tkharisov7 commited on
Commit
58500c6
·
verified ·
1 Parent(s): 066777e

Upload folder using huggingface_hub

Browse files
Files changed (2) hide show
  1. lr_0p005/config.json +117 -0
  2. lr_0p005/metrics_epoch.csv +170 -0
lr_0p005/config.json ADDED
@@ -0,0 +1,117 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "args": {
3
+ "data_dir": "./data",
4
+ "dataset": "cifar10",
5
+ "dataset_kwargs": "{}",
6
+ "criterion": "ce",
7
+ "model": "resnet",
8
+ "model_kwargs": "{}",
9
+ "image_size": 32,
10
+ "patch_size": 4,
11
+ "embed_dim": 256,
12
+ "depth": 6,
13
+ "num_heads": 8,
14
+ "mlp_ratio": 4.0,
15
+ "dropout": 0.0,
16
+ "vit_preset": "default",
17
+ "epochs": 300,
18
+ "early_stop_converged": true,
19
+ "early_stop_patience": 12,
20
+ "early_stop_rel_tol": 0.0005,
21
+ "early_stop_abs_tol": 0.0001,
22
+ "early_stop_min_epochs": 30,
23
+ "batch_size": 8192,
24
+ "lr": 0.005,
25
+ "weight_decay": 0.0,
26
+ "grad_clip": 0.0,
27
+ "num_workers": 32,
28
+ "seed": 42,
29
+ "device": "cuda",
30
+ "deterministic": false,
31
+ "train_shuffle": false,
32
+ "dataset_cache": "ram",
33
+ "momentum": 0.0,
34
+ "optimizer": "sgd",
35
+ "adam_beta1": 0.9,
36
+ "adam_beta2": 0.999,
37
+ "adam_eps": 1e-08,
38
+ "mlp_width": 256,
39
+ "mlp_activation": "gelu",
40
+ "cnn_width": 32,
41
+ "cnn_activation": "gelu",
42
+ "resnet_width": 16,
43
+ "resnet_activation": "gelu",
44
+ "dataset_n": 50000,
45
+ "dataset_n_test": 1000,
46
+ "cf_cifar_classes": "10",
47
+ "sorting_vocab_size": 4,
48
+ "sorting_length": 8,
49
+ "sharpness_batches": 1,
50
+ "sharpness_batch_size": 0,
51
+ "eig_iters": 100,
52
+ "eig_warm_iters": 0,
53
+ "eig_refresh": 5,
54
+ "eig_frequency": null,
55
+ "eig_tol": 1e-10,
56
+ "trace_samples": 16,
57
+ "trace_refresh": 5,
58
+ "eig_method": "lobpcg",
59
+ "log_dir": "logs/resnet_cifar10_gd_lr_0p005",
60
+ "log_steps": false,
61
+ "log_every": 50,
62
+ "wandb_project": "loss-landscape",
63
+ "wandb_entity": null,
64
+ "wandb_name": "resnet-gd-lr-0.005",
65
+ "wandb_offline": false,
66
+ "wandb_disabled": false,
67
+ "debug": false,
68
+ "profile_sharpness": false,
69
+ "profile_epoch": 1,
70
+ "all": false,
71
+ "local_smoothness": true,
72
+ "local_smoothness_every": 5,
73
+ "eig": false,
74
+ "trace": false,
75
+ "checkpoint_every_epochs": 5,
76
+ "checkpoint_dir": null,
77
+ "save_final_checkpoint": true,
78
+ "save_optimizer_state": false,
79
+ "train_only": false,
80
+ "hf_push": true,
81
+ "hf_repo_id": "tkharisov7/resnet-fullbatch-gd-sweep",
82
+ "hf_repo_type": "model",
83
+ "hf_private": false,
84
+ "hf_token_env": "HF_TOKEN",
85
+ "hf_path_in_repo": "lr_0p005",
86
+ "hf_include_checkpoints": true,
87
+ "hf_checkpoints_as_branches": true,
88
+ "hf_checkpoint_branch_prefix": "lr-0p005-epoch-",
89
+ "list_models": false,
90
+ "list_datasets": false
91
+ },
92
+ "data_meta": {
93
+ "dataset_type": "torchvision_cifar",
94
+ "train_config": {
95
+ "data_dir": "./data",
96
+ "dataset": "cifar10",
97
+ "image_size": 32,
98
+ "batch_size": 8192,
99
+ "num_workers": 32,
100
+ "train_shuffle": false,
101
+ "pin_memory": true,
102
+ "dataset_cache": "ram",
103
+ "cache_device": "cpu"
104
+ },
105
+ "sharpness_config": {
106
