andrew-healey commited on
Commit
e58c6e9
·
verified ·
1 Parent(s): ba4e120

Upload folder using huggingface_hub

Browse files
logs/64_head_dim_two_latent_masks/2_latent_mask_seed_1339/args.json ADDED
@@ -0,0 +1 @@
 
 
1
+ {"hellaswag": true, "attention_kind": "selective", "log_dir": "logs/64_head_dim_two_latent_masks/2_latent_mask_seed_1339", "resume_checkpoint": null, "resume_optimizer": false, "add_a_head": false, "add_head_to_start": true, "new_head_init": "normal", "n_heads": 12, "n_embd": 768, "head_dim": 64, "protect_bos_token": true, "prevent_from_masking_myself": true, "max_steps": 4375, "warmup_steps": 250, "group": "64_head_dim_two_latent_masks", "use_wandb": true, "kill_self_after_run": false, "random_seed": 1339, "memory_penalty_epsilon": 0.1, "selection_head_linear_combo": "n_latent_masks", "selection_head_linear_combo_scale": 1.0, "disable_selection_head_linear_combo_bias": false, "assert_latent_matches_no_head": false, "n_latent_masks": 2, "init_latent_masks_to_identity": true, "init_latent_masks_to_inverse": false, "latent_mask_scale": null, "latent_mask_runtime_multiplier": null, "latent_mask_sigmoid": false, "S_layernorm": false, "one_head_per_latent_mask": false, "protection_kind": "none", "leaky_relu_alpha": null, "leaky_relu_bias": null, "use_compile": false, "use_mini_model": false, "upload_to_hf": true, "seq_len": 256, "batch_size": 16, "total_batch_size": 131072, "protection_head_scaling_factor": 1.0, "protection_head_bias": 0.0, "n_sliced_masks": null, "mask_layernorm": false, "residual_attention_masks": false, "compute_base_shapes": false, "base_shapes_savefile": null, "mup": true, "disable_selection": false, "mup_enable_coord_check_logging": false, "max_lr": 0.003, "decay_lr": true, "readout_zero_init": false, "query_zero_init": false, "l1_loss": false, "debugpy": false, "key": "2_latent_mask", "latent_mask_precision": "float32"}
logs/64_head_dim_two_latent_masks/2_latent_mask_seed_1339/dataloader_04374.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:994809a5ef065b3b39ff8d94ca69f209f3eedb103954655f09fa348a0918b20e
3
+ size 964
logs/64_head_dim_two_latent_masks/2_latent_mask_seed_1339/log2.txt ADDED
@@ -0,0 +1,548 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ max_steps: 4375
2
+ 0 val loss 10.8752
3
+ 0 val perplexity 52848.6719
4
+ 0 hella 0.2578
5
+ 0 train 10.874926 (lr=4.1958e-06) (hash(x)=19499830)
6
+ 10 train 9.601612 (lr=4.6154e-05) (hash(x)=19266129)
7
+ 20 train 9.057961 (lr=8.8112e-05) (hash(x)=17425714)
8
+ 30 train 8.415224 (lr=1.3007e-04) (hash(x)=19434488)
9
+ 40 train 7.784925 (lr=1.7203e-04) (hash(x)=18289960)
10
+ 50 train 7.675522 (lr=2.1399e-04) (hash(x)=25963856)
11
+ 60 train 7.695436 (lr=2.5594e-04) (hash(x)=23798061)
12
+ 70 train 7.394673 (lr=2.9790e-04) (hash(x)=19309727)
13
+ 80 train 7.279800 (lr=3.3986e-04) (hash(x)=20092642)
14
+ 90 train 7.047365 (lr=3.8182e-04) (hash(x)=21094807)
15
+ 100 val loss 6.9673
16
+ 100 val perplexity 1061.3724
17
+ 100 train 6.817717 (lr=4.2378e-04) (hash(x)=18910383)
18
+ 110 train 6.732147 (lr=4.6573e-04) (hash(x)=17739111)
19
+ 120 train 6.590297 (lr=5.0769e-04) (hash(x)=20068667)
20
+ 130 train 6.440167 (lr=5.4965e-04) (hash(x)=19548561)
21
+ 140 train 6.414361 (lr=5.9161e-04) (hash(x)=19990057)
22
+ 150 train 6.487254 (lr=6.3357e-04) (hash(x)=17559733)
23
+ 160 train 6.403044 (lr=6.7552e-04) (hash(x)=21295815)
24
+ 170 train 6.300015 (lr=7.1748e-04) (hash(x)=21771608)
25
+ 180 train 6.228881 (lr=7.5944e-04) (hash(x)=18484596)
26
+ 190 train 6.139716 (lr=8.0140e-04) (hash(x)=18663948)
27
+ 200 val loss 6.0694
28
+ 200 val perplexity 432.4034
29
+ 200 train 6.085604 (lr=8.4336e-04) (hash(x)=19306031)
30
+ 210 train 5.968075 (lr=8.8531e-04) (hash(x)=18345886)
31
+ 220 train 5.887372 (lr=9.2727e-04) (hash(x)=18477726)
32
+ 230 train 6.008586 (lr=9.6923e-04) (hash(x)=22315538)
33
+ 240 train 5.700725 (lr=1.0112e-03) (hash(x)=16529598)
34
+ 250 hella 0.2387
35
+ 250 train 5.774494 (lr=1.0531e-03) (hash(x)=23415380)
36
+ 260 train 5.648863 (lr=1.0951e-03) (hash(x)=21452526)
