 0.000029  loss: 2.4770 (2.4677)  grad: 0.2064 (0.2146)  time: 0.6108  data: 0.0039  max mem: 57344
+train: [16] Total time: 0:04:04 (0.6101 s / it)
+train: [16] Summary: lr: 0.000029  loss: 2.4770 (2.4677)  grad: 0.2064 (0.2146)
+eval (validation): [16]  [ 0/85]  eta: 0:01:26    time: 1.0221  data: 0.6628  max mem: 57344
+eval (validation): [16]  [20/85]  eta: 0:00:26    time: 0.3701  data: 0.0025  max mem: 57344
+eval (validation): [16]  [40/85]  eta: 0:00:17    time: 0.3707  data: 0.0036  max mem: 57344
+eval (validation): [16]  [60/85]  eta: 0:00:09    time: 0.3710  data: 0.0037  max mem: 57344
+eval (validation): [16]  [80/85]  eta: 0:00:01    time: 0.3709  data: 0.0036  max mem: 57344
+eval (validation): [16]  [84/85]  eta: 0:00:00    time: 0.3645  data: 0.0036  max mem: 57344
+eval (validation): [16] Total time: 0:00:32 (0.3782 s / it)
+cv: [16]  best hparam: (1.4, 1.0) (026) ('026_lr1.4e+00_wd1.0e+00')  loss: 2.391  acc: 0.287  f1: 0.240
+saving checkpoint experiments/schaefer1000/output/schaefer1000/schaefer1000_lr3e-4_4/eval_v2/nsd_cococlip__patch__attn/checkpoint-last.pth
+train: [17]  [  0/400]  eta: 0:08:31  lr: nan  time: 1.2791  data: 0.6821  max mem: 57344
+train: [17]  [ 20/400]  eta: 0:04:02  lr: 0.000028  loss: 2.4422 (2.4889)  grad: 0.2093 (0.2081)  time: 0.6073  data: 0.0022  max mem: 57344
+train: [17]  [ 40/400]  eta: 0:03:44  lr: 0.000027  loss: 2.4304 (2.4381)  grad: 0.2145 (0.2129)  time: 0.6089  data: 0.0034  max mem: 57344
+train: [17]  [ 60/400]  eta: 0:03:30  lr: 0.000026  loss: 2.4138 (2.4312)  grad: 0.2172 (0.2129)  time: 0.6089  data: 0.0036  max mem: 57344
+train: [17]  [ 80/400]  eta: 0:03:17  lr: 0.000025  loss: 2.4775 (2.4504)  grad: 0.2145 (0.2130)  time: 0.6086  data: 0.0035  max mem: 57344
+train: [17]  [100/400]  eta: 0:03:04  lr: 0.000024  loss: 2.4899 (2.4572)  grad: 0.2098 (0.2128)  time: 0.6085  data: 0.0035  max mem: 57344
+train: [17]  [120/400]  eta: 0:02:51  lr: 0.000023  loss: 2.4849 (2.4515)  grad: 0.2090 (0.2134)  time: 0.6082  data: 0.0035  max mem: 57344
+train: [17]  [140/400]  eta: 0:02:39  lr: 0.000023  loss: 2.3853 (2.4434)  grad: 0.2092 (0.2128)  time: 0.6072  data: 0.0033  max mem: 57344
+train: [17]  [160/400]  eta: 0:02:26  lr: 0.000022  loss: 2.4453 (2.4485)  grad: 0.2119 (0.2132)  time: 0.6072  data: 0.0033  max mem: 57344
+train: [17]  [180/400]  eta: 0:02:14  lr: 0.000021  loss: 2.4629 (2.4491)  grad: 0.2168 (0.2134)  time: 0.6073  data: 0.0033  max mem: 57344
+train: [17]  [200/400]  eta: 0:02:02  lr: 0.000020  loss: 2.4566 (2.4500)  grad: 0.2100 (0.2134)  time: 0.6068  data: 0.0032  max mem: 57344
+train: [17]  [220/400]  eta: 0:01:49  lr: 0.000019  loss: 2.4566 (2.4492)  grad: 0.2121 (0.2133)  time: 0.6072  data: 0.0032  max mem: 57344
+train: [17]  [240/400]  eta: 0:01:37  lr: 0.000019  loss: 2.4190 (2.4481)  grad: 0.2080 (0.2128)  time: 0.6073  data: 0.0033  max mem: 57344
+train: [17]  [260/400]  eta: 0:01:25  lr: 0.000018  loss: 2.4102 (2.4488)  grad: 0.2053 (0.2132)  time: 0.6072  data: 0.0032  max mem: 57344
+train: [17]  [280/400]  eta: 0:01:13  lr: 0.000017  loss: 2.4240 (2.4486)  grad: 0.2143 (0.2135)  time: 0.6102  data: 0.0039  max mem: 57344
+train: [17]  [300/400]  eta: 0:01:01  lr: 0.000016  loss: 2.4240 (2.4462)  grad: 0.2137 (0.2134)  time: 0.6106  data: 0.0039  max mem: 57344
+train: [17]  [320/400]  eta: 0:00:48  lr: 0.000016  loss: 2.4264 (2.4484)  grad: 0.2108 (0.2131)  time: 0.6075  data: 0.0034  max mem: 57344
+train: [17]  [340/400]  eta: 0:00:36  lr: 0.000015  loss: 2.4559 (2.4491)  grad: 0.2084 (0.2130)  time: 0.6079  data: 0.0034  max mem: 57344
+train: [17]  [360/400]  eta: 0:00:24  lr: 0.000014  loss: 2.4419 (2.4480)  grad: 0.2108 (0.2132)  time: 0.6076  data: 0.0034  max mem: 57344
+train: [17]  [380/400]  eta: 0:00:12  lr: 0.000014  loss: 2.4419 (2.4495)  grad: 0.2108 (0.2132)  time: 0.6084  data: 0.0035  max mem: 57344
+train: [17]  [399/400]  eta: 0:00:00  lr: 0.000013  loss: 2.4350 (2.4483)  grad: 0.2089 (0.2134)  time: 0.6083  data: 0.0035  max mem: 57344
+train: [17] Total time: 0:04:04 (0.6100 s / it)
+train: [17] Summary: lr: 0.000013  loss: 2.4350 (2.4483)  grad: 0.2089 (0.2134)
+eval (validation): [17]  [ 0/85]  eta: 0:01:20    time: 0.9423  data: 0.5793  max mem: 57344
+eval (validation): [17]  [20/85]  eta: 0:00:25    time: 0.3693  data: 0.0020  max mem: 57344
+eval (validation): [17]  [40/85]  eta: 0:00:17    time: 0.3713  data: 0.0036  max mem: 57344
+eval (validation): [17]  [60/85]  eta: 0:00:09    time: 0.3719  data: 0.0037  max mem: 57344
+eval (validation): [17]  [80/85]  eta: 0:00:01    time: 0.3728  data: 0.0038  max mem: 57344
+eval (validation): [17]  [84/85]  eta: 0:00:00    time: 0.3663  data: 0.0037  max mem: 57344
+eval (validation): [17] Total time: 0:00:32 (0.3780 s / it)
+cv: [17]  best hparam: (1.4, 1.0) (026) ('026_lr1.4e+00_wd1.0e+00')  loss: 2.387  acc: 0.288  f1: 0.238
+saving checkpoint experiments/schaefer1000/output/schaefer1000/schaefer1000_lr3e-4_4/eval_v2/nsd_cococlip__patch__attn/checkpoint-last.pth
+train: [18]  [  0/400]  eta: 0:08:50  lr: nan  time: 1.3263  data: 0.7242  max mem: 57344
+train: [18]  [ 20/400]  eta: 0:04:04  lr: 0.000012  loss: 2.3850 (2.3971)  grad: 0.2095 (0.2103)  time: 0.6104  data: 0.0038  max mem: 57344
+train: [18]  [ 40/400]  eta: 0:03:45  lr: 0.000012  loss: 2.3515 (2.3829)  grad: 0.2095 (0.2091)  time: 0.6095  data: 0.0037  max mem: 57344
+train: [18]  [ 60/400]  eta: 0:03:31  lr: 0.000011  loss: 2.3506 (2.3891)  grad: 0.2102 (0.2120)  time: 0.6087  data: 0.0035  max mem: 57344
+train: [18]  [ 80/400]  eta: 0:03:17  lr: 0.000011  loss: 2.4201 (2.3989)  grad: 0.2074 (0.2114)  time: 0.6092  data: 0.0035  max mem: 57344
+train: [18]  [100/400]  eta: 0:03:04  lr: 0.000010  loss: 2.4519 (2.4133)  grad: 0.2111 (0.2143)  time: 0.6086  data: 0.0035  max mem: 57344
