andrew-healey commited on
Commit
c357701
·
verified ·
1 Parent(s): 288339a

Upload folder using huggingface_hub

Browse files
logs/64_heads_bump_comparison/13_heads_no_compile_seed_1340/args.json ADDED
@@ -0,0 +1 @@
 
 
1
+ {"hellaswag": true, "attention_kind": "selective", "log_dir": "logs/64_heads_bump_comparison/13_heads_no_compile_seed_1340", "resume_checkpoint": null, "resume_optimizer": false, "add_a_head": false, "add_head_to_start": true, "new_head_init": "normal", "n_heads": 13, "n_embd": 768, "head_dim": 64, "protect_bos_token": true, "prevent_from_masking_myself": true, "max_steps": 4375, "warmup_steps": 250, "group": "64_heads_bump_comparison", "use_wandb": true, "kill_self_after_run": false, "random_seed": 1340, "memory_penalty_epsilon": 0.1, "selection_head_linear_combo": "none", "selection_head_linear_combo_scale": 1.0, "disable_selection_head_linear_combo_bias": false, "assert_latent_matches_no_head": false, "n_latent_masks": null, "init_latent_masks_to_identity": false, "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": 1024, "batch_size": 8, "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": "13_heads_no_compile", "latent_mask_precision": "bfloat16"}
logs/64_heads_bump_comparison/13_heads_no_compile_seed_1340/dataloader_04374.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:6031fd3e2855a036f7a5531cc24555aabd1115f9dd6618b8b2ca6f55279ef0b2
3
+ size 964
logs/64_heads_bump_comparison/13_heads_no_compile_seed_1340/log2.txt ADDED
@@ -0,0 +1,548 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ max_steps: 4375
2
+ 0 val loss 10.9161
3
+ 0 val perplexity 55056.2148
4
+ 0 hella 0.2574
5
+ 0 train 10.924518 (lr=4.1958e-06) (hash(x)=45482580)
6
+ 10 train 9.599250 (lr=4.6154e-05) (hash(x)=38414730)
7
+ 20 train 9.113592 (lr=8.8112e-05) (hash(x)=39783917)
8
+ 30 train 8.409630 (lr=1.3007e-04) (hash(x)=38395733)
9
+ 40 train 7.788647 (lr=1.7203e-04) (hash(x)=37486707)
10
+ 50 train 7.580905 (lr=2.1399e-04) (hash(x)=39897505)
11
+ 60 train 7.485894 (lr=2.5594e-04) (hash(x)=40933473)
12
+ 70 train 7.444473 (lr=2.9790e-04) (hash(x)=37690746)
13
+ 80 train 7.294475 (lr=3.3986e-04) (hash(x)=38609656)
14
+ 90 train 7.066379 (lr=3.8182e-04) (hash(x)=38356571)
15
+ 100 val loss 6.9566
16
+ 100 val perplexity 1050.0975
17
+ 100 train 6.938846 (lr=4.2378e-04) (hash(x)=35980376)
18
+ 110 train 6.829484 (lr=4.6573e-04) (hash(x)=42011042)
19
+ 120 train 6.743187 (lr=5.0769e-04) (hash(x)=40266823)
20
+ 130 train 6.594794 (lr=5.4965e-04) (hash(x)=38645447)
21
+ 140 train 6.449816 (lr=5.9161e-04) (hash(x)=36059313)
22
+ 150 train 6.413119 (lr=6.3357e-04) (hash(x)=34161947)
23
+ 160 train 6.285947 (lr=6.7552e-04) (hash(x)=42263375)
24
+ 170 train 6.343897 (lr=7.1748e-04) (hash(x)=42957725)
25
+ 180 train 6.328916 (lr=7.5944e-04) (hash(x)=40198018)
26
+ 190 train 6.082541 (lr=8.0140e-04) (hash(x)=41666215)