37
+ 270 train 5.595823 (lr=1.1371e-03) (hash(x)=16964237)
38
+ 280 train 5.517642 (lr=1.1790e-03) (hash(x)=20851273)
39
+ 290 train 5.451559 (lr=1.2210e-03) (hash(x)=14094390)
40
+ 300 val loss 5.5591
41
+ 300 val perplexity 259.5919
42
+ 300 train 5.306190 (lr=1.2629e-03) (hash(x)=21184674)
43
+ 310 train 5.248900 (lr=1.3049e-03) (hash(x)=18777755)
44
+ 320 train 5.361280 (lr=1.3469e-03) (hash(x)=23869658)
45
+ 330 train 5.445915 (lr=1.3888e-03) (hash(x)=18147471)
46
+ 340 train 5.507410 (lr=1.4308e-03) (hash(x)=19672778)
47
+ 350 train 5.470810 (lr=1.4727e-03) (hash(x)=21868542)
48
+ 360 train 5.391451 (lr=1.5147e-03) (hash(x)=20606890)
49
+ 370 train 5.455467 (lr=1.5566e-03) (hash(x)=22900825)
50
+ 380 train 5.247479 (lr=1.5986e-03) (hash(x)=20332562)
51
+ 390 train 5.203438 (lr=1.6406e-03) (hash(x)=16917496)
52
+ 400 val loss 5.2265
53
+ 400 val perplexity 186.1392
54
+ 400 train 5.195853 (lr=1.6825e-03) (hash(x)=20419326)
55
+ 410 train 5.150535 (lr=1.7245e-03) (hash(x)=18934133)
56
+ 420 train 5.164592 (lr=1.7664e-03) (hash(x)=17991367)
57
+ 430 train 5.175559 (lr=1.8084e-03) (hash(x)=17769433)
58
+ 440 train 5.071062 (lr=1.8503e-03) (hash(x)=21720430)
59
+ 450 train 5.068007 (lr=1.8923e-03) (hash(x)=14866747)
60
+ 460 train 4.957835 (lr=1.9343e-03) (hash(x)=22009033)
61
+ 470 train 4.750209 (lr=1.9762e-03) (hash(x)=19327881)
62
+ 480 train 4.912211 (lr=2.0182e-03) (hash(x)=19973633)
63
+ 490 train 4.695123 (lr=2.0601e-03) (hash(x)=20294459)
64
+ 500 val loss 5.0014
65
+ 500 val perplexity 148.6156
66
+ 500 hella 0.2382
67
+ 500 train 4.771401 (lr=2.1021e-03) (hash(x)=16902401)
68
+ 510 train 4.893031 (lr=2.1441e-03) (hash(x)=20028937)
69
+ 520 train 4.950193 (lr=2.1860e-03) (hash(x)=18001903)
70
+ 530 train 5.020234 (lr=2.2280e-03) (hash(x)=21435244)
71
+ 540 train 4.947384 (lr=2.2699e-03) (hash(x)=16973801)
72
+ 550 train 4.884551 (lr=2.3119e-03) (hash(x)=21158829)
73
+ 560 train 4.945761 (lr=2.3538e-03) (hash(x)=15382082)
74
+ 570 train 4.871669 (lr=2.3958e-03) (hash(x)=18999843)
75
+ 580 train 4.807076 (lr=2.4378e-03) (hash(x)=19873877)
76
+ 590 train 4.882966 (lr=2.4797e-03) (hash(x)=22520351)
77
+ 600 val loss 4.8208
78
+ 600 val perplexity 124.0684
79
+ 600 train 4.839802 (lr=2.5217e-03) (hash(x)=18681131)
80
+ 610 train 4.736097 (lr=2.5636e-03) (hash(x)=18401442)
81
+ 620 train 4.715189 (lr=2.6056e-03) (hash(x)=21171777)
82
+ 630 train 4.616986 (lr=2.6476e-03) (hash(x)=20257940)
83
+ 640 train 4.616564 (lr=2.6895e-03) (hash(x)=21673710)
84
+ 650 train 4.563687 (lr=2.7315e-03) (hash(x)=16969297)
85
+ 660 train 4.518816 (lr=2.7734e-03) (hash(x)=18464518)
86
+ 670 train 4.472268 (lr=2.8154e-03) (hash(x)=17196302)
87
+ 680 train 4.368213 (lr=2.8573e-03) (hash(x)=18376792)
88
+ 690 train 4.405716 (lr=2.8993e-03) (hash(x)=17928860)
89
+ 700 val loss 4.6541
90
+ 700 val perplexity 105.0163
91
+ 700 train 4.378140 (lr=2.9413e-03) (hash(x)=22330314)
92
+ 710 train 4.698320 (lr=2.9832e-03) (hash(x)=17776819)
93
+ 720 train 4.675519 (lr=3.0000e-03) (hash(x)=20367943)
94
+ 730 train 4.668781 (lr=2.9999e-03) (hash(x)=20244235)
95
+ 740 train 4.605052 (lr=2.9997e-03) (hash(x)=25068819)
96
+ 750 hella 0.2529
97
+ 750 train 4.549994 (lr=2.9994e-03) (hash(x)=20552971)
98
+ 760 train 4.620965 (lr=2.9990e-03) (hash(x)=19170549)
99
+ 770 train 4.551469 (lr=2.9985e-03) (hash(x)=20863809)
100
+ 780 train 4.568525 (lr=2.9979e-03) (hash(x)=23003114)
101
+ 790 train 4.519841 (lr=2.9972e-03) (hash(x)=21939064)
102
+ 800 val loss 4.5255
103
+ 800 val perplexity 92.3399
104
+ 800 train 4.500484 (lr=2.9964e-03) (hash(x)=19245118)
105
+ 810 train 4.402371 (lr=2.9955e-03) (hash(x)=22642355)
106
+ 820 train 4.429706 (lr=2.9945e-03) (hash(x)=23447029)
107
+ 830 train 4.486634 (lr=2.9934e-03) (hash(x)=21708229)
108
+ 840 train 4.399207 (lr=2.9922e-03) (hash(x)=18377450)
109
+ 850 train 4.397586 (lr=2.9909e-03) (hash(x)=16701034)
110
+ 860 train 4.535068 (lr=2.9896e-03) (hash(x)=21348338)
111
+ 870 train 4.490294 (lr=2.9881e-03) (hash(x)=20919606)
112
+ 880 train 4.458325 (lr=2.9865e-03) (hash(x)=22912154)