+train: [18]  [120/400]  eta: 0:02:52  lr: 0.000009  loss: 2.4609 (2.4197)  grad: 0.2264 (0.2145)  time: 0.6089  data: 0.0036  max mem: 57344
+train: [18]  [140/400]  eta: 0:02:39  lr: 0.000009  loss: 2.4148 (2.4169)  grad: 0.2125 (0.2127)  time: 0.6094  data: 0.0035  max mem: 57344
+train: [18]  [160/400]  eta: 0:02:27  lr: 0.000008  loss: 2.4454 (2.4218)  grad: 0.2079 (0.2130)  time: 0.6088  data: 0.0035  max mem: 57344
+train: [18]  [180/400]  eta: 0:02:14  lr: 0.000008  loss: 2.4609 (2.4235)  grad: 0.2098 (0.2131)  time: 0.6091  data: 0.0035  max mem: 57344
+train: [18]  [200/400]  eta: 0:02:02  lr: 0.000007  loss: 2.4093 (2.4186)  grad: 0.2089 (0.2126)  time: 0.6082  data: 0.0034  max mem: 57344
+train: [18]  [220/400]  eta: 0:01:50  lr: 0.000007  loss: 2.4246 (2.4221)  grad: 0.2042 (0.2130)  time: 0.6070  data: 0.0033  max mem: 57344
+train: [18]  [240/400]  eta: 0:01:37  lr: 0.000006  loss: 2.4490 (2.4257)  grad: 0.2138 (0.2135)  time: 0.6069  data: 0.0033  max mem: 57344
+train: [18]  [260/400]  eta: 0:01:25  lr: 0.000006  loss: 2.4250 (2.4215)  grad: 0.2139 (0.2139)  time: 0.6081  data: 0.0033  max mem: 57344
+train: [18]  [280/400]  eta: 0:01:13  lr: 0.000006  loss: 2.3775 (2.4196)  grad: 0.2069 (0.2133)  time: 0.6072  data: 0.0032  max mem: 57344
+train: [18]  [300/400]  eta: 0:01:01  lr: 0.000005  loss: 2.4055 (2.4213)  grad: 0.2077 (0.2137)  time: 0.6077  data: 0.0032  max mem: 57344
+train: [18]  [320/400]  eta: 0:00:48  lr: 0.000005  loss: 2.4348 (2.4209)  grad: 0.2146 (0.2140)  time: 0.6069  data: 0.0033  max mem: 57344
+train: [18]  [340/400]  eta: 0:00:36  lr: 0.000004  loss: 2.3994 (2.4202)  grad: 0.2145 (0.2141)  time: 0.6074  data: 0.0033  max mem: 57344
+train: [18]  [360/400]  eta: 0:00:24  lr: 0.000004  loss: 2.4112 (2.4214)  grad: 0.2116 (0.2137)  time: 0.6099  data: 0.0038  max mem: 57344
+train: [18]  [380/400]  eta: 0:00:12  lr: 0.000004  loss: 2.4351 (2.4220)  grad: 0.2097 (0.2138)  time: 0.6102  data: 0.0039  max mem: 57344
+train: [18]  [399/400]  eta: 0:00:00  lr: 0.000003  loss: 2.4403 (2.4239)  grad: 0.2098 (0.2142)  time: 0.6084  data: 0.0034  max mem: 57344
+train: [18] Total time: 0:04:04 (0.6106 s / it)
+train: [18] Summary: lr: 0.000003  loss: 2.4403 (2.4239)  grad: 0.2098 (0.2142)
+eval (validation): [18]  [ 0/85]  eta: 0:01:20    time: 0.9419  data: 0.5791  max mem: 57344
+eval (validation): [18]  [20/85]  eta: 0:00:25    time: 0.3702  data: 0.0021  max mem: 57344
+eval (validation): [18]  [40/85]  eta: 0:00:17    time: 0.3703  data: 0.0030  max mem: 57344
+eval (validation): [18]  [60/85]  eta: 0:00:09    time: 0.3704  data: 0.0029  max mem: 57344
+eval (validation): [18]  [80/85]  eta: 0:00:01    time: 0.3709  data: 0.0032  max mem: 57344
+eval (validation): [18]  [84/85]  eta: 0:00:00    time: 0.3645  data: 0.0032  max mem: 57344
+eval (validation): [18] Total time: 0:00:32 (0.3769 s / it)
+cv: [18]  best hparam: (1.4, 1.0) (026) ('026_lr1.4e+00_wd1.0e+00')  loss: 2.386  acc: 0.288  f1: 0.239
+saving checkpoint experiments/schaefer1000/output/schaefer1000/schaefer1000_lr3e-4_4/eval_v2/nsd_cococlip__patch__attn/checkpoint-last.pth
+train: [19]  [  0/400]  eta: 0:08:09  lr: nan  time: 1.2232  data: 0.6271  max mem: 57344
+train: [19]  [ 20/400]  eta: 0:04:01  lr: 0.000003  loss: 2.4032 (2.4314)  grad: 0.2080 (0.2109)  time: 0.6062  data: 0.0021  max mem: 57344
+train: [19]  [ 40/400]  eta: 0:03:44  lr: 0.000003  loss: 2.4181 (2.4385)  grad: 0.2111 (0.2130)  time: 0.6090  data: 0.0036  max mem: 57344
+train: [19]  [ 60/400]  eta: 0:03:30  lr: 0.000002  loss: 2.4100 (2.4325)  grad: 0.2098 (0.2131)  time: 0.6105  data: 0.0040  max mem: 57344
+train: [19]  [ 80/400]  eta: 0:03:17  lr: 0.000002  loss: 2.3918 (2.4313)  grad: 0.2173 (0.2152)  time: 0.6106  data: 0.0039  max mem: 57344
+train: [19]  [100/400]  eta: 0:03:04  lr: 0.000002  loss: 2.4219 (2.4386)  grad: 0.2173 (0.2142)  time: 0.6104  data: 0.0038  max mem: 57344
+train: [19]  [120/400]  eta: 0:02:52  lr: 0.000002  loss: 2.4219 (2.4299)  grad: 0.2008 (0.2121)  time: 0.6092  data: 0.0037  max mem: 57344
+train: [19]  [140/400]  eta: 0:02:39  lr: 0.000001  loss: 2.3927 (2.4278)  grad: 0.2036 (0.2118)  time: 0.6089  data: 0.0035  max mem: 57344
+train: [19]  [160/400]  eta: 0:02:27  lr: 0.000001  loss: 2.4265 (2.4294)  grad: 0.2115 (0.2121)  time: 0.6092  data: 0.0035  max mem: 57344
+train: [19]  [180/400]  eta: 0:02:14  lr: 0.000001  loss: 2.4217 (2.4276)  grad: 0.2136 (0.2118)  time: 0.6092  data: 0.0035  max mem: 57344
+train: [19]  [200/400]  eta: 0:02:02  lr: 0.000001  loss: 2.4492 (2.4305)  grad: 0.2147 (0.2123)  time: 0.6092  data: 0.0036  max mem: 57344
+train: [19]  [220/400]  eta: 0:01:50  lr: 0.000001  loss: 2.4370 (2.4281)  grad: 0.2116 (0.2123)  time: 0.6089  data: 0.0035  max mem: 57344
+train: [19]  [240/400]  eta: 0:01:37  lr: 0.000001  loss: 2.3758 (2.4242)  grad: 0.2078 (0.2118)  time: 0.6086  data: 0.0035  max mem: 57344
+train: [19]  [260/400]  eta: 0:01:25  lr: 0.000000  loss: 2.3758 (2.4249)  grad: 0.2072 (0.2115)  time: 0.6088  data: 0.0035  max mem: 57344
+train: [19]  [280/400]  eta: 0:01:13  lr: 0.000000  loss: 2.4608 (2.4294)  grad: 0.2084 (0.2116)  time: 0.6073  data: 0.0033  max mem: 57344
+train: [19]  [300/400]  eta: 0:01:01  lr: 0.000000  loss: 2.4530 (2.4291)  grad: 0.2087 (0.2112)  time: 0.6076  data: 0.0033  max mem: 57344
+train: [19]  [320/400]  eta: 0:00:48  lr: 0.000000  loss: 2.4109 (2.4279)  grad: 0.2087 (0.2114)  time: 0.6073  data: 0.0033  max mem: 57344
+train: [19]  [340/400]  eta: 0:00:36  lr: 0.000000  loss: 2.4298 (2.4271)  grad: 0.2079 (0.2109)  time: 0.6082  data: 0.0034  max mem: 57344
+train: [19]  [360/400]  eta: 0:00:24  lr: 0.000000  loss: 2.4366 (2.4296)  grad: 0.2113 (0.2113)  time: 0.6071  data: 0.0033  max mem: 57344