27
+ 200 val loss 6.0501
28
+ 200 val perplexity 424.1606
29
+ 200 train 6.013837 (lr=8.4336e-04) (hash(x)=54060482)
30
+ 210 train 5.929295 (lr=8.8531e-04) (hash(x)=42332778)
31
+ 220 train 5.900323 (lr=9.2727e-04) (hash(x)=39551486)
32
+ 230 train 5.866964 (lr=9.6923e-04) (hash(x)=38462018)
33
+ 240 train 5.927802 (lr=1.0112e-03) (hash(x)=36591442)
34
+ 250 hella 0.2389
35
+ 250 train 5.893656 (lr=1.0531e-03) (hash(x)=40861237)
36
+ 260 train 5.808354 (lr=1.0951e-03) (hash(x)=41739369)
37
+ 270 train 5.662393 (lr=1.1371e-03) (hash(x)=39415360)
38
+ 280 train 5.631592 (lr=1.1790e-03) (hash(x)=44665934)
39
+ 290 train 5.597699 (lr=1.2210e-03) (hash(x)=37035579)
40
+ 300 val loss 5.5746
41
+ 300 val perplexity 263.6502
42
+ 300 train 5.593894 (lr=1.2629e-03) (hash(x)=38301011)
43
+ 310 train 5.432870 (lr=1.3049e-03) (hash(x)=42850980)
44
+ 320 train 5.492656 (lr=1.3469e-03) (hash(x)=37476222)
45
+ 330 train 5.418432 (lr=1.3888e-03) (hash(x)=53028205)
46
+ 340 train 5.450152 (lr=1.4308e-03) (hash(x)=41466008)
47
+ 350 train 5.331077 (lr=1.4727e-03) (hash(x)=37802865)
48
+ 360 train 5.274551 (lr=1.5147e-03) (hash(x)=42769282)
49
+ 370 train 5.306739 (lr=1.5566e-03) (hash(x)=39319256)
50
+ 380 train 5.273731 (lr=1.5986e-03) (hash(x)=42637402)
51
+ 390 train 5.302903 (lr=1.6406e-03) (hash(x)=31141514)
52
+ 400 val loss 5.3063
53
+ 400 val perplexity 201.5979
54
+ 400 train 5.372399 (lr=1.6825e-03) (hash(x)=38151157)
55
+ 410 train 5.289481 (lr=1.7245e-03) (hash(x)=46033439)
56
+ 420 train 5.242544 (lr=1.7664e-03) (hash(x)=41365246)
57
+ 430 train 5.234932 (lr=1.8084e-03) (hash(x)=42369184)
58
+ 440 train 5.210252 (lr=1.8503e-03) (hash(x)=42004840)
59
+ 450 train 5.112603 (lr=1.8923e-03) (hash(x)=37181172)
60
+ 460 train 5.219237 (lr=1.9343e-03) (hash(x)=31630797)
61
+ 470 train 5.087325 (lr=1.9762e-03) (hash(x)=42135747)
62
+ 480 train 5.010891 (lr=2.0182e-03) (hash(x)=30023651)
63
+ 490 train 5.082321 (lr=2.0601e-03) (hash(x)=40463476)
64
+ 500 val loss 5.0222
65
+ 500 val perplexity 151.7520
66
+ 500 hella 0.2449
67
+ 500 train 4.946929 (lr=2.1021e-03) (hash(x)=37089842)
68
+ 510 train 4.944039 (lr=2.1441e-03) (hash(x)=41095974)
69
+ 520 train 4.993733 (lr=2.1860e-03) (hash(x)=41525258)
70
+ 530 train 4.934231 (lr=2.2280e-03) (hash(x)=37697291)
71
+ 540 train 4.965906 (lr=2.2699e-03) (hash(x)=43459894)
72
+ 550 train 4.825506 (lr=2.3119e-03) (hash(x)=38993956)
73
+ 560 train 4.828552 (lr=2.3538e-03) (hash(x)=40677277)
74
+ 570 train 4.955185 (lr=2.3958e-03) (hash(x)=43103127)
75
+ 580 train 4.811969 (lr=2.4378e-03) (hash(x)=41871776)
76
+ 590 train 4.819898 (lr=2.4797e-03) (hash(x)=43230967)
77
+ 600 val loss 4.7296
78
+ 600 val perplexity 113.2460
79
+ 600 train 4.759600 (lr=2.5217e-03) (hash(x)=42226172)
80
+ 610 train 5.009989 (lr=2.5636e-03) (hash(x)=43007164)
81