+ "data_dir": "./data",
107
+ "dataset": "cifar10",
108
+ "image_size": 32,
109
+ "batch_size": 8192,
110
+ "num_workers": 32,
111
+ "train_shuffle": false,
112
+ "pin_memory": true,
113
+ "dataset_cache": "ram",
114
+ "cache_device": "cpu"
115
+ }
116
+ }
117
+ }
lr_0p005/metrics_epoch.csv ADDED
@@ -0,0 +1,170 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ epoch,loss,local_smoothness,max_eig,trace,grad_norm
2
+ 0,2.3025853974478587,2.2358601093292236,0.0,0.0,0.7141000032424927
3
+ 1,2.3025853974478587,2.2358601093292236,0.0,0.0,0.7141000032424927
4
+ 2,2.3000237260546004,2.2358601093292236,0.0,0.0,0.7098240852355957
5
+ 3,2.2974928447178433,2.2358601093292236,0.0,0.0,0.705694854259491
6
+ 4,2.2949910504477367,2.2358601093292236,0.0,0.0,0.7017277479171753
7
+ 5,2.292517219270979,2.135153293609619,0.0,0.0,0.6978868842124939
8
+ 6,2.2900702272142683,2.135153293609619,0.0,0.0,0.6941871047019958
9
+ 7,2.2876491206032887,2.135153293609619,0.0,0.0,0.6906062364578247
10
+ 8,2.285252945763724,2.135153293609619,0.0,0.0,0.6871404647827148
11
+ 9,2.282880578722273,2.135153293609619,0.0,0.0,0.6837828159332275
12
+ 10,2.2805310658046176,2.0295581817626953,0.0,0.0,0.6805695295333862
13
+ 11,2.278204543249948,2.0295581817626953,0.0,0.0,0.6774080395698547
14
+ 12,2.2758980137961253,2.0295581817626953,0.0,0.0,0.6743999719619751
15
+ 13,2.2736123289380754,2.0295581817626953,0.0,0.0,0.6714915037155151
16
+ 14,2.2713468755994524,2.0295581817626953,0.0,0.0,0.6686530709266663
17
+ 15,2.2691005298069546,1.931995153427124,0.0,0.0,0.6659135818481445
18
+ 16,2.266874756131853,1.931995153427124,0.0,0.0,0.6632252931594849
19
+ 17,2.26466281073434,1.931995153427124,0.0,0.0,0.660692036151886
20
+ 18,2.262468303952898,1.931995153427124,0.0,0.0,0.6582223773002625
21
+ 19,2.2602918488638744,1.931995153427124,0.0,0.0,0.6558142304420471
22
+ 20,2.2581284046173096,1.8383731842041016,0.0,0.0,0.6535186171531677
23
+ 21,2.2559837273189,1.8383731842041016,0.0,0.0,0.6512542366981506
24
+ 22,2.2538517883845737,1.8383731842041016,0.0,0.0,0.6490786075592041
25
+ 23,2.251734869820731,1.8383731842041016,0.0,0.0,0.6469637751579285
26
+ 24,2.249630962099348,1.8383731842041016,0.0,0.0,0.6449211835861206
27
+ 25,2.2475394180842807,1.7518352270126343,0.0,0.0,0.642931342124939
28
+ 26,2.2454637459346225,1.7518352270126343,0.0,0.0,0.6409814357757568
29
+ 27,2.2433977127075195,1.7518352270126343,0.0,0.0,0.6391183733940125
30
+ 28,2.241342680794852,1.7518352270126343,0.0,0.0,0.6373159885406494
31
+ 29,2.239302635192871,1.7518352270126343,0.0,0.0,0.6355343461036682
32
+ 30,2.2372744424002513,1.6734156608581543,0.0,0.0,0.6337972283363342
33
+ 31,2.2352539471217563,1.6734156608581543,0.0,0.0,0.6321338415145874
34
+ 32,2.233245985848563,1.6734156608581543,0.0,0.0,0.6304828524589539
35
+ 33,2.2312486512320384,1.6734156608581543,0.0,0.0,0.6288774609565735
36
+ 34,2.229260342461722,1.6734156608581543,0.0,0.0,0.6273267865180969
37
+ 35,2.2272797312055315,1.608928918838501,0.0,0.0,0.6258020997047424
38
+ 36,2.2253157751900807,1.608928918838501,0.0,0.0,0.6242772340774536
39
+ 37,2.2233564172472273,1.608928918838501,0.0,0.0,0.6228348016738892
40
+ 38,2.2214065279279436,1.608928918838501,0.0,0.0,0.6214076280593872