+train: [19]  [380/400]  eta: 0:00:12  lr: 0.000000  loss: 2.4452 (2.4303)  grad: 0.2113 (0.2111)  time: 0.6077  data: 0.0034  max mem: 57344
+train: [19]  [399/400]  eta: 0:00:00  lr: 0.000000  loss: 2.4378 (2.4317)  grad: 0.2037 (0.2108)  time: 0.6074  data: 0.0034  max mem: 57344
+train: [19] Total time: 0:04:04 (0.6104 s / it)
+train: [19] Summary: lr: 0.000000  loss: 2.4378 (2.4317)  grad: 0.2037 (0.2108)
+eval (validation): [19]  [ 0/85]  eta: 0:01:26    time: 1.0161  data: 0.6533  max mem: 57344
+eval (validation): [19]  [20/85]  eta: 0:00:26    time: 0.3696  data: 0.0024  max mem: 57344
+eval (validation): [19]  [40/85]  eta: 0:00:17    time: 0.3712  data: 0.0037  max mem: 57344
+eval (validation): [19]  [60/85]  eta: 0:00:09    time: 0.3716  data: 0.0037  max mem: 57344
+eval (validation): [19]  [80/85]  eta: 0:00:01    time: 0.3721  data: 0.0037  max mem: 57344
+eval (validation): [19]  [84/85]  eta: 0:00:00    time: 0.3654  data: 0.0037  max mem: 57344
+eval (validation): [19] Total time: 0:00:32 (0.3785 s / it)
+cv: [19]  best hparam: (1.4, 1.0) (026) ('026_lr1.4e+00_wd1.0e+00')  loss: 2.388  acc: 0.289  f1: 0.239
+saving checkpoint experiments/schaefer1000/output/schaefer1000/schaefer1000_lr3e-4_4/eval_v2/nsd_cococlip__patch__attn/checkpoint-last.pth
+evaluating last checkpoint: experiments/schaefer1000/output/schaefer1000/schaefer1000_lr3e-4_4/eval_v2/nsd_cococlip__patch__attn/checkpoint-last.pth
+eval model info:
+{"score": 0.2888519748984865, "hparam": [1.4, 1.0], "hparam_id": 26, "epoch": 19, "is_best": false, "best_score": 0.29088224437061644}
+eval (train): [20]  [  0/509]  eta: 0:08:13    time: 0.9698  data: 0.6105  max mem: 57344
+eval (train): [20]  [ 20/509]  eta: 0:03:14    time: 0.3686  data: 0.0022  max mem: 57344
+eval (train): [20]  [ 40/509]  eta: 0:03:00    time: 0.3702  data: 0.0033  max mem: 57344
+eval (train): [20]  [ 60/509]  eta: 0:02:50    time: 0.3704  data: 0.0034  max mem: 57344
+eval (train): [20]  [ 80/509]  eta: 0:02:41    time: 0.3701  data: 0.0032  max mem: 57344
+eval (train): [20]  [100/509]  eta: 0:02:33    time: 0.3710  data: 0.0034  max mem: 57344
+eval (train): [20]  [120/509]  eta: 0:02:25    time: 0.3712  data: 0.0033  max mem: 57344
+eval (train): [20]  [140/509]  eta: 0:02:18    time: 0.3705  data: 0.0033  max mem: 57344
+eval (train): [20]  [160/509]  eta: 0:02:10    time: 0.3710  data: 0.0033  max mem: 57344
+eval (train): [20]  [180/509]  eta: 0:02:02    time: 0.3714  data: 0.0036  max mem: 57344
+eval (train): [20]  [200/509]  eta: 0:01:55    time: 0.3723  data: 0.0038  max mem: 57344
+eval (train): [20]  [220/509]  eta: 0:01:47    time: 0.3724  data: 0.0037  max mem: 57344
+eval (train): [20]  [240/509]  eta: 0:01:40    time: 0.3726  data: 0.0039  max mem: 57344
+eval (train): [20]  [260/509]  eta: 0:01:32    time: 0.3723  data: 0.0037  max mem: 57344
+eval (train): [20]  [280/509]  eta: 0:01:25    time: 0.3714  data: 0.0036  max mem: 57344
+eval (train): [20]  [300/509]  eta: 0:01:17    time: 0.3714  data: 0.0036  max mem: 57344
+eval (train): [20]  [320/509]  eta: 0:01:10    time: 0.3709  data: 0.0036  max mem: 57344
+eval (train): [20]  [340/509]  eta: 0:01:03    time: 0.3709  data: 0.0033  max mem: 57344
+eval (train): [20]  [360/509]  eta: 0:00:55    time: 0.3716  data: 0.0035  max mem: 57344
+eval (train): [20]  [380/509]  eta: 0:00:48    time: 0.3718  data: 0.0034  max mem: 57344
+eval (train): [20]  [400/509]  eta: 0:00:40    time: 0.3716  data: 0.0035  max mem: 57344
+eval (train): [20]  [420/509]  eta: 0:00:33    time: 0.3710  data: 0.0034  max mem: 57344
+eval (train): [20]  [440/509]  eta: 0:00:25    time: 0.3711  data: 0.0034  max mem: 57344
+eval (train): [20]  [460/509]  eta: 0:00:18    time: 0.3712  data: 0.0034  max mem: 57344
+eval (train): [20]  [480/509]  eta: 0:00:10    time: 0.3714  data: 0.0035  max mem: 57344
+eval (train): [20]  [500/509]  eta: 0:00:03    time: 0.3710  data: 0.0034  max mem: 57344
+eval (train): [20]  [508/509]  eta: 0:00:00    time: 0.3608  data: 0.0034  max mem: 57344
+eval (train): [20] Total time: 0:03:09 (0.3723 s / it)
+eval (validation): [20]  [ 0/85]  eta: 0:01:27    time: 1.0348  data: 0.6717  max mem: 57344
+eval (validation): [20]  [20/85]  eta: 0:00:26    time: 0.3696  data: 0.0022  max mem: 57344
+eval (validation): [20]  [40/85]  eta: 0:00:17    time: 0.3700  data: 0.0032  max mem: 57344
+eval (validation): [20]  [60/85]  eta: 0:00:09    time: 0.3708  data: 0.0032  max mem: 57344
+eval (validation): [20]  [80/85]  eta: 0:00:01    time: 0.3702  data: 0.0031  max mem: 57344
+eval (validation): [20]  [84/85]  eta: 0:00:00    time: 0.3643  data: 0.0032  max mem: 57344
+eval (validation): [20] Total time: 0:00:32 (0.3777 s / it)
+eval (test): [20]  [ 0/85]  eta: 0:01:17    time: 0.9086  data: 0.5476  max mem: 57344
+eval (test): [20]  [20/85]  eta: 0:00:25    time: 0.3690  data: 0.0024  max mem: 57344
+eval (test): [20]  [40/85]  eta: 0:00:17    time: 0.3709  data: 0.0032  max mem: 57344
+eval (test): [20]  [60/85]  eta: 0:00:09    time: 0.3706  data: 0.0032  max mem: 57344
+eval (test): [20]  [80/85]  eta: 0:00:01    time: 0.3699  data: 0.0031  max mem: 57344
+eval (test): [20]  [84/85]  eta: 0:00:00    time: 0.3560  data: 0.0031  max mem: 57344
+eval (test): [20] Total time: 0:00:31 (0.3743 s / it)
+eval (testid): [20]  [ 0/82]  eta: 0:01:06    time: 0.8123  data: 0.4519  max mem: 57344
+eval (testid): [20]  [20/82]  eta: 0:00:24    time: 0.3700  data: 0.0030  max mem: 57344
+eval (testid): [20]  [40/82]  eta: 0:00:15    time: 0.3698  data: 0.0032  max mem: 57344
+eval (testid): [20]  [60/82]  eta: 0:00:08    time: 0.3700  data: 0.0031  max mem: 57344
+eval (testid): [20]  [80/82]  eta: 0:00:00    time: 0.3701  data: 0.0033  max mem: 57344
+eval (testid): [20]  [81/82]  eta: 0:00:00    time: 0.3530  data: 0.0033  max mem: 57344