+ 620 train 4.624640 (lr=2.6056e-03) (hash(x)=35752341)
82
+ 630 train 4.708325 (lr=2.6476e-03) (hash(x)=39453504)
83
+ 640 train 4.640758 (lr=2.6895e-03) (hash(x)=43635712)
84
+ 650 train 4.655717 (lr=2.7315e-03) (hash(x)=41852456)
85
+ 660 train 4.654819 (lr=2.7734e-03) (hash(x)=49125742)
86
+ 670 train 4.670001 (lr=2.8154e-03) (hash(x)=34886275)
87
+ 680 train 4.767265 (lr=2.8573e-03) (hash(x)=40546482)
88
+ 690 train 4.566515 (lr=2.8993e-03) (hash(x)=40850901)
89
+ 700 val loss 4.5407
90
+ 700 val perplexity 93.7571
91
+ 700 train 4.401430 (lr=2.9413e-03) (hash(x)=37426993)
92
+ 710 train 4.708692 (lr=2.9832e-03) (hash(x)=42679743)
93
+ 720 train 4.508728 (lr=3.0000e-03) (hash(x)=44856043)
94
+ 730 train 4.508308 (lr=2.9999e-03) (hash(x)=36457471)
95
+ 740 train 4.462440 (lr=2.9997e-03) (hash(x)=47203534)
96
+ 750 hella 0.2490
97
+ 750 train 4.480333 (lr=2.9994e-03) (hash(x)=35585657)
98
+ 760 train 4.377686 (lr=2.9990e-03) (hash(x)=37259264)
99
+ 770 train 4.296279 (lr=2.9985e-03) (hash(x)=44161997)
100
+ 780 train 4.262340 (lr=2.9979e-03) (hash(x)=41272886)
101
+ 790 train 4.270710 (lr=2.9972e-03) (hash(x)=43303662)
102
+ 800 val loss 4.4378
103
+ 800 val perplexity 84.5928
104
+ 800 train 4.126956 (lr=2.9964e-03) (hash(x)=39067231)
105
+ 810 train 4.391942 (lr=2.9955e-03) (hash(x)=44046732)
106
+ 820 train 4.371235 (lr=2.9945e-03) (hash(x)=39783962)
107
+ 830 train 4.405455 (lr=2.9934e-03) (hash(x)=40973974)
108
+ 840 train 4.405073 (lr=2.9922e-03) (hash(x)=41619968)
109
+ 850 train 4.440304 (lr=2.9909e-03) (hash(x)=36939960)
110
+ 860 train 4.315139 (lr=2.9896e-03) (hash(x)=42942377)
111
+ 870 train 4.334913 (lr=2.9881e-03) (hash(x)=39659455)
112
+ 880 train 4.348632 (lr=2.9865e-03) (hash(x)=40331986)
113
+ 890 train 4.263987 (lr=2.9848e-03) (hash(x)=38084814)
114
+ 900 val loss 4.2924
115
+ 900 val perplexity 73.1429
116
+ 900 train 4.287558 (lr=2.9830e-03) (hash(x)=30830367)
117
+ 910 train 4.201886 (lr=2.9811e-03) (hash(x)=39664356)
118
+ 920 train 4.203846 (lr=2.9792e-03) (hash(x)=39007775)
119
+ 930 train 4.220542 (lr=2.9771e-03) (hash(x)=39319254)
120
+ 940 train 4.345969 (lr=2.9749e-03) (hash(x)=36347051)
121
+ 950 train 4.241587 (lr=2.9726e-03) (hash(x)=39243577)
122
+ 960 train 4.223720 (lr=2.9703e-03) (hash(x)=45603839)
123
+ 970 train 4.042967 (lr=2.9678e-03) (hash(x)=40036075)
124
+ 980 train 4.098536 (lr=2.9652e-03) (hash(x)=43068524)
125
+ 990 train 4.088611 (lr=2.9626e-03) (hash(x)=33798472)
126
+ 1000 val loss 4.2410
127
+ 1000 val perplexity 69.4794
128
+ 1000 hella 0.2562
129
+ 1000 train 4.454307 (lr=2.9598e-03) (hash(x)=37728665)
130
+ 1010 train 4.304788 (lr=2.9570e-03) (hash(x)=43678636)
131
+ 1020 train 4.301694 (lr=2.9540e-03) (hash(x)=48464376)
132
+ 1030 train 4.306663 (lr=2.9510e-03) (hash(x)=43771461)
133
+ 1040 train 4.158973 (lr=2.9478e-03) (hash(x)=44643085)
134