41
+ 39,2.219465902873448,1.608928918838501,0.0,0.0,0.6199925541877747
42
+ 40,2.2175347123827254,1.5483938455581665,0.0,0.0,0.6186236143112183
43
+ 41,2.2156100613730296,1.5483938455581665,0.0,0.0,0.6172664761543274
44
+ 42,2.2136951855250766,1.5483938455581665,0.0,0.0,0.6159298419952393
45
+ 43,2.2117933886391774,1.5483938455581665,0.0,0.0,0.6145972013473511
46
+ 44,2.2098924773080006,1.5483938455581665,0.0,0.0,0.6132972240447998
47
+ 45,2.2079999106270924,1.4897335767745972,0.0,0.0,0.6120378375053406
48
+ 46,2.2061173915863037,1.4897335767745972,0.0,0.0,0.6107847690582275
49
+ 47,2.2042504719325473,1.4897335767745972,0.0,0.0,0.609488844871521
50
+ 48,2.202383892876761,1.4897335767745972,0.0,0.0,0.6082617044448853
51
+ 49,2.20052238873073,1.4897335767745972,0.0,0.0,0.6070461273193359
52
+ 50,2.1986733845302036,1.4454704523086548,0.0,0.0,0.6058071255683899
53
+ 51,2.196823699133737,1.4454704523086548,0.0,0.0,0.6046218872070312
54
+ 52,2.1949963910239085,1.4454704523086548,0.0,0.0,0.6033784747123718
55
+ 53,2.1931637014661516,1.4454704523086548,0.0,0.0,0.6021950840950012
56
+ 54,2.1913465772356306,1.4454704523086548,0.0,0.0,0.6009792685508728
57
+ 55,2.1895300320216586,1.3999360799789429,0.0,0.0,0.5998073220252991
58
+ 56,2.187718459538051,1.3999360799789429,0.0,0.0,0.5986286401748657
59
+ 57,2.185917241232736,1.3999360799789429,0.0,0.0,0.5974422693252563
60
+ 58,2.184128829411098,1.3999360799789429,0.0,0.0,0.5962486267089844
61
+ 59,2.1823412009647916,1.3999360799789429,0.0,0.0,0.595075249671936
62
+ 60,2.1805634839194163,1.363907814025879,0.0,0.0,0.5938907861709595
63
+ 61,2.178795371736799,1.363907814025879,0.0,0.0,0.5926961898803711
64
+ 62,2.177029711859567,1.363907814025879,0.0,0.0,0.5914927124977112
65
+ 63,2.1752738612038747,1.363907814025879,0.0,0.0,0.5902932286262512
66
+ 64,2.1735288075038364,1.363907814025879,0.0,0.0,0.5890975594520569
67
+ 65,2.1717838559831892,1.3281036615371704,0.0,0.0,0.5878918170928955
68
+ 66,2.1700524943215505,1.3281036615371704,0.0,0.0,0.5866866111755371
69
+ 67,2.168320758002145,1.3281036615371704,0.0,0.0,0.5854839086532593
70
+ 68,2.166610070637294,1.3281036615371704,0.0,0.0,0.5842536091804504
71
+ 69,2.1648852484566823,1.3281036615371704,0.0,0.0,0.5830514430999756
72
+ 70,2.163188798086984,1.3208885192871094,0.0,0.0,0.5818049311637878
73
+ 71,2.161487238747733,1.3208885192871094,0.0,0.0,0.5805606245994568
74
+ 72,2.159791742052351,1.3208885192871094,0.0,0.0,0.5793185830116272
75
+ 73,2.158107723508562,1.3208885192871094,0.0,0.0,0.5780618190765381
76
+ 74,2.1564315387180875,1.3208885192871094,0.0,0.0,0.5767974853515625
77
+ 75,2.1547651631491527,1.2850087881088257,0.0,0.0,0.5755208134651184
78
+ 76,2.1531037943703786,1.2850087881088257,0.0,0.0,0.5742555856704712
79
+ 77,2.1514528819492886,1.2850087881088257,0.0,0.0,0.5729508399963379
80
+ 78,2.1498031616210938,1.2850087881088257,0.0,0.0,0.571647584438324
81
+ 79,2.1481611047472273,1.2850087881088257,0.0,0.0,0.5703531503677368
82
+ 80,2.14653526033674,1.2819403409957886,0.0,0.0,0.5690321922302246
83
+ 81,2.144913945879255,1.2819403409957886,0.0,0.0,0.5677058100700378
84
+ 82,2.1432860919407437,1.2819403409957886,0.0,0.0,0.5663838386535645