+eval (testid): [20] Total time: 0:00:30 (0.3725 s / it)
+evaluating best checkpoint: experiments/schaefer1000/output/schaefer1000/schaefer1000_lr3e-4_4/eval_v2/nsd_cococlip__patch__attn/checkpoint-best.pth
+eval model info:
+{"score": 0.29088224437061644, "hparam": [1.2, 1.0], "hparam_id": 25, "epoch": 15, "is_best": true, "best_score": 0.29088224437061644}
+eval (train): [20]  [  0/509]  eta: 0:07:46    time: 0.9157  data: 0.5578  max mem: 57344
+eval (train): [20]  [ 20/509]  eta: 0:03:13    time: 0.3700  data: 0.0028  max mem: 57344
+eval (train): [20]  [ 40/509]  eta: 0:03:00    time: 0.3720  data: 0.0040  max mem: 57344
+eval (train): [20]  [ 60/509]  eta: 0:02:50    time: 0.3720  data: 0.0040  max mem: 57344
+eval (train): [20]  [ 80/509]  eta: 0:02:42    time: 0.3714  data: 0.0034  max mem: 57344
+eval (train): [20]  [100/509]  eta: 0:02:34    time: 0.3711  data: 0.0033  max mem: 57344
+eval (train): [20]  [120/509]  eta: 0:02:26    time: 0.3706  data: 0.0033  max mem: 57344
+eval (train): [20]  [140/509]  eta: 0:02:18    time: 0.3714  data: 0.0033  max mem: 57344
+eval (train): [20]  [160/509]  eta: 0:02:10    time: 0.3706  data: 0.0032  max mem: 57344
+eval (train): [20]  [180/509]  eta: 0:02:03    time: 0.3711  data: 0.0033  max mem: 57344
+eval (train): [20]  [200/509]  eta: 0:01:55    time: 0.3712  data: 0.0033  max mem: 57344
+eval (train): [20]  [220/509]  eta: 0:01:47    time: 0.3713  data: 0.0033  max mem: 57344
+eval (train): [20]  [240/509]  eta: 0:01:40    time: 0.3715  data: 0.0033  max mem: 57344
+eval (train): [20]  [260/509]  eta: 0:01:32    time: 0.3718  data: 0.0035  max mem: 57344
+eval (train): [20]  [280/509]  eta: 0:01:25    time: 0.3725  data: 0.0038  max mem: 57344
+eval (train): [20]  [300/509]  eta: 0:01:18    time: 0.3728  data: 0.0037  max mem: 57344
+eval (train): [20]  [320/509]  eta: 0:01:10    time: 0.3725  data: 0.0036  max mem: 57344
+eval (train): [20]  [340/509]  eta: 0:01:03    time: 0.3734  data: 0.0039  max mem: 57344
+eval (train): [20]  [360/509]  eta: 0:00:55    time: 0.3720  data: 0.0038  max mem: 57344
+eval (train): [20]  [380/509]  eta: 0:00:48    time: 0.3712  data: 0.0037  max mem: 57344
+eval (train): [20]  [400/509]  eta: 0:00:40    time: 0.3718  data: 0.0035  max mem: 57344
+eval (train): [20]  [420/509]  eta: 0:00:33    time: 0.3716  data: 0.0034  max mem: 57344
+eval (train): [20]  [440/509]  eta: 0:00:25    time: 0.3719  data: 0.0034  max mem: 57344
+eval (train): [20]  [460/509]  eta: 0:00:18    time: 0.3718  data: 0.0035  max mem: 57344
+eval (train): [20]  [480/509]  eta: 0:00:10    time: 0.3716  data: 0.0034  max mem: 57344
+eval (train): [20]  [500/509]  eta: 0:00:03    time: 0.3716  data: 0.0034  max mem: 57344
+eval (train): [20]  [508/509]  eta: 0:00:00    time: 0.3612  data: 0.0034  max mem: 57344
+eval (train): [20] Total time: 0:03:09 (0.3727 s / it)
+eval (validation): [20]  [ 0/85]  eta: 0:01:24    time: 0.9980  data: 0.6353  max mem: 57344
+eval (validation): [20]  [20/85]  eta: 0:00:26    time: 0.3702  data: 0.0032  max mem: 57344
+eval (validation): [20]  [40/85]  eta: 0:00:17    time: 0.3705  data: 0.0035  max mem: 57344
+eval (validation): [20]  [60/85]  eta: 0:00:09    time: 0.3704  data: 0.0034  max mem: 57344
+eval (validation): [20]  [80/85]  eta: 0:00:01    time: 0.3707  data: 0.0033  max mem: 57344
+eval (validation): [20]  [84/85]  eta: 0:00:00    time: 0.3647  data: 0.0033  max mem: 57344
+eval (validation): [20] Total time: 0:00:32 (0.3778 s / it)
+eval (test): [20]  [ 0/85]  eta: 0:01:26    time: 1.0203  data: 0.6558  max mem: 57344
+eval (test): [20]  [20/85]  eta: 0:00:26    time: 0.3699  data: 0.0029  max mem: 57344
+eval (test): [20]  [40/85]  eta: 0:00:17    time: 0.3698  data: 0.0028  max mem: 57344
+eval (test): [20]  [60/85]  eta: 0:00:09    time: 0.3707  data: 0.0032  max mem: 57344
+eval (test): [20]  [80/85]  eta: 0:00:01    time: 0.3701  data: 0.0032  max mem: 57344
+eval (test): [20]  [84/85]  eta: 0:00:00    time: 0.3561  data: 0.0032  max mem: 57344
+eval (test): [20] Total time: 0:00:31 (0.3757 s / it)
+eval (testid): [20]  [ 0/82]  eta: 0:01:15    time: 0.9256  data: 0.5668  max mem: 57344
+eval (testid): [20]  [20/82]  eta: 0:00:24    time: 0.3689  data: 0.0021  max mem: 57344
+eval (testid): [20]  [40/82]  eta: 0:00:16    time: 0.3702  data: 0.0035  max mem: 57344
+eval (testid): [20]  [60/82]  eta: 0:00:08    time: 0.3698  data: 0.0031  max mem: 57344
+eval (testid): [20]  [80/82]  eta: 0:00:00    time: 0.3704  data: 0.0032  max mem: 57344
+eval (testid): [20]  [81/82]  eta: 0:00:00    time: 0.3532  data: 0.0032  max mem: 57344
+eval (testid): [20] Total time: 0:00:30 (0.3736 s / it)
+eval results:
+
+| model            | repr   | clf   | dataset      | ckpt   |   epoch |      lr |   wd |   hparam_id | hparam     | split      |   loss |     acc |   acc_std |      f1 |    f1_std |
+|:-----------------|:-------|:------|:-------------|:-------|--------:|--------:|-----:|------------:|:-----------|:-----------|-------:|--------:|----------:|--------:|----------:|
+| schaefer1000_mae | patch  | attn  | nsd_cococlip | best   |      15 | 0.00036 | 0.05 |          25 | [1.2, 1.0] | train      | 1.9453 | 0.41138 | 0.0024169 | 0.36819 | 0.0026383 |
+| schaefer1000_mae | patch  | attn  | nsd_cococlip | best   |      15 | 0.00036 | 0.05 |          25 | [1.2, 1.0] | validation | 2.3598 | 0.29088 | 0.0055433 | 0.24273 | 0.0056067 |
+| schaefer1000_mae | patch  | attn  | nsd_cococlip | best   |      15 | 0.00036 | 0.05 |          25 | [1.2, 1.0] | test       | 2.2764 | 0.30501 | 0.0054423 | 0.24336 | 0.005427  |
+| schaefer1000_mae | patch  | attn  | nsd_cococlip | best   |      15 | 0.00036 | 0.05 |          25 | [1.2, 1.0] | testid     | 2.2771 | 0.30017 | 0.0057446 | 0.2568  | 0.005662  |
+
+