+ 1050 train 4.239938 (lr=2.9446e-03) (hash(x)=39201716)
135
+ 1060 train 4.178282 (lr=2.9412e-03) (hash(x)=36396343)
136
+ 1070 train 4.169182 (lr=2.9378e-03) (hash(x)=37504111)
137
+ 1080 train 4.156756 (lr=2.9343e-03) (hash(x)=37454537)
138
+ 1090 train 4.168472 (lr=2.9307e-03) (hash(x)=36888183)
139
+ 1100 val loss 4.1535
140
+ 1100 val perplexity 63.6566
141
+ 1100 train 4.113934 (lr=2.9270e-03) (hash(x)=39549045)
142
+ 1110 train 4.300883 (lr=2.9231e-03) (hash(x)=56818420)
143
+ 1120 train 4.121250 (lr=2.9192e-03) (hash(x)=36573357)
144
+ 1130 train 4.037148 (lr=2.9152e-03) (hash(x)=38540878)
145
+ 1140 train 4.053041 (lr=2.9112e-03) (hash(x)=38645859)
146
+ 1150 train 4.238201 (lr=2.9070e-03) (hash(x)=41972503)
147
+ 1160 train 3.846027 (lr=2.9027e-03) (hash(x)=38151868)
148
+ 1170 train 3.887326 (lr=2.8983e-03) (hash(x)=42375436)
149
+ 1180 train 3.867134 (lr=2.8939e-03) (hash(x)=42868028)
150
+ 1190 train 4.118700 (lr=2.8893e-03) (hash(x)=41060868)
151
+ 1200 val loss 4.1032
152
+ 1200 val perplexity 60.5309
153
+ 1200 train 4.121489 (lr=2.8847e-03) (hash(x)=40852814)
154
+ 1210 train 4.091335 (lr=2.8800e-03) (hash(x)=43365554)
155
+ 1220 train 4.034972 (lr=2.8751e-03) (hash(x)=39121134)
156
+ 1230 train 4.134510 (lr=2.8702e-03) (hash(x)=40565426)
157
+ 1240 train 4.208247 (lr=2.8652e-03) (hash(x)=43378926)
158
+ 1250 hella 0.2560
159
+ 1250 train 4.023478 (lr=2.8601e-03) (hash(x)=35227381)
160
+ 1260 train 4.129640 (lr=2.8550e-03) (hash(x)=41267226)
161
+ 1270 train 4.064462 (lr=2.8497e-03) (hash(x)=31395565)
162
+ 1280 train 4.122903 (lr=2.8443e-03) (hash(x)=41470963)
163
+ 1290 train 3.941321 (lr=2.8389e-03) (hash(x)=40729366)
164
+ 1300 val loss 4.0370
165
+ 1300 val perplexity 56.6553
166
+ 1300 train 4.026610 (lr=2.8333e-03) (hash(x)=39916663)
167
+ 1310 train 4.040397 (lr=2.8277e-03) (hash(x)=42179962)
168
+ 1320 train 3.924713 (lr=2.8220e-03) (hash(x)=44991932)
169
+ 1330 train 3.880450 (lr=2.8162e-03) (hash(x)=38925911)
170
+ 1340 train 3.828296 (lr=2.8103e-03) (hash(x)=38565560)
171
+ 1350 train 3.879773 (lr=2.8044e-03) (hash(x)=43384902)
172
+ 1360 train 3.834180 (lr=2.7983e-03) (hash(x)=45422261)
173
+ 1370 train 3.817152 (lr=2.7922e-03) (hash(x)=40167325)
174
+ 1380 train 3.958490 (lr=2.7860e-03) (hash(x)=51258162)
175
+ 1390 train 4.027603 (lr=2.7797e-03) (hash(x)=43229106)
176
+ 1400 val loss 3.9889
177
+ 1400 val perplexity 53.9946
178
+ 1400 train 3.953212 (lr=2.7733e-03) (hash(x)=42537494)
179
+ 1410 train 4.229206 (lr=2.7668e-03) (hash(x)=51236490)
180
+ 1420 train 4.061996 (lr=2.7603e-03) (hash(x)=45243347)
181
+ 1430 train 3.921724 (lr=2.7536e-03) (hash(x)=37658566)
182
+ 1440 train 3.988309 (lr=2.7469e-03) (hash(x)=38856106)
183
+ 1450 train 4.055024 (lr=2.7401e-03) (hash(x)=36649488)
184
+ 1460 train 4.044413 (lr=2.7333e-03) (hash(x)=40445407)
185
+ 1470 train 3.988968 (lr=2.7263e-03) (hash(x)=39916387)