85
+ 83,2.1416804449898854,1.2819403409957886,0.0,0.0,0.5650356411933899
86
+ 84,2.1400816781180247,1.2819403409957886,0.0,0.0,0.5636682510375977
87
+ 85,2.1384896550859724,1.2795381546020508,0.0,0.0,0.5623053908348083
88
+ 86,2.1369054658072337,1.2795381546020508,0.0,0.0,0.560926079750061
89
+ 87,2.1353304386138916,1.2795381546020508,0.0,0.0,0.5595468878746033
90
+ 88,2.133760520390102,1.2795381546020508,0.0,0.0,0.5581508874893188
91
+ 89,2.132202454975673,1.2795381546020508,0.0,0.0,0.5567569732666016
92
+ 90,2.1306412560599193,1.256364107131958,0.0,0.0,0.5553502440452576
93
+ 91,2.129103490284511,1.256364107131958,0.0,0.0,0.5539290904998779
94
+ 92,2.1275566646030972,1.256364107131958,0.0,0.0,0.55250483751297
95
+ 93,2.1260369505201067,1.256364107131958,0.0,0.0,0.5510669946670532
96
+ 94,2.1244995253426686,1.256364107131958,0.0,0.0,0.5496078729629517
97
+ 95,2.12299336705889,1.2720335721969604,0.0,0.0,0.5481586456298828
98
+ 96,2.121490342276437,1.2720335721969604,0.0,0.0,0.546687662601471
99
+ 97,2.1199916090284074,1.2720335721969604,0.0,0.0,0.545239269733429
100
+ 98,2.118504626410348,1.2720335721969604,0.0,0.0,0.5437584519386292
101
+ 99,2.1170271124158586,1.2720335721969604,0.0,0.0,0.5422608256340027
102
+ 100,2.115548236029489,1.2254526615142822,0.0,0.0,0.5407824516296387
103
+ 101,2.1140879903520857,1.2254526615142822,0.0,0.0,0.5393025875091553
104
+ 102,2.112626756940569,1.2254526615142822,0.0,0.0,0.5377830862998962
105
+ 103,2.1111771379198347,1.2254526615142822,0.0,0.0,0.5362664461135864
106
+ 104,2.1097400188446045,1.2254526615142822,0.0,0.0,0.5347505807876587
107
+ 105,2.1083037853240967,1.2275124788284302,0.0,0.0,0.5332365036010742
108
+ 106,2.1068853650774275,1.2275124788284302,0.0,0.0,0.5317116379737854
109
+ 107,2.1054602691105435,1.2275124788284302,0.0,0.0,0.5301925539970398
110
+ 108,2.104061876024519,1.2275124788284302,0.0,0.0,0.5286245346069336
111
+ 109,2.10266045161656,1.2275124788284302,0.0,0.0,0.5270918011665344
112
+ 110,2.101273400442941,1.2259693145751953,0.0,0.0,0.5255476236343384
113
+ 111,2.0998948982783725,1.2259693145751953,0.0,0.0,0.5240098237991333
114
+ 112,2.0985117639814104,1.2259693145751953,0.0,0.0,0.5224480628967285
115
+ 113,2.0971417086465016,1.2259693145751953,0.0,0.0,0.5208921432495117
116
+ 114,2.0957865715026855,1.2259693145751953,0.0,0.0,0.51932692527771
117
+ 115,2.094431059701102,1.2240465879440308,0.0,0.0,0.5177609324455261
118
+ 116,2.0931010246276855,1.2240465879440308,0.0,0.0,0.5161899328231812
119
+ 117,2.091764824731009,1.2240465879440308,0.0,0.0,0.5146190524101257
120
+ 118,2.0904359817504883,1.2240465879440308,0.0,0.0,0.5130354762077332
121
+ 119,2.089118276323591,1.2240465879440308,0.0,0.0,0.5114806890487671
122
+ 120,2.087810584477016,1.225411295890808,0.0,0.0,0.509895920753479
123
+ 121,2.0865061623709544,1.225411295890808,0.0,0.0,0.5083146691322327
124
+ 122,2.0852196557181224,1.225411295890808,0.0,0.0,0.5067394375801086
125
+ 123,2.0839291981288364,1.225411295890808,0.0,0.0,0.5051525831222534
126
+ 124,2.0826611518859863,1.225411295890808,0.0,0.0,0.503578245639801
127
+ 125,2.08138370513916,1.1863303184509277,0.0,0.0,0.5019927620887756
128