+done! total time: 1:44:21

schaefer1000/schaefer1000_lr3e-4_4/pretrain/config.yaml

@@ -0,0 +1,102 @@
+name: schaefer1000/schaefer1000_lr3e-4_4/pretrain
+notes: schaefer1000 ablation schaefer1000_lr3e-4_4 (input_space=schaefer1000 base_lr=3e-4
+  seed=5404)
+output_dir: experiments/schaefer1000/output/schaefer1000/schaefer1000_lr3e-4_4/pretrain
+input_space: schaefer1000
+patch_size: 1
+num_frames: 16
+t_patch_size: 4
+mask_ratio: 0.9
+pred_mask_ratio: null
+masking: tube
+masking_kwargs: {}
+mask_patch_size: null
+model: mae_vit_base
+model_kwargs:
+  decoding: attn
+  pos_embed: sep
+  target_norm: null
+  pca_norm_nc: 2
+  t_pred_stride: 2
+  no_decode_pos: true
+  mask_drop_scale: false
+  pred_edge_pad: 0
+  gauss_sigma: null
+  class_token: true
+  reg_tokens: 0
+  no_embed_class: true
+  head_init_scale: 0.0
+  decoder_depth: 4
+  drop_path_rate: 0.0
+datasets:
+  hcp-train:
+    type: wds
+    url: /data/fmri-datasets/pretrain/hcpya-all.${input_space}.wds/hcpya-all-${input_space}-{00000..01799}.tar
+    clipping: random
+    clipping_kwargs:
+      oversample: 4.0
+    shuffle: true
+    buffer_size: 2000
+    samples_per_epoch: 200000
+  hcp-train-subset:
+    type: arrow
+    root: s3://medarc/fmri-datasets/eval/hcpya-clips.${input_space}.arrow/validation
+    split_range:
+    - 0
+    - 2000
+    shuffle: false
+  hcp-val:
+    type: arrow
+    root: s3://medarc/fmri-datasets/eval/hcpya-clips.${input_space}.arrow/test
+    split_range:
+    - 0
+    - 2000
+    shuffle: false
+train_dataset: hcp-train
+eval_datasets:
+- hcp-train-subset
+- hcp-val
+val_dataset: null
+clip_vmax: 3.0
+normalize: frame
+tr_scale: null
+crop_scale: null
+crop_aspect: null
+gray_jitter: null
+num_workers: 16
+epochs: 100
+batch_size: 32
+accum_iter: 1
+base_lr: 0.0003
+min_lr: 0.0
+warmup_epochs: 5
+weight_decay: 0.05
+betas:
+- 0.9
+- 0.95
+clip_grad: 1.0
+amp: true
+amp_dtype: float16
+ckpt: null
+resume: true
+auto_resume: true
+start_epoch: 0
+max_checkpoints: 0
+checkpoint_period: null
+plot_period: 5
+device: cuda
+presend_cuda: false
+seed: 5404
+debug: false
+wandb: true
+wandb_entity: null
+wandb_project: fMRI-foundation-model
+rank: 0
+world_size: 1
+gpu: 0
+distributed: true
+dist_backend: nccl
+in_chans: 1
+img_size:
+- 1000
+- 1

schaefer1000/schaefer1000_lr3e-4_4/pretrain/log.json

@@ -0,0 +1,100 @@
+{"epoch": 0, "train/lr": 3.7507200230407366e-06, "train/grad": 2.134240338090502, "train/loss": 0.9846526712608338, "eval/hcp-train-subset/loss": 0.9654733163695182, "eval/hcp-val/loss": 0.9612428363292448}
+{"epoch": 1, "train/lr": 1.125096003072098e-05, "train/grad": 3.6396139876800144, "train/loss": 0.9181011641407013, "eval/hcp-train-subset/loss": 0.8792350820956691, "eval/hcp-val/loss": 0.8735126458829449}
+{"epoch": 2, "train/lr": 1.875120003840123e-05, "train/grad": 2.7922826140179184, "train/loss": 0.8421830167293548, "eval/hcp-train-subset/loss": 0.8048923784686673, "eval/hcp-val/loss": 0.796557883101125}
+{"epoch": 3, "train/lr": 2.625144004608147e-05, "train/grad": 2.021520472834037, "train/loss": 0.7766621382427216, "eval/hcp-train-subset/loss": 0.7720528091153791, "eval/hcp-val/loss": 0.7628026604652405}
+{"epoch": 4, "train/lr": 3.375167986175557e-05, "train/grad": 1.3682986922849554, "train/loss": 0.7622566637039184, "eval/hcp-train-subset/loss": 0.7622794739661678, "eval/hcp-val/loss": 0.7536938632688215}
+{"epoch": 5, "train/lr": 3.749658191365373e-05, "train/grad": 0.9582835134695449, "train/loss": 0.7487169290542602, "eval/hcp-train-subset/loss": 0.750582133570025, "eval/hcp-val/loss": 0.7412037291834431}
+{"epoch": 6, "train/lr": 3.74760811613377e-05, "train/grad": 0.742696480894707, "train/loss": 0.7394470161914826, "eval/hcp-train-subset/loss": 0.7451939996211759, "eval/hcp-val/loss": 0.7367144486596507}
+{"epoch": 7, "train/lr": 3.743510371420547e-05, "train/grad": 0.6169777847784738, "train/loss": 0.7343027816295624, "eval/hcp-train-subset/loss": 0.7401283892893022, "eval/hcp-val/loss": 0.731248878663586}
+{"epoch": 8, "train/lr": 3.73736943804934e-05, "train/grad": 0.5466088711671568, "train/loss": 0.7325876157665253, "eval/hcp-train-subset/loss": 0.7359288719392592, "eval/hcp-val/loss": 0.7265200115019276}
+{"epoch": 9, "train/lr": 3.7291920310406644e-05, "train/grad": 0.4898914724888946, "train/loss": 0.7269160768222809, "eval/hcp-train-subset/loss": 0.7329672709588082, "eval/hcp-val/loss": 0.7245836719389884}
+{"epoch": 10, "train/lr": 3.718987092269037e-05, "train/grad": 0.4512696563147995, "train/loss": 0.7258539622783661, "eval/hcp-train-subset/loss": 0.7308645267640391, "eval/hcp-val/loss": 0.721330375440659}
+{"epoch": 11, "train/lr": 3.706765780685143e-05, "train/grad": 0.42288669853644006, "train/loss": 0.7198157748031616, "eval/hcp-train-subset/loss": 0.7274544479385499, "eval/hcp-val/loss": 0.7187711952194091}
+{"epoch": 12, "train/lr": 3.692541460113792e-05, "train/grad": 0.3955775134674507, "train/loss": 0.7173363891506195, "eval/hcp-train-subset/loss": 0.7259065191591939, "eval/hcp-val/loss": 0.7173883857265595}
+{"epoch": 13, "train/lr": 3.6763296846406675e-05, "train/grad": 0.3801466648041849, "train/loss": 0.7161822625541687, "eval/hcp-train-subset/loss": 0.7241568988369357, "eval/hcp-val/loss": 0.7155001317301104}
+{"epoch": 14, "train/lr": 3.658148181604263e-05, "train/grad": 0.36002728717506877, "train/loss": 0.7169261291027069, "eval/hcp-train-subset/loss": 0.7230653897408517, "eval/hcp-val/loss": 0.7144085562998249}
+{"epoch": 15, "train/lr": 3.6380168322111824e-05, "train/grad": 0.35119914208982056, "train/loss": 0.7107337213039399, "eval/hcp-train-subset/loss": 0.7210659182840778, "eval/hcp-val/loss": 0.7120572616977077}
+{"epoch": 16, "train/lr": 3.615957649796421e-05, "train/grad": 0.34019720957472704, "train/loss": 0.7087781484031678, "eval/hcp-train-subset/loss": 0.7194117384572183, "eval/hcp-val/loss": 0.7110743464962128}