113
+ 890 train 4.412510 (lr=2.9848e-03) (hash(x)=20629816)
114
+ 900 val loss 4.4192
115
+ 900 val perplexity 83.0311
116
+ 900 train 4.403420 (lr=2.9830e-03) (hash(x)=21120493)
117
+ 910 train 4.386687 (lr=2.9811e-03) (hash(x)=19354333)
118
+ 920 train 4.434529 (lr=2.9792e-03) (hash(x)=22251784)
119
+ 930 train 4.356070 (lr=2.9771e-03) (hash(x)=17462001)
120
+ 940 train 4.312666 (lr=2.9749e-03) (hash(x)=21939624)
121
+ 950 train 4.351430 (lr=2.9726e-03) (hash(x)=20914464)
122
+ 960 train 4.364164 (lr=2.9703e-03) (hash(x)=18974331)
123
+ 970 train 4.309196 (lr=2.9678e-03) (hash(x)=21032787)
124
+ 980 train 4.285992 (lr=2.9652e-03) (hash(x)=18848940)
125
+ 990 train 4.339186 (lr=2.9626e-03) (hash(x)=20074037)
126
+ 1000 val loss 4.3204
127
+ 1000 val perplexity 75.2223
128
+ 1000 hella 0.2557
129
+ 1000 train 4.327174 (lr=2.9598e-03) (hash(x)=18791811)
130
+ 1010 train 4.496412 (lr=2.9570e-03) (hash(x)=16870030)
131
+ 1020 train 4.352316 (lr=2.9540e-03) (hash(x)=17382263)
132
+ 1030 train 4.335989 (lr=2.9510e-03) (hash(x)=19518020)
133
+ 1040 train 4.290406 (lr=2.9478e-03) (hash(x)=18479196)
134
+ 1050 train 4.304888 (lr=2.9446e-03) (hash(x)=20240434)
135
+ 1060 train 4.323289 (lr=2.9412e-03) (hash(x)=20176962)
136
+ 1070 train 4.326409 (lr=2.9378e-03) (hash(x)=22429292)
137
+ 1080 train 4.191626 (lr=2.9343e-03) (hash(x)=20635325)
138
+ 1090 train 4.247191 (lr=2.9307e-03) (hash(x)=23524462)
139
+ 1100 val loss 4.2473
140
+ 1100 val perplexity 69.9160
141
+ 1100 train 4.277756 (lr=2.9270e-03) (hash(x)=25636544)
142
+ 1110 train 4.195518 (lr=2.9231e-03) (hash(x)=21984040)
143
+ 1120 train 4.179825 (lr=2.9192e-03) (hash(x)=21955314)
144
+ 1130 train 4.256575 (lr=2.9152e-03) (hash(x)=24632576)
145
+ 1140 train 4.243897 (lr=2.9112e-03) (hash(x)=22566911)
146
+ 1150 train 4.361907 (lr=2.9070e-03) (hash(x)=19348391)
147
+ 1160 train 4.298800 (lr=2.9027e-03) (hash(x)=13676213)
148
+ 1170 train 4.261055 (lr=2.8983e-03) (hash(x)=21126914)
149
+ 1180 train 4.197734 (lr=2.8939e-03) (hash(x)=21859315)
150
+ 1190 train 4.197610 (lr=2.8893e-03) (hash(x)=22140691)
151
+ 1200 val loss 4.1740
152
+ 1200 val perplexity 64.9736
153
+ 1200 train 4.092720 (lr=2.8847e-03) (hash(x)=21010060)
154
+ 1210 train 4.173311 (lr=2.8800e-03) (hash(x)=25216648)
155
+ 1220 train 4.145973 (lr=2.8751e-03) (hash(x)=19447714)
156
+ 1230 train 4.094602 (lr=2.8702e-03) (hash(x)=20268415)
157
+ 1240 train 4.176362 (lr=2.8652e-03) (hash(x)=21742969)
158
+ 1250 hella 0.2546
159
+ 1250 train 4.036061 (lr=2.8601e-03) (hash(x)=23680287)
160
+ 1260 train 4.121371 (lr=2.8550e-03) (hash(x)=21568309)
161
+ 1270 train 4.063969 (lr=2.8497e-03) (hash(x)=20176939)
162
+ 1280 train 4.127291 (lr=2.8443e-03) (hash(x)=21657918)
163
+ 1290 train 4.161637 (lr=2.8389e-03) (hash(x)=20749696)
164
+ 1300 val loss 4.1165
165
+ 1300 val perplexity 61.3431
166
+ 1300 train 4.153109 (lr=2.8333e-03) (hash(x)=21178552)
167
+ 1310 train 4.277659 (lr=2.8277e-03) (hash(x)=24008496)
168
+ 1320 train 4.024350 (lr=2.8220e-03) (hash(x)=20529379)
169
+ 1330 train 4.100066 (lr=2.8162e-03) (hash(x)=19580783)
170
+ 1340 train 4.208639 (lr=2.8103e-03) (hash(x)=22031001)
171
+ 1350 train 4.024272 (lr=2.8044e-03) (hash(x)=19758453)
172
+ 1360 train 4.144370 (lr=2.7983e-03) (hash(x)=20524989)
173
+ 1370 train 4.033062 (lr=2.7922e-03) (hash(x)=18611857)
174
+ 1380 train 4.067211 (lr=2.7860e-03) (hash(x)=18097851)
175
+ 1390 train 4.096242 (lr=2.7797e-03) (hash(x)=18841925)
176
+ 1400 val loss 4.0935
177
+ 1400 val perplexity 59.9472
178
+ 1400 train 4.048141 (lr=2.7733e-03) (hash(x)=20959768)
179
+ 1410 train 4.202372 (lr=2.7668e-03) (hash(x)=18836875)
180
+ 1420 train 4.264233 (lr=2.7603e-03) (hash(x)=19657234)
181
+ 1430 train 4.022685 (lr=2.7536e-03) (hash(x)=21047683)
182
+ 1440 train 4.171171 (lr=2.7469e-03) (hash(x)=26743461)
183
+ 1450 train 4.098897 (lr=2.7401e-03) (hash(x)=22881324)
184
+ 1460 train 4.021103 (lr=2.7333e-03) (hash(x)=24194800)
185
+ 1470 train 4.063319 (lr=2.7263e-03) (hash(x)=23399930)
186