186
+ 1480 train 3.959047 (lr=2.7193e-03) (hash(x)=40934555)
187
+ 1490 train 3.919458 (lr=2.7121e-03) (hash(x)=42276527)
188
+ 1500 val loss 3.9593
189
+ 1500 val perplexity 52.4200
190
+ 1500 hella 0.2573
191
+ 1500 train 3.909872 (lr=2.7050e-03) (hash(x)=37973883)
192
+ 1510 train 3.935476 (lr=2.6977e-03) (hash(x)=38929286)
193
+ 1520 train 3.778986 (lr=2.6903e-03) (hash(x)=36855837)
194
+ 1530 train 4.036331 (lr=2.6829e-03) (hash(x)=42146431)
195
+ 1540 train 3.974432 (lr=2.6754e-03) (hash(x)=41898411)
196
+ 1550 train 3.882111 (lr=2.6678e-03) (hash(x)=44906272)
197
+ 1560 train 3.908237 (lr=2.6602e-03) (hash(x)=40414353)
198
+ 1570 train 3.956314 (lr=2.6525e-03) (hash(x)=38723190)
199
+ 1580 train 3.840225 (lr=2.6447e-03) (hash(x)=39065271)
200
+ 1590 train 3.872326 (lr=2.6368e-03) (hash(x)=40000886)
201
+ 1600 val loss 3.9208
202
+ 1600 val perplexity 50.4418
203
+ 1600 train 3.775952 (lr=2.6289e-03) (hash(x)=37898571)
204
+ 1610 train 3.882262 (lr=2.6208e-03) (hash(x)=40447863)
205
+ 1620 train 3.834979 (lr=2.6128e-03) (hash(x)=40119318)
206
+ 1630 train 3.785786 (lr=2.6046e-03) (hash(x)=40664074)
207
+ 1640 train 3.745187 (lr=2.5964e-03) (hash(x)=37404741)
208
+ 1650 train 3.895607 (lr=2.5881e-03) (hash(x)=39214928)
209
+ 1660 train 3.927749 (lr=2.5797e-03) (hash(x)=35005019)
210
+ 1670 train 3.952534 (lr=2.5713e-03) (hash(x)=37675832)
211
+ 1680 train 3.920050 (lr=2.5628e-03) (hash(x)=40489680)
212
+ 1690 train 3.842745 (lr=2.5542e-03) (hash(x)=45537879)
213
+ 1700 val loss 3.8889
214
+ 1700 val perplexity 48.8565
215
+ 1700 train 3.828368 (lr=2.5455e-03) (hash(x)=38866100)
216
+ 1710 train 3.875422 (lr=2.5368e-03) (hash(x)=44726254)
217
+ 1720 train 3.864232 (lr=2.5281e-03) (hash(x)=36917792)
218
+ 1730 train 3.837669 (lr=2.5192e-03) (hash(x)=42814805)
219
+ 1740 train 3.917092 (lr=2.5103e-03) (hash(x)=51812216)
220
+ 1750 hella 0.2568
221
+ 1750 train 3.791716 (lr=2.5014e-03) (hash(x)=33549014)
222
+ 1760 train 3.794625 (lr=2.4924e-03) (hash(x)=40354215)
223
+ 1770 train 3.822738 (lr=2.4833e-03) (hash(x)=40718606)
224
+ 1780 train 3.750740 (lr=2.4741e-03) (hash(x)=38274164)
225
+ 1790 train 3.884122 (lr=2.4649e-03) (hash(x)=42531471)
226
+ 1800 val loss 3.8615
227
+ 1800 val perplexity 47.5379
228
+ 1800 train 3.844220 (lr=2.4556e-03) (hash(x)=35616519)
229
+ 1810 train 3.864695 (lr=2.4463e-03) (hash(x)=33803118)
230
+ 1820 train 3.885228 (lr=2.4369e-03) (hash(x)=36973525)
231
+ 1830 train 3.852531 (lr=2.4275e-03) (hash(x)=38162549)
232
+ 1840 train 3.833591 (lr=2.4180e-03) (hash(x)=33937159)
233
+ 1850 train 3.878129 (lr=2.4084e-03) (hash(x)=39887546)
234
+ 1860 train 3.752393 (lr=2.3988e-03) (hash(x)=37818525)
235
+ 1870 train 3.812887 (lr=2.3891e-03) (hash(x)=37250478)
236
+ 1880 train 3.704835 (lr=2.3794e-03) (hash(x)=36412167)
237
+ 1890 train 3.806083 (lr=2.3696e-03) (hash(x)=40261189)