+{"epoch": 17, "train/lr": 3.591994755752113e-05, "train/grad": 0.33343158180482374, "train/loss": 0.7122406131267548, "eval/hcp-train-subset/loss": 0.7180126693940931, "eval/hcp-val/loss": 0.7096054092530282}
+{"epoch": 18, "train/lr": 3.5661543531510486e-05, "train/grad": 0.32558281325150895, "train/loss": 0.7074071199512482, "eval/hcp-train-subset/loss": 0.7171267280655522, "eval/hcp-val/loss": 0.7081384812631915}
+{"epoch": 19, "train/lr": 3.538464698094067e-05, "train/grad": 0.32715296563299556, "train/loss": 0.7026601014709473, "eval/hcp-train-subset/loss": 0.7164860223570177, "eval/hcp-val/loss": 0.7076563719780214}
+{"epoch": 20, "train/lr": 3.508956068812486e-05, "train/grad": 0.3302246261419669, "train/loss": 0.703736708984375, "eval/hcp-train-subset/loss": 0.7145615095092405, "eval/hcp-val/loss": 0.7070394100681427}
+{"epoch": 21, "train/lr": 3.4776607325591504e-05, "train/grad": 0.3221865713229062, "train/loss": 0.7028804015827179, "eval/hcp-train-subset/loss": 0.7144837437137481, "eval/hcp-val/loss": 0.7066356203248424}
+{"epoch": 22, "train/lr": 3.4446129103247903e-05, "train/grad": 0.32087653478129796, "train/loss": 0.6997607206630707, "eval/hcp-train-subset/loss": 0.7143522049150159, "eval/hcp-val/loss": 0.707490578774483}
+{"epoch": 23, "train/lr": 3.4098487394178203e-05, "train/grad": 0.32510492152904, "train/loss": 0.697772089395523, "eval/hcp-train-subset/loss": 0.7129459458012735, "eval/hcp-val/loss": 0.7055421356231936}
+{"epoch": 24, "train/lr": 3.37340623394871e-05, "train/grad": 0.3196508876152834, "train/loss": 0.6962286649417877, "eval/hcp-train-subset/loss": 0.714224309690537, "eval/hcp-val/loss": 0.7050911726490143}
+{"epoch": 25, "train/lr": 3.335325243262167e-05, "train/grad": 0.3138868521313983, "train/loss": 0.6969752832794189, "eval/hcp-train-subset/loss": 0.7130602819304313, "eval/hcp-val/loss": 0.70523198381547}
+{"epoch": 26, "train/lr": 3.295647408362393e-05, "train/grad": 0.31687828264886025, "train/loss": 0.6975533834075928, "eval/hcp-train-subset/loss": 0.7111225474265314, "eval/hcp-val/loss": 0.7035650755128553}
+{"epoch": 27, "train/lr": 3.254416116379294e-05, "train/grad": 0.31702788980155633, "train/loss": 0.6959725891113281, "eval/hcp-train-subset/loss": 0.7113673052480144, "eval/hcp-val/loss": 0.7029983487821394}
+{"epoch": 28, "train/lr": 3.211676453125335e-05, "train/grad": 0.31785055456704503, "train/loss": 0.6939874038982391, "eval/hcp-train-subset/loss": 0.7097854306620937, "eval/hcp-val/loss": 0.7037077478824123}
+{"epoch": 29, "train/lr": 3.1674751537947714e-05, "train/grad": 0.3272139544393289, "train/loss": 0.6893004694080352, "eval/hcp-train-subset/loss": 0.7089365240066282, "eval/hcp-val/loss": 0.7027350885252799}
+{"epoch": 30, "train/lr": 3.1218605518594614e-05, "train/grad": 0.3309577675162076, "train/loss": 0.6897910982227325, "eval/hcp-train-subset/loss": 0.7083662786791401, "eval/hcp-val/loss": 0.7029534020731526}
+{"epoch": 31, "train/lr": 3.074882526216893e-05, "train/grad": 0.33356781011202513, "train/loss": 0.6880362207221985, "eval/hcp-train-subset/loss": 0.710823162909477, "eval/hcp-val/loss": 0.7042398596963575}
+{"epoch": 32, "train/lr": 3.0265924466484304e-05, "train/grad": 0.3316847203597271, "train/loss": 0.6895733369255066, "eval/hcp-train-subset/loss": 0.7097758525802244, "eval/hcp-val/loss": 0.7037569351734654}
+{"epoch": 33, "train/lr": 2.977043117647054e-05, "train/grad": 0.3208396506248445, "train/loss": 0.688268856306076, "eval/hcp-train-subset/loss": 0.7084683199082652, "eval/hcp-val/loss": 0.7040522406178136}
+{"epoch": 34, "train/lr": 2.9262887206765643e-05, "train/grad": 0.3336025366825887, "train/loss": 0.6858533309650421, "eval/hcp-train-subset/loss": 0.7081845931468471, "eval/hcp-val/loss": 0.7020506378143064}
+{"epoch": 35, "train/lr": 2.8743847549248834e-05, "train/grad": 0.32760474263753475, "train/loss": 0.6880840663051605, "eval/hcp-train-subset/loss": 0.7067941946368064, "eval/hcp-val/loss": 0.702783836472419}
+{"epoch": 36, "train/lr": 2.821387976616516e-05, "train/grad": 0.3350355001317037, "train/loss": 0.6861513663291932, "eval/hcp-train-subset/loss": 0.7081379842373633, "eval/hcp-val/loss": 0.7026558287682072}
+{"epoch": 37, "train/lr": 2.7673563369504416e-05, "train/grad": 0.33524440620610363, "train/loss": 0.6858542539978028, "eval/hcp-train-subset/loss": 0.7058599523959621, "eval/hcp-val/loss": 0.6989784615655099}
+{"epoch": 38, "train/lr": 2.7123489187313515e-05, "train/grad": 0.34380480698610244, "train/loss": 0.6816445820236207, "eval/hcp-train-subset/loss": 0.7045277184055697, "eval/hcp-val/loss": 0.6996139845540447}
+{"epoch": 39, "train/lr": 2.6564258717634662e-05, "train/grad": 0.34406968581737546, "train/loss": 0.6845705462360382, "eval/hcp-train-subset/loss": 0.7047922332440654, "eval/hcp-val/loss": 0.7015491185649749}
+{"epoch": 40, "train/lr": 2.5996483470775956e-05, "train/grad": 0.3355967952849408, "train/loss": 0.6845515926265716, "eval/hcp-train-subset/loss": 0.7049455354290624, "eval/hcp-val/loss": 0.7015802312281824}
+{"epoch": 41, "train/lr": 2.5420784300633598e-05, "train/grad": 0.3422153883620951, "train/loss": 0.6831561270809173, "eval/hcp-train-subset/loss": 0.7025749298834032, "eval/hcp-val/loss": 0.697069639159787}
+{"epoch": 42, "train/lr": 2.4837790725798037e-05, "train/grad": 0.3484169677775326, "train/loss": 0.6804185666179657, "eval/hcp-train-subset/loss": 0.7032374316646207, "eval/hcp-val/loss": 0.6998570532568039}
+{"epoch": 43, "train/lr": 2.4248140241183872e-05, "train/grad": 0.35316575852769727, "train/loss": 0.6783222032356262, "eval/hcp-train-subset/loss": 0.7055750104688829, "eval/hcp-val/loss": 0.6999571284940166}