+ 1480 train 4.099722 (lr=2.7193e-03) (hash(x)=20569902)
187
+ 1490 train 4.012094 (lr=2.7121e-03) (hash(x)=18223528)
188
+ 1500 val loss 4.0417
189
+ 1500 val perplexity 56.9251
190
+ 1500 hella 0.2521
191
+ 1500 train 4.143293 (lr=2.7050e-03) (hash(x)=19550097)
192
+ 1510 train 4.014848 (lr=2.6977e-03) (hash(x)=19347158)
193
+ 1520 train 4.036420 (lr=2.6903e-03) (hash(x)=16014775)
194
+ 1530 train 3.952625 (lr=2.6829e-03) (hash(x)=19187039)
195
+ 1540 train 4.011573 (lr=2.6754e-03) (hash(x)=19718445)
196
+ 1550 train 4.054626 (lr=2.6678e-03) (hash(x)=20755662)
197
+ 1560 train 3.985566 (lr=2.6602e-03) (hash(x)=20026689)
198
+ 1570 train 4.055885 (lr=2.6525e-03) (hash(x)=20481658)
199
+ 1580 train 4.014137 (lr=2.6447e-03) (hash(x)=20415281)
200
+ 1590 train 4.139642 (lr=2.6368e-03) (hash(x)=21683290)
201
+ 1600 val loss 4.0153
202
+ 1600 val perplexity 55.4408
203
+ 1600 train 4.094338 (lr=2.6289e-03) (hash(x)=21287292)
204
+ 1610 train 4.178652 (lr=2.6208e-03) (hash(x)=20072672)
205
+ 1620 train 4.033861 (lr=2.6128e-03) (hash(x)=15774154)
206
+ 1630 train 4.030692 (lr=2.6046e-03) (hash(x)=19253426)
207
+ 1640 train 4.118616 (lr=2.5964e-03) (hash(x)=18235717)
208
+ 1650 train 4.015238 (lr=2.5881e-03) (hash(x)=17940066)
209
+ 1660 train 3.996684 (lr=2.5797e-03) (hash(x)=20731779)
210
+ 1670 train 4.000323 (lr=2.5713e-03) (hash(x)=21326011)
211
+ 1680 train 3.824627 (lr=2.5628e-03) (hash(x)=19569086)
212
+ 1690 train 3.873015 (lr=2.5542e-03) (hash(x)=17136153)
213
+ 1700 val loss 4.0154
214
+ 1700 val perplexity 55.4470
215
+ 1700 train 4.070844 (lr=2.5455e-03) (hash(x)=24947450)
216
+ 1710 train 4.050337 (lr=2.5368e-03) (hash(x)=20835959)
217
+ 1720 train 4.057382 (lr=2.5281e-03) (hash(x)=19803138)
218
+ 1730 train 4.020662 (lr=2.5192e-03) (hash(x)=23602127)
219
+ 1740 train 4.058943 (lr=2.5103e-03) (hash(x)=21588728)
220
+ 1750 hella 0.2588
221
+ 1750 train 4.012644 (lr=2.5014e-03) (hash(x)=22545660)
222
+ 1760 train 4.030592 (lr=2.4924e-03) (hash(x)=22464423)
223
+ 1770 train 3.886310 (lr=2.4833e-03) (hash(x)=22560125)
224
+ 1780 train 3.982537 (lr=2.4741e-03) (hash(x)=20628381)
225
+ 1790 train 3.995685 (lr=2.4649e-03) (hash(x)=19928492)
226
+ 1800 val loss 3.9661
227
+ 1800 val perplexity 52.7762
228
+ 1800 train 4.027374 (lr=2.4556e-03) (hash(x)=19855177)
229
+ 1810 train 3.950476 (lr=2.4463e-03) (hash(x)=17417243)
230
+ 1820 train 3.792014 (lr=2.4369e-03) (hash(x)=20821803)
231
+ 1830 train 3.927718 (lr=2.4275e-03) (hash(x)=24879137)
232
+ 1840 train 3.864922 (lr=2.4180e-03) (hash(x)=20844335)
233
+ 1850 train 3.913505 (lr=2.4084e-03) (hash(x)=18445411)
234
+ 1860 train 4.040217 (lr=2.3988e-03) (hash(x)=22194917)
235
+ 1870 train 4.067190 (lr=2.3891e-03) (hash(x)=21291954)
236
+ 1880 train 3.892105 (lr=2.3794e-03) (hash(x)=18252425)
237
+ 1890 train 3.996315 (lr=2.3696e-03) (hash(x)=17424764)
238
+ 1900 val loss 3.9280
239
+ 1900 val perplexity 50.8075
240
+ 1900 train 3.919645 (lr=2.3598e-03) (hash(x)=19759252)
241
+ 1910 train 3.857271 (lr=2.3499e-03) (hash(x)=17053540)
242
+ 1920 train 3.964164 (lr=2.3400e-03) (hash(x)=20059888)
243
+ 1930 train 3.812448 (lr=2.3300e-03) (hash(x)=18717513)
244
+ 1940 train 3.899640 (lr=2.3200e-03) (hash(x)=21252145)
245
+ 1950 train 3.886501 (lr=2.3099e-03) (hash(x)=17166480)
246
+ 1960 train 3.790443 (lr=2.2998e-03) (hash(x)=19791836)
247
+ 1970 train 3.961019 (lr=2.2896e-03) (hash(x)=20908476)
248
+ 1980 train 3.905114 (lr=2.2793e-03) (hash(x)=17252455)
249
+ 1990 train 3.912068 (lr=2.2691e-03) (hash(x)=19187120)
250
+ 2000 val loss 3.9042
251
+ 2000 val perplexity 49.6096
252
+ 2000 hella 0.2552
253
+ 2000 train 3.943335 (lr=2.2588e-03) (hash(x)=20996686)
254
+ 2010 train 3.952033 (lr=2.2484e-03) (hash(x)=22420273)
255
+ 2020 train 3.937759 (lr=2.2380e-03) (hash(x)=21035175)
256
+ 2030 train 3.966361 (lr=2.2275e-03) (hash(x)=21464100)
257
+ 2040 train 4.005024 (lr=2.2170e-03) (hash(x)=19332299)
258
+ 2050 train 3.861417 (lr=2.2065e-03) (hash(x)=16978105)
259
+ 2060 train 4.074130 (lr=2.1959e-03) (hash(x)=17517702)
260
+ 2070 train 3.921298 (lr=2.1853e-03) (hash(x)=19999968)