+ 126,2.080125536237444,1.1863303184509277,0.0,0.0,0.5004093050956726
129
+ 127,2.0788757460457936,1.1863303184509277,0.0,0.0,0.49882829189300537
130
+ 128,2.0776283059801375,1.1863303184509277,0.0,0.0,0.49723339080810547
131
+ 129,2.0763910838535855,1.1863303184509277,0.0,0.0,0.4956563413143158
132
+ 130,2.07516234261649,1.1765127182006836,0.0,0.0,0.49407583475112915
133
+ 131,2.073941264833723,1.1765127182006836,0.0,0.0,0.4924870431423187
134
+ 132,2.0727226052965437,1.1765127182006836,0.0,0.0,0.4908667206764221
135
+ 133,2.0715180465153287,1.1765127182006836,0.0,0.0,0.4892737567424774
136
+ 134,2.070324557168143,1.1765127182006836,0.0,0.0,0.4876975119113922
137
+ 135,2.0691348143986295,1.1735858917236328,0.0,0.0,0.48611873388290405
138
+ 136,2.0679495334625244,1.1735858917236328,0.0,0.0,0.48452842235565186
139
+ 137,2.0667817252022878,1.1735858917236328,0.0,0.0,0.48295870423316956
140
+ 138,2.0656093869890486,1.1735858917236328,0.0,0.0,0.48134151101112366
141
+ 139,2.0644491740635464,1.1735858917236328,0.0,0.0,0.47976046800613403
142
+ 140,2.0632990087781633,1.1621063947677612,0.0,0.0,0.47817882895469666
143
+ 141,2.0621513979775563,1.1621063947677612,0.0,0.0,0.47659242153167725
144
+ 142,2.061021055494036,1.1621063947677612,0.0,0.0,0.47501033544540405
145
+ 143,2.059889657156808,1.1621063947677612,0.0,0.0,0.47341006994247437
146
+ 144,2.0587706565856934,1.1621063947677612,0.0,0.0,0.47182947397232056
147
+ 145,2.0576548916952953,1.1518938541412354,0.0,0.0,0.47025221586227417
148
+ 146,2.0565509455544606,1.1518938541412354,0.0,0.0,0.46867817640304565
149
+ 147,2.055452993937901,1.1518938541412354,0.0,0.0,0.4670979678630829
150
+ 148,2.0543621948787143,1.1518938541412354,0.0,0.0,0.4655107259750366
151
+ 149,2.0532752445765903,1.1518938541412354,0.0,0.0,0.4639209806919098
152
+ 150,2.052203689302717,1.1420392990112305,0.0,0.0,0.46235546469688416
153
+ 151,2.051131316593715,1.1420392990112305,0.0,0.0,0.4607672095298767
154
+ 152,2.050067799431937,1.1420392990112305,0.0,0.0,0.45920616388320923
155
+ 153,2.049022606440953,1.1420392990112305,0.0,0.0,0.4576500952243805
156
+ 154,2.0479680470057895,1.1420392990112305,0.0,0.0,0.4560665488243103
157
+ 155,2.04692428452628,1.151491403579712,0.0,0.0,0.4545044004917145
158
+ 156,2.0458905356270924,1.151491403579712,0.0,0.0,0.4529186189174652
159
+ 157,2.0448694229125977,1.151491403579712,0.0,0.0,0.4513695240020752
160
+ 158,2.0438489573342458,1.151491403579712,0.0,0.0,0.44979241490364075
161
+ 159,2.0428435461861745,1.151491403579712,0.0,0.0,0.44824448227882385
162
+ 160,2.0418271677834645,1.1125946044921875,0.0,0.0,0.4466685652732849
163
+ 161,2.0408336775643483,1.1125946044921875,0.0,0.0,0.4451257884502411
164
+ 162,2.0398451260157993,1.1125946044921875,0.0,0.0,0.4435855448246002
165
+ 163,2.038860661642892,1.1125946044921875,0.0,0.0,0.4420276880264282
166
+ 164,2.0378808634621755,1.1125946044921875,0.0,0.0,0.44047263264656067
167
+ 165,2.036912270954677,1.1083577871322632,0.0,0.0,0.4389345347881317
168
+ 166,2.035951954977853,1.1083577871322632,0.0,0.0,0.43738991022109985
169
+ 167,2.0349923201969693,1.1083577871322632,0.0,0.0,0.43583154678344727
170
+ 168,2.034046036856515,1.1083577871322632,0.0,0.0,0.434312641620636