+{"epoch": 44, "train/lr": 2.36524776209388e-05, "train/grad": 0.3588160075925914, "train/loss": 0.6782692975234985, "eval/hcp-train-subset/loss": 0.703335226543488, "eval/hcp-val/loss": 0.6992921646564237}
+{"epoch": 45, "train/lr": 2.305145421339275e-05, "train/grad": 0.35886130435325175, "train/loss": 0.6787263076257706, "eval/hcp-train-subset/loss": 0.7026810146147205, "eval/hcp-val/loss": 0.6985472834879353}
+{"epoch": 46, "train/lr": 2.2445727228817648e-05, "train/grad": 0.370833120623531, "train/loss": 0.674669801120758, "eval/hcp-train-subset/loss": 0.7032893569238724, "eval/hcp-val/loss": 0.6974262133721383}
+{"epoch": 47, "train/lr": 2.1835959020777867e-05, "train/grad": 0.36264895949150744, "train/loss": 0.680853259677887, "eval/hcp-train-subset/loss": 0.7006202840035961, "eval/hcp-val/loss": 0.6979691165108834}
+{"epoch": 48, "train/lr": 2.1222816361856524e-05, "train/grad": 0.3528911003579669, "train/loss": 0.6756647857284546, "eval/hcp-train-subset/loss": 0.7013878889622227, "eval/hcp-val/loss": 0.6964491692281538}
+{"epoch": 49, "train/lr": 2.0606969714549076e-05, "train/grad": 0.35628366630869673, "train/loss": 0.6758407590007782, "eval/hcp-train-subset/loss": 0.7014905743060573, "eval/hcp-val/loss": 0.6975252464894326}
+{"epoch": 50, "train/lr": 1.9989092498122458e-05, "train/grad": 0.3668795470482218, "train/loss": 0.6762814596939087, "eval/hcp-train-subset/loss": 0.6998322038881241, "eval/hcp-val/loss": 0.6973843411091836}
+{"epoch": 51, "train/lr": 1.936986035224125e-05, "train/grad": 0.36420247071081685, "train/loss": 0.6770081124782562, "eval/hcp-train-subset/loss": 0.6999283288755724, "eval/hcp-val/loss": 0.6960841321176098}
+{"epoch": 52, "train/lr": 1.8749950398165728e-05, "train/grad": 0.3727717510938377, "train/loss": 0.6752724985408783, "eval/hcp-train-subset/loss": 0.6991962819330154, "eval/hcp-val/loss": 0.6959551236321849}
+{"epoch": 53, "train/lr": 1.8130040498328958e-05, "train/grad": 0.37164728342532655, "train/loss": 0.6739503014087677, "eval/hcp-train-subset/loss": 0.6991718709468842, "eval/hcp-val/loss": 0.6969349143966552}
+{"epoch": 54, "train/lr": 1.751080851510517e-05, "train/grad": 0.37264427514301024, "train/loss": 0.6749159032535553, "eval/hcp-train-subset/loss": 0.6989205864167982, "eval/hcp-val/loss": 0.6969586966499206}
+{"epoch": 55, "train/lr": 1.689293156957646e-05, "train/grad": 0.3792606828660035, "train/loss": 0.6748916484260559, "eval/hcp-train-subset/loss": 0.6966157065283868, "eval/hcp-val/loss": 0.6952348341864925}
+{"epoch": 56, "train/lr": 1.627708530111125e-05, "train/grad": 0.3790141936777419, "train/loss": 0.6783035042285919, "eval/hcp-train-subset/loss": 0.6983174229821851, "eval/hcp-val/loss": 0.6941245448204779}
+{"epoch": 57, "train/lr": 1.566394312856223e-05, "train/grad": 0.38432285797075105, "train/loss": 0.6749071781158448, "eval/hcp-train-subset/loss": 0.6954342549847018, "eval/hcp-val/loss": 0.6953892275210349}
+{"epoch": 58, "train/lr": 1.5054175513893184e-05, "train/grad": 0.38692361151235133, "train/loss": 0.6732528396224976, "eval/hcp-train-subset/loss": 0.6951014812915556, "eval/hcp-val/loss": 0.6941599768977011}
+{"epoch": 59, "train/lr": 1.444844922903809e-05, "train/grad": 0.3809004788644336, "train/loss": 0.6745267806816101, "eval/hcp-train-subset/loss": 0.6949481050814351, "eval/hcp-val/loss": 0.6931752437545408}
+{"epoch": 60, "train/lr": 1.3847426626796371e-05, "train/grad": 0.3827359745286109, "train/loss": 0.6744501443767548, "eval/hcp-train-subset/loss": 0.6949766710881264, "eval/hcp-val/loss": 0.6958676653523599}
+{"epoch": 61, "train/lr": 1.3251764916559323e-05, "train/grad": 0.38366791838042086, "train/loss": 0.6736010522270203, "eval/hcp-train-subset/loss": 0.6938982480956662, "eval/hcp-val/loss": 0.6940203895491939}
+{"epoch": 62, "train/lr": 1.2662115445661782e-05, "train/grad": 0.40649806928382753, "train/loss": 0.6698238531112671, "eval/hcp-train-subset/loss": 0.6929766439622448, "eval/hcp-val/loss": 0.6928124687364025}
+{"epoch": 63, "train/lr": 1.2079122987143025e-05, "train/grad": 0.4028031749116224, "train/loss": 0.6683569884300232, "eval/hcp-train-subset/loss": 0.6921698806747314, "eval/hcp-val/loss": 0.6921964633849359}
+{"epoch": 64, "train/lr": 1.1503425034696866e-05, "train/grad": 0.41001333235482784, "train/loss": 0.6674102717590332, "eval/hcp-train-subset/loss": 0.6874344454657647, "eval/hcp-val/loss": 0.6916894182082145}
+{"epoch": 65, "train/lr": 1.0935651105582284e-05, "train/grad": 0.4027227966861189, "train/loss": 0.6693318142032624, "eval/hcp-train-subset/loss": 0.691916145624653, "eval/hcp-val/loss": 0.6937962674325512}
+{"epoch": 66, "train/lr": 1.0376422052254546e-05, "train/grad": 0.41232374464585836, "train/loss": 0.6680827937841416, "eval/hcp-train-subset/loss": 0.6922092255084745, "eval/hcp-val/loss": 0.6923996892667585}
+{"epoch": 67, "train/lr": 9.82634938347307e-06, "train/grad": 0.41621980361793093, "train/loss": 0.6720570600700378, "eval/hcp-train-subset/loss": 0.6908039498713708, "eval/hcp-val/loss": 0.6922565054508948}
+{"epoch": 68, "train/lr": 9.286034595625e-06, "train/grad": 0.4279620386403676, "train/loss": 0.6639180937099457, "eval/hcp-train-subset/loss": 0.69130925113155, "eval/hcp-val/loss": 0.6931661223211596}
+{"epoch": 69, "train/lr": 8.756068514998114e-06, "train/grad": 0.43183043976666147, "train/loss": 0.6655102820110321, "eval/hcp-train-subset/loss": 0.6873442607541238, "eval/hcp-val/loss": 0.6919620440852258}
+{"epoch": 70, "train/lr": 8.237030651720665e-06, "train/grad": 0.4274061983521316, "train/loss": 0.667318902850151, "eval/hcp-train-subset/loss": 0.6885113100851735, "eval/hcp-val/loss": 0.6898962864952702}
+{"epoch": 71, "train/lr": 7.729488566075681e-06, "train/grad": 0.4312722896095771, "train/loss": 0.6679157211780548, "eval/hcp-train-subset/loss": 0.6901250548901097, "eval/hcp-val/loss": 0.6912191414063976}