261
+ 2080 train 3.923304 (lr=2.1746e-03) (hash(x)=17974202)
262
+ 2090 train 3.763221 (lr=2.1639e-03) (hash(x)=18093857)
263
+ 2100 val loss 3.8777
264
+ 2100 val perplexity 48.3127
265
+ 2100 train 3.824392 (lr=2.1532e-03) (hash(x)=18025348)
266
+ 2110 train 3.856955 (lr=2.1424e-03) (hash(x)=18562385)
267
+ 2120 train 3.829606 (lr=2.1316e-03) (hash(x)=21660073)
268
+ 2130 train 3.950062 (lr=2.1208e-03) (hash(x)=23688255)
269
+ 2140 train 4.023461 (lr=2.1099e-03) (hash(x)=22201770)
270
+ 2150 train 3.925009 (lr=2.0990e-03) (hash(x)=19338752)
271
+ 2160 train 3.915506 (lr=2.0881e-03) (hash(x)=15027506)
272
+ 2170 train 3.901565 (lr=2.0771e-03) (hash(x)=16328757)
273
+ 2180 train 3.915625 (lr=2.0661e-03) (hash(x)=25769566)
274
+ 2190 train 3.923048 (lr=2.0550e-03) (hash(x)=21688947)
275
+ 2200 val loss 3.8570
276
+ 2200 val perplexity 47.3237
277
+ 2200 train 3.898511 (lr=2.0440e-03) (hash(x)=15500010)
278
+ 2210 train 3.901591 (lr=2.0329e-03) (hash(x)=17836120)
279
+ 2220 train 3.828845 (lr=2.0217e-03) (hash(x)=19443168)
280
+ 2230 train 3.751292 (lr=2.0106e-03) (hash(x)=17297013)
281
+ 2240 train 3.818310 (lr=1.9994e-03) (hash(x)=18650030)
282
+ 2250 hella 0.2553
283
+ 2250 train 3.777676 (lr=1.9882e-03) (hash(x)=15805613)
284
+ 2260 train 3.756160 (lr=1.9770e-03) (hash(x)=19942534)
285
+ 2270 train 3.768357 (lr=1.9657e-03) (hash(x)=22599430)
286
+ 2280 train 4.031559 (lr=1.9544e-03) (hash(x)=22218719)
287
+ 2290 train 3.927412 (lr=1.9431e-03) (hash(x)=19342048)
288
+ 2300 val loss 3.8332
289
+ 2300 val perplexity 46.2117
290
+ 2300 train 3.809177 (lr=1.9318e-03) (hash(x)=18810407)
291
+ 2310 train 3.883594 (lr=1.9205e-03) (hash(x)=22073167)
292
+ 2320 train 3.921335 (lr=1.9091e-03) (hash(x)=20883748)
293
+ 2330 train 3.775849 (lr=1.8977e-03) (hash(x)=15010975)
294
+ 2340 train 3.890758 (lr=1.8863e-03) (hash(x)=17153145)
295
+ 2350 train 3.817680 (lr=1.8749e-03) (hash(x)=25451452)
296
+ 2360 train 3.897727 (lr=1.8635e-03) (hash(x)=19036737)
297
+ 2370 train 3.947877 (lr=1.8520e-03) (hash(x)=18582711)
298
+ 2380 train 3.835198 (lr=1.8406e-03) (hash(x)=22311667)
299
+ 2390 train 3.897584 (lr=1.8291e-03) (hash(x)=23289834)
300
+ 2400 val loss 3.8216
301
+ 2400 val perplexity 45.6771
302
+ 2400 train 3.765887 (lr=1.8176e-03) (hash(x)=19343835)
303
+ 2410 train 3.669539 (lr=1.8061e-03) (hash(x)=19070868)
304
+ 2420 train 3.762300 (lr=1.7946e-03) (hash(x)=21492358)
305
+ 2430 train 3.698888 (lr=1.7830e-03) (hash(x)=21355438)
306
+ 2440 train 3.718751 (lr=1.7715e-03) (hash(x)=18688923)
307
+ 2450 train 3.556384 (lr=1.7600e-03) (hash(x)=18655016)
308
+ 2460 train 3.585841 (lr=1.7484e-03) (hash(x)=23966823)
309
+ 2470 train 3.542913 (lr=1.7368e-03) (hash(x)=19789602)
310
+ 2480 train 3.897294 (lr=1.7253e-03) (hash(x)=18904369)
311
+ 2490 train 3.890482 (lr=1.7137e-03) (hash(x)=18638893)
312
+ 2500 val loss 3.7961
313
+ 2500 val perplexity 44.5262
314
+ 2500 hella 0.2578
315
+ 2500 train 3.812621 (lr=1.7021e-03) (hash(x)=16815212)
316
+ 2510 train 3.757810 (lr=1.6906e-03) (hash(x)=24781988)
317
+ 2520 train 3.858606 (lr=1.6790e-03) (hash(x)=18249583)
318
+ 2530 train 3.800295 (lr=1.6674e-03) (hash(x)=18203346)
319
+ 2540 train 3.815923 (lr=1.6558e-03) (hash(x)=19029376)
320
+ 2550 train 3.786716 (lr=1.6442e-03) (hash(x)=21525059)
321
+ 2560 train 3.750834 (lr=1.6326e-03) (hash(x)=21419471)
322
+ 2570 train 3.799153 (lr=1.6210e-03) (hash(x)=19425528)
323
+ 2580 train 3.863064 (lr=1.6094e-03) (hash(x)=16653977)
324
+ 2590 train 3.754469 (lr=1.5979e-03) (hash(x)=23372914)
325
+ 2600 val loss 3.7750
326
+ 2600 val perplexity 43.5969
327
+ 2600 train 3.729886 (lr=1.5863e-03) (hash(x)=15676604)
328
+ 2610 train 3.713174 (lr=1.5747e-03) (hash(x)=24943464)
329
+ 2620 train 3.601118 (lr=1.5632e-03) (hash(x)=18637598)
330
+ 2630 train 3.546050 (lr=1.5516e-03) (hash(x)=20671881)
331
+ 2640 train 3.600868 (lr=1.5400e-03) (hash(x)=21040846)
332
+ 2650 train 3.654497 (lr=1.5285e-03) (hash(x)=22880651)
333
+ 2660 train 3.748867 (lr=1.5170e-03) (hash(x)=18865768)
334
+ 2670 train 3.935880 (lr=1.5054e-03) (hash(x)=21044077)
335
+ 2680 train 3.824387 (lr=1.4939e-03) (hash(x)=20980745)
336
+ 2690 train 3.766754 (lr=1.4824e-03) (hash(x)=19495821)
337
+ 2700 val loss 3.7605
338
+ 2700 val perplexity 42.9682
339
+ 2700 train 3.901769 (lr=1.4709e-03) (hash(x)=24282313)
340
+ 2710 train 3.917942 (lr=1.4594e-03) (hash(x)=16434484)
341
+ 2720 train 3.745023 (lr=1.4480e-03) (hash(x)=19401384)
342
+ 2730 train 3.673296 (lr=1.4365e-03) (hash(x)=29013935)
343
+ 2740 train 3.635990 (lr=1.4251e-03) (hash(x)=20089946)
344
+ 2750 hella 0.2644
345
+ 2750 train 3.772944 (lr=1.4137e-03) (hash(x)=19869457)
346
+ 2760 train 3.687065 (lr=1.4023e-03) (hash(x)=16709635)
347
+ 2770 train 3.706210 (lr=1.3909e-03) (hash(x)=22743096)
348
+ 2780 train 3.691381 (lr=1.3795e-03) (hash(x)=21850887)
349
+ 2790 train 3.656559 (lr=1.3682e-03) (hash(x)=21898297)
350
+ 2800 val loss 3.7390
351
+ 2800 val perplexity 42.0570
352
+ 2800 train 3.545223 (lr=1.3569e-03) (hash(x)=21413897)
353
+ 2810 train 3.564178 (lr=1.3456e-03) (hash(x)=19920845)
354
+ 2820 train 3.547002 (lr=1.3343e-03) (hash(x)=24043370)
355
+ 2830 train 3.619495 (lr=1.3230e-03) (hash(x)=18524290)
356
+ 2840 train 3.554548 (lr=1.3118e-03) (hash(x)=19600591)
357
+ 2850 train 3.751623 (lr=1.3006e-03) (hash(x)=21409231)
358
+ 2860 train 3.658887 (lr=1.2894e-03) (hash(x)=22752705)
359
+ 2870 train 3.863405 (lr=1.2783e-03) (hash(x)=19111240)
360
+ 2880 train 3.808617 (lr=1.2671e-03) (hash(x)=19723832)
361
+ 2890 train 3.619238 (lr=1.2560e-03) (hash(x)=20931007)
362
+ 2900 val loss 3.7101
363
+ 2900 val perplexity 40.8577
364
+ 2900 train 3.697971 (lr=1.2450e-03) (hash(x)=18157571)
365
+ 2910 train 3.914078 (lr=1.2339e-03) (hash(x)=15451342)
366
+ 2920 train 3.762803 (lr=1.2229e-03) (hash(x)=20840624)
367
+ 2930 train 3.809642 (lr=1.2119e-03) (hash(x)=20592314)
368
+ 2940 train 3.678662 (lr=1.2010e-03) (hash(x)=18543854)
369
+ 2950 train 3.676887 (lr=1.1901e-03) (hash(x)=18949144)
370
+ 2960 train 3.644515 (lr=1.1792e-03) (hash(x)=21711836)
371
+ 2970 train 3.625829 (lr=1.1684e-03) (hash(x)=22932108)
372
+ 2980 train 3.670794 (lr=1.1576e-03) (hash(x)=21153871)
373
+ 2990 train 3.594872 (lr=1.1468e-03) (hash(x)=20821229)
374
+ 3000 val loss 3.7081
375
+ 3000 val perplexity 40.7770
376
+ 3000 hella 0.2639
377
+ 3000 train 3.590554 (lr=1.1361e-03) (hash(x)=20308294)
378
+ 3010 train 3.477707 (lr=1.1254e-03) (hash(x)=15568285)
379
+ 3020 train 3.532680 (lr=1.1147e-03) (hash(x)=25333743)
380
+ 3030 train 3.787381 (lr=1.1041e-03) (hash(x)=21412609)
381
+ 3040 train 3.736506 (lr=1.0935e-03) (hash(x)=20306459)
382
+ 3050 train 3.757171 (lr=1.0830e-03) (hash(x)=22401325)
383
+ 3060 train 3.891738 (lr=1.0725e-03) (hash(x)=17461443)
384
+ 3070 train 3.682438 (lr=1.0620e-03) (hash(x)=21060753)
385
+ 3080 train 3.720884 (lr=1.0516e-03) (hash(x)=18276937)
386
+ 3090 train 3.767801 (lr=1.0412e-03) (hash(x)=22919536)
387
+ 3100 val loss 3.6816
388
+ 3100 val perplexity 39.7081
389
+ 3100 train 3.649293 (lr=1.0309e-03) (hash(x)=22068258)
390
+ 3110 train 3.679260 (lr=1.0207e-03) (hash(x)=22238174)
391
+ 3120 train 3.542986 (lr=1.0104e-03) (hash(x)=19841337)
392
+ 3130 train 3.551080 (lr=1.0002e-03) (hash(x)=18274815)
393
+ 3140 train 3.655715 (lr=9.9011e-04) (hash(x)=22162670)
394
+ 3150 train 3.694702 (lr=9.8002e-04) (hash(x)=22571495)
395
+ 3160 train 3.703209 (lr=9.6999e-04) (hash(x)=22642982)
396
+ 3170 train 3.715561 (lr=9.6000e-04) (hash(x)=20033229)
397
+ 3180 train 3.749094 (lr=9.5007e-04) (hash(x)=19437509)
398
+ 3190 train 3.756117 (lr=9.4019e-04) (hash(x)=16783200)
399
+ 3200 val loss 3.6606
400
+ 3200 val perplexity 38.8866
401
+ 3200 train 3.725704 (lr=9.3036e-04) (hash(x)=24078985)
402
+ 3210 train 3.671267 (lr=9.2058e-04) (hash(x)=17215036)
403
+ 3220 train 3.730839 (lr=9.1085e-04) (hash(x)=20834133)
404
+ 3230 train 3.777707 (lr=9.0118e-04) (hash(x)=22319183)
405
+ 3240 train 3.761367 (lr=8.9157e-04) (hash(x)=18488174)
406
+ 3250 hella 0.2623
407
+ 3250 train 3.669786 (lr=8.8201e-04) (hash(x)=18309688)
408