+{"epoch": 72, "train/lr": 7.233997247882178e-06, "train/grad": 0.43071094500175205, "train/loss": 0.6669733953619004, "eval/hcp-train-subset/loss": 0.6897142539101262, "eval/hcp-val/loss": 0.6914407389779245}
+{"epoch": 73, "train/lr": 6.751098509621876e-06, "train/grad": 0.4412515031274307, "train/loss": 0.6673387251901627, "eval/hcp-train-subset/loss": 0.6881275301979434, "eval/hcp-val/loss": 0.6909299002539727}
+{"epoch": 74, "train/lr": 6.281320393975191e-06, "train/grad": 0.44053667582685985, "train/loss": 0.669289624967575, "eval/hcp-train-subset/loss": 0.6893632306206611, "eval/hcp-val/loss": 0.6904609078361142}
+{"epoch": 75, "train/lr": 5.825176596414363e-06, "train/grad": 0.43496521191061144, "train/loss": 0.6734399849414825, "eval/hcp-train-subset/loss": 0.6876642684782704, "eval/hcp-val/loss": 0.6914007817545245}
+{"epoch": 76, "train/lr": 5.3831659034850535e-06, "train/grad": 0.45597105717498726, "train/loss": 0.666034353427887, "eval/hcp-train-subset/loss": 0.6874219054175962, "eval/hcp-val/loss": 0.6907763577276661}
+{"epoch": 77, "train/lr": 4.955771647390373e-06, "train/grad": 0.4545531209543939, "train/loss": 0.6668989979457856, "eval/hcp-train-subset/loss": 0.6873618643130025, "eval/hcp-val/loss": 0.6890721840243186}
+{"epoch": 78, "train/lr": 4.543461177474475e-06, "train/grad": 0.4535468465716563, "train/loss": 0.6672739129829407, "eval/hcp-train-subset/loss": 0.6868328404042029, "eval/hcp-val/loss": 0.6883130573457287}
+{"epoch": 79, "train/lr": 4.146685349182691e-06, "train/grad": 0.44424814653001593, "train/loss": 0.6698996903419495, "eval/hcp-train-subset/loss": 0.6856929063796997, "eval/hcp-val/loss": 0.6891837254647286}
+{"epoch": 80, "train/lr": 3.7658780310578768e-06, "train/grad": 0.45109315363954733, "train/loss": 0.6720970470714569, "eval/hcp-train-subset/loss": 0.6844730290674395, "eval/hcp-val/loss": 0.6896578736843602}
+{"epoch": 81, "train/lr": 3.401455630311446e-06, "train/grad": 0.4698446486145258, "train/loss": 0.6708381364536286, "eval/hcp-train-subset/loss": 0.6862277378959041, "eval/hcp-val/loss": 0.6880580129161957}
+{"epoch": 82, "train/lr": 3.0538166374881993e-06, "train/grad": 0.464902892129143, "train/loss": 0.6687176738452911, "eval/hcp-train-subset/loss": 0.6849273722017964, "eval/hcp-val/loss": 0.6871779916747924}
+{"epoch": 83, "train/lr": 2.723341190722833e-06, "train/grad": 0.475637408705588, "train/loss": 0.6663692991352081, "eval/hcp-train-subset/loss": 0.6820439532879861, "eval/hcp-val/loss": 0.6873984432989552}
+{"epoch": 84, "train/lr": 2.410390660064364e-06, "train/grad": 0.4663630254138503, "train/loss": 0.6710951271533966, "eval/hcp-train-subset/loss": 0.6824239684689429, "eval/hcp-val/loss": 0.6886381641510995}
+{"epoch": 85, "train/lr": 2.115307252323394e-06, "train/grad": 0.48215299390170546, "train/loss": 0.6690464607810974, "eval/hcp-train-subset/loss": 0.681869472226789, "eval/hcp-val/loss": 0.6876782746084275}
+{"epoch": 86, "train/lr": 1.838413636873912e-06, "train/grad": 0.4709623700220057, "train/loss": 0.6709921840953826, "eval/hcp-train-subset/loss": 0.6812579699100987, "eval/hcp-val/loss": 0.6874729529503854}
+{"epoch": 87, "train/lr": 1.5800125928190943e-06, "train/grad": 0.48566645129081515, "train/loss": 0.6706987867164612, "eval/hcp-train-subset/loss": 0.6808112436725248, "eval/hcp-val/loss": 0.6863994098478748}
+{"epoch": 88, "train/lr": 1.3403866779068222e-06, "train/grad": 0.48721769308952273, "train/loss": 0.6707881694173813, "eval/hcp-train-subset/loss": 0.6788989988065535, "eval/hcp-val/loss": 0.6875241244992902}
+{"epoch": 89, "train/lr": 1.1197979195568888e-06, "train/grad": 0.47596169312127323, "train/loss": 0.6712960547351837, "eval/hcp-train-subset/loss": 0.678637828557722, "eval/hcp-val/loss": 0.6871613196788295}
+{"epoch": 90, "train/lr": 9.184875283379039e-07, "train/grad": 0.4665178111243748, "train/loss": 0.6741295422649384, "eval/hcp-train-subset/loss": 0.678850241245762, "eval/hcp-val/loss": 0.6865319869210643}
+{"epoch": 91, "train/lr": 7.366756342070463e-07, "train/grad": 0.483484634378312, "train/loss": 0.671332836265564, "eval/hcp-train-subset/loss": 0.6778488120725078, "eval/hcp-val/loss": 0.6861398364267042}
+{"epoch": 92, "train/lr": 5.745610458012273e-07, "train/grad": 0.500627132726895, "train/loss": 0.6735539878368377, "eval/hcp-train-subset/loss": 0.6781710722754078, "eval/hcp-val/loss": 0.6866355672959359}
+{"epoch": 93, "train/lr": 4.323210330427178e-07, "train/grad": 0.5063482916496345, "train/loss": 0.67325744638443, "eval/hcp-train-subset/loss": 0.6781107398771471, "eval/hcp-val/loss": 0.687229331462614}
+{"epoch": 94, "train/lr": 3.1011113329712343e-07, "train/grad": 0.49369753631464247, "train/loss": 0.6738336009216308, "eval/hcp-train-subset/loss": 0.6768757152941919, "eval/hcp-val/loss": 0.6858220792585804}
+{"epoch": 95, "train/lr": 2.080649812955481e-07, "train/grad": 0.5092636348629578, "train/loss": 0.6724010057258606, "eval/hcp-train-subset/loss": 0.6768554285649331, "eval/hcp-val/loss": 0.6860781677307621}
+{"epoch": 96, "train/lr": 1.2629416300698208e-07, "train/grad": 0.5040310152090891, "train/loss": 0.6712429201889039, "eval/hcp-train-subset/loss": 0.6755554810647042, "eval/hcp-val/loss": 0.6847032971920506}
+{"epoch": 97, "train/lr": 6.488809362067338e-08, "train/grad": 0.4952824810775004, "train/loss": 0.6759873937702179, "eval/hcp-train-subset/loss": 0.6758405931534306, "eval/hcp-val/loss": 0.6857471514132715}
+{"epoch": 98, "train/lr": 2.391391977194211e-08, "train/grad": 0.5098741576492987, "train/loss": 0.6739185288238525, "eval/hcp-train-subset/loss": 0.6769252640585746, "eval/hcp-val/loss": 0.685934986798994}
+{"epoch": 99, "train/lr": 3.4164461183156008e-09, "train/grad": 0.5314042434845451, "train/loss": 0.6743520196533204, "eval/hcp-train-subset/loss": 0.675778518761358, "eval/hcp-val/loss": 0.6860132419293926}