+ 3260 train 3.671246 (lr=8.7251e-04) (hash(x)=22420069)
409
+ 3270 train 3.662454 (lr=8.6307e-04) (hash(x)=20198122)
410
+ 3280 train 3.621376 (lr=8.5368e-04) (hash(x)=19296839)
411
+ 3290 train 3.743758 (lr=8.4435e-04) (hash(x)=18856940)
412
+ 3300 val loss 3.6507
413
+ 3300 val perplexity 38.5031
414
+ 3300 train 3.545310 (lr=8.3508e-04) (hash(x)=18860544)
415
+ 3310 train 3.575811 (lr=8.2588e-04) (hash(x)=18540569)
416
+ 3320 train 3.706502 (lr=8.1673e-04) (hash(x)=19524980)
417
+ 3330 train 3.423438 (lr=8.0764e-04) (hash(x)=20135353)
418
+ 3340 train 3.569072 (lr=7.9862e-04) (hash(x)=18231276)
419
+ 3350 train 3.694188 (lr=7.8966e-04) (hash(x)=21460655)
420
+ 3360 train 3.681391 (lr=7.8076e-04) (hash(x)=21427500)
421
+ 3370 train 3.639767 (lr=7.7192e-04) (hash(x)=20983682)
422
+ 3380 train 3.713173 (lr=7.6315e-04) (hash(x)=19255896)
423
+ 3390 train 3.859157 (lr=7.5445e-04) (hash(x)=22402624)
424
+ 3400 val loss 3.6282
425
+ 3400 val perplexity 37.6458
426
+ 3400 train 3.736673 (lr=7.4581e-04) (hash(x)=18912715)
427
+ 3410 train 3.736694 (lr=7.3724e-04) (hash(x)=21257398)
428
+ 3420 train 3.720812 (lr=7.2874e-04) (hash(x)=21948097)
429
+ 3430 train 3.825389 (lr=7.2030e-04) (hash(x)=20347030)
430
+ 3440 train 3.632965 (lr=7.1193e-04) (hash(x)=21481378)
431
+ 3450 train 3.640884 (lr=7.0363e-04) (hash(x)=17546482)
432
+ 3460 train 3.707560 (lr=6.9541e-04) (hash(x)=18550105)
433
+ 3470 train 3.672097 (lr=6.8725e-04) (hash(x)=24547129)
434
+ 3480 train 3.607415 (lr=6.7916e-04) (hash(x)=20098288)
435
+ 3490 train 3.484494 (lr=6.7114e-04) (hash(x)=21452276)
436
+ 3500 val loss 3.6218
437
+ 3500 val perplexity 37.4044
438
+ 3500 hella 0.2621
439
+ 3500 train 3.442814 (lr=6.6320e-04) (hash(x)=23225933)
440
+ 3510 train 3.586346 (lr=6.5533e-04) (hash(x)=21071905)
441
+ 3520 train 3.475165 (lr=6.4753e-04) (hash(x)=16906922)
442
+ 3530 train 3.734577 (lr=6.3981e-04) (hash(x)=21899151)
443
+ 3540 train 3.719867 (lr=6.3216e-04) (hash(x)=23041756)
444
+ 3550 train 3.740219 (lr=6.2458e-04) (hash(x)=21732674)
445
+ 3560 train 3.671240 (lr=6.1708e-04) (hash(x)=18473192)
446
+ 3570 train 3.700329 (lr=6.0966e-04) (hash(x)=14969150)
447
+ 3580 train 3.647456 (lr=6.0231e-04) (hash(x)=24237738)
448
+ 3590 train 3.573905 (lr=5.9504e-04) (hash(x)=17182881)
449
+ 3600 val loss 3.6018
450
+ 3600 val perplexity 36.6659
451
+ 3600 train 3.603187 (lr=5.8785e-04) (hash(x)=21408499)
452
+ 3610 train 3.615844 (lr=5.8074e-04) (hash(x)=20004592)
453
+ 3620 train 3.654868 (lr=5.7370e-04) (hash(x)=19807690)
454
+ 3630 train 3.617193 (lr=5.6675e-04) (hash(x)=15802211)
455
+ 3640 train 3.608238 (lr=5.5987e-04) (hash(x)=17076697)
456
+ 3650 train 3.673433 (lr=5.5308e-04) (hash(x)=18190785)
457
+ 3660 train 3.478796 (lr=5.4636e-04) (hash(x)=19759253)
458
+ 3670 train 3.549374 (lr=5.3973e-04) (hash(x)=17023532)
459
+ 3680 train 3.412514 (lr=5.3318e-04) (hash(x)=17433843)
460
+ 3690 train 3.415391 (lr=5.2671e-04) (hash(x)=18793933)
461
+ 3700 val loss 3.6011
462
+ 3700 val perplexity 36.6395
463
+ 3700 train 3.537344 (lr=5.2033e-04) (hash(x)=17153326)
464
+ 3710 train 3.724280 (lr=5.1402e-04) (hash(x)=18663421)
465
+ 3720 train 3.632090 (lr=5.0780e-04) (hash(x)=18726507)
466
+ 3730 train 3.626199 (lr=5.0167e-04) (hash(x)=21535668)
467
+ 3740 train 3.841143 (lr=4.9562e-04) (hash(x)=23817997)
468
+ 3750 hella 0.2626
469
+ 3750 train 3.625765 (lr=4.8965e-04) (hash(x)=16956634)
470
+ 3760 train 3.708429 (lr=4.8377e-04) (hash(x)=19896983)
471
+ 3770 train 3.685248 (lr=4.7798e-04) (hash(x)=20117113)
472
+ 3780 train 3.625684 (lr=4.7227e-04) (hash(x)=15844634)
473
+ 3790 train 3.724803 (lr=4.6665e-04) (hash(x)=20319750)
474
+ 3800 val loss 3.5812
475
+ 3800 val perplexity 35.9184
476
+ 3800 train 3.579834 (lr=4.6112e-04) (hash(x)=20435350)
477
+ 3810 train 3.599582 (lr=4.5567e-04) (hash(x)=20641047)
478
+ 3820 train 3.653370 (lr=4.5031e-04) (hash(x)=18297741)
479
+ 3830 train 3.582731 (lr=4.4504e-04) (hash(x)=17763890)
480
+ 3840 train 3.573376 (lr=4.3986e-04) (hash(x)=23487198)
481
+ 3850 train 3.556420 (lr=4.3477e-04) (hash(x)=21778602)
482
+ 3860 train 3.543508 (lr=4.2977e-04) (hash(x)=20416853)
483
+ 3870 train 3.478567 (lr=4.2486e-04) (hash(x)=20694619)
484
+ 3880 train 3.530393 (lr=4.2004e-04) (hash(x)=20581652)
485
+ 3890 train 3.610345 (lr=4.1530e-04) (hash(x)=24613563)
486
+ 3900 val loss 3.5716
487
+ 3900 val perplexity 35.5729
488
+ 3900 train 3.578161 (lr=4.1066e-04) (hash(x)=19662796)
489
+ 3910 train 3.542918 (lr=4.0611e-04) (hash(x)=23741407)
490
+ 3920 train 3.609020 (lr=4.0166e-04) (hash(x)=23031285)
491
+ 3930 train 3.635436 (lr=3.9729e-04) (hash(x)=23239084)
492
+ 3940 train 3.540740 (lr=3.9302e-04) (hash(x)=19296933)
493
+ 3950 train 3.546000 (lr=3.8884e-04) (hash(x)=20099740)
494
+ 3960 train 3.603016 (lr=3.8475e-04) (hash(x)=22043697)
495
+ 3970 train 3.616661 (lr=3.8076e-04) (hash(x)=27563931)
496
+ 3980 train 3.647436 (lr=3.7685e-04) (hash(x)=21251196)
497
+ 3990 train 3.462097 (lr=3.7305e-04) (hash(x)=22898068)
498
+ 4000 val loss 3.5644
499
+ 4000 val perplexity 35.3197
500
+ 4000 hella 0.2662
501
+ 4000 train 3.551707 (lr=3.6933e-04) (hash(x)=19449887)
502
+ 4010 train 3.560866 (lr=3.6572e-04) (hash(x)=21147640)
503
+ 4020 train 3.478668 (lr=3.6219e-04) (hash(x)=19826876)
504
+ 4030 train 3.580366 (lr=3.5876e-04) (hash(x)=23648072)
505
+ 4040 train 3.552593 (lr=3.5543e-04) (hash(x)=21568106)
506
+ 4050 train 3.546541 (lr=3.5219e-04) (hash(x)=28260637)
507
+ 4060 train 3.562685 (lr=3.4905e-04) (hash(x)=20127545)
508
+ 4070 train 3.536802 (lr=3.4600e-04) (hash(x)=20608325)
509
+ 4080 train 3.688708 (lr=3.4305e-04) (hash(x)=22853386)
510
+ 4090 train 3.555515 (lr=3.4019e-04) (hash(x)=21675157)
511
+ 4100 val loss 3.5503
512
+ 4100 val perplexity 34.8231
513
+ 4100 train 3.594016 (lr=3.3744e-04) (hash(x)=17377009)
514
+ 4110 train 3.571661 (lr=3.3477e-04) (hash(x)=22277706)
515
+ 4120 train 3.603137 (lr=3.3221e-04) (hash(x)=19781296)
516
+ 4130 train 3.605687 (lr=3.2974e-04) (hash(x)=14644070)
517
+ 4140 train 3.547884 (lr=3.2737e-04) (hash(x)=17324623)
518
+ 4150 train 3.553865 (lr=3.2510e-04) (hash(x)=18640674)
519
+ 4160 train 3.481996 (lr=3.2292e-04) (hash(x)=20754523)
520
+ 4170 train 3.435157 (lr=3.2085e-04) (hash(x)=22334132)
521
+ 4180 train 3.567030 (lr=3.1887e-04) (hash(x)=21104828)
522
+ 4190 train 3.586621 (lr=3.1699e-04) (hash(x)=19725849)
523
+ 4200 val loss 3.5459
524
+ 4200 val perplexity 34.6712
525
+ 4200 train 3.689755 (lr=3.1520e-04) (hash(x)=24693657)
526
+ 4210 train 3.539202 (lr=3.1352e-04) (hash(x)=20148669)
527
+ 4220 train 3.550663 (lr=3.1193e-04) (hash(x)=26860969)
528
+ 4230 train 3.578574 (lr=3.1044e-04) (hash(x)=20146192)
529
+ 4240 train 3.573604 (lr=3.0905e-04) (hash(x)=15710025)
530
+ 4250 hella 0.2675
531
+ 4250 train 3.607403 (lr=3.0776e-04) (hash(x)=16513041)
532
+ 4260 train 3.606173 (lr=3.0657e-04) (hash(x)=17073880)
533
+ 4270 train 3.492221 (lr=3.0548e-04) (hash(x)=21596384)
534
+ 4280 train 3.546285 (lr=3.0449e-04) (hash(x)=18548369)
535
+ 4290 train 3.439570 (lr=3.0359e-04) (hash(x)=20524519)
536
+ 4300 val loss 3.5443
537
+ 4300 val perplexity 34.6154
538
+ 4300 train 3.538421 (lr=3.0280e-04) (hash(x)=20906140)
539
+ 4310 train 3.480823 (lr=3.0210e-04) (hash(x)=21519922)
540
+ 4320 train 3.539315 (lr=3.0150e-04) (hash(x)=20493589)
541
+ 4330 train 3.634491 (lr=3.0101e-04) (hash(x)=18964922)
542
+ 4340 train 3.513443 (lr=3.0061e-04) (hash(x)=20832448)
543
+ 4350 train 3.526926 (lr=3.0031e-04) (hash(x)=18631764)
544
+ 4360 train 3.536441 (lr=3.0011e-04) (hash(x)=19542477)
545
+ 4370 train 3.613957 (lr=3.0001e-04) (hash(x)=17072278)
546
+ 4374 val loss 3.5375
547
+ 4374 val perplexity 34.3821
548
+ 4374 hella 0.2667
logs/64_head_dim_two_latent_masks/2_latent_mask_seed_1339/model_04374.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:4854ccd632fef6526a560f219c8228c9859a097ef1816eb1705a5c1302b7f7e8
3
+ size 505704434
logs/64_head_dim_two_latent_masks/2_latent_mask_seed_1339/optimizer_04374.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:d6e943b52ce72c6f51f852235fd2148692ff28fc5bfa20a6b53e1b625551a4ae
3
+ size 1005130806