238
+ 1900 val loss 3.8499
239
+ 1900 val perplexity 46.9875
240
+ 1900 train 3.783599 (lr=2.3598e-03) (hash(x)=38654303)
241
+ 1910 train 3.727170 (lr=2.3499e-03) (hash(x)=37929515)
242
+ 1920 train 3.886823 (lr=2.3400e-03) (hash(x)=45580146)
243
+ 1930 train 3.830049 (lr=2.3300e-03) (hash(x)=32386330)
244
+ 1940 train 3.871968 (lr=2.3200e-03) (hash(x)=36331864)
245
+ 1950 train 3.857353 (lr=2.3099e-03) (hash(x)=41181727)
246
+ 1960 train 3.656847 (lr=2.2998e-03) (hash(x)=42705152)
247
+ 1970 train 3.785862 (lr=2.2896e-03) (hash(x)=40251511)
248
+ 1980 train 3.720516 (lr=2.2793e-03) (hash(x)=37525551)
249
+ 1990 train 3.764929 (lr=2.2691e-03) (hash(x)=41849618)
250
+ 2000 val loss 3.8112
251
+ 2000 val perplexity 45.2042
252
+ 2000 hella 0.2619
253
+ 2000 train 3.814750 (lr=2.2588e-03) (hash(x)=37642582)
254
+ 2010 train 3.788179 (lr=2.2484e-03) (hash(x)=42131121)
255
+ 2020 train 3.681581 (lr=2.2380e-03) (hash(x)=39000209)
256
+ 2030 train 3.774901 (lr=2.2275e-03) (hash(x)=43641355)
257
+ 2040 train 3.655588 (lr=2.2170e-03) (hash(x)=47910507)
258
+ 2050 train 3.677436 (lr=2.2065e-03) (hash(x)=36670359)
259
+ 2060 train 3.846431 (lr=2.1959e-03) (hash(x)=36477755)
260
+ 2070 train 3.913362 (lr=2.1853e-03) (hash(x)=43262487)
261
+ 2080 train 3.714719 (lr=2.1746e-03) (hash(x)=49546029)
262
+ 2090 train 3.877703 (lr=2.1639e-03) (hash(x)=39924731)
263
+ 2100 val loss 3.7903
264
+ 2100 val perplexity 44.2697
265
+ 2100 train 3.738570 (lr=2.1532e-03) (hash(x)=39921304)
266
+ 2110 train 3.702976 (lr=2.1424e-03) (hash(x)=38152788)
267
+ 2120 train 3.761880 (lr=2.1316e-03) (hash(x)=37977911)
268
+ 2130 train 3.735536 (lr=2.1208e-03) (hash(x)=34748760)
269
+ 2140 train 3.766486 (lr=2.1099e-03) (hash(x)=40509369)
270
+ 2150 train 3.738815 (lr=2.0990e-03) (hash(x)=37654262)
271
+ 2160 train 3.700902 (lr=2.0881e-03) (hash(x)=38139543)
272
+ 2170 train 3.602980 (lr=2.0771e-03) (hash(x)=42501806)
273
+ 2180 train 3.742223 (lr=2.0661e-03) (hash(x)=40085092)
274
+ 2190 train 3.709488 (lr=2.0550e-03) (hash(x)=50480193)
275
+ 2200 val loss 3.7701
276
+ 2200 val perplexity 43.3862
277
+ 2200 train 3.777882 (lr=2.0440e-03) (hash(x)=40604084)
278
+ 2210 train 3.865965 (lr=2.0329e-03) (hash(x)=41555823)
279
+ 2220 train 3.702904 (lr=2.0217e-03) (hash(x)=50441765)
280
+ 2230 train 3.817141 (lr=2.0106e-03) (hash(x)=39796580)
281
+ 2240 train 3.814400 (lr=1.9994e-03) (hash(x)=44127022)
282
+ 2250 hella 0.2579
283
+ 2250 train 3.724712 (lr=1.9882e-03) (hash(x)=37026826)
284
+ 2260 train 3.763681 (lr=1.9770e-03) (hash(x)=42133839)
285
+ 2270 train 3.742120 (lr=1.9657e-03) (hash(x)=38500664)
286
+ 2280 train 3.710731 (lr=1.9544e-03) (hash(x)=40538661)
287
+ 2290 train 3.708076 (lr=1.9431e-03) (hash(x)=51509210)
288
+ 2300 val loss 3.7415
289
+ 2300 val perplexity 42.1592
290
+ 2300 train 3.725485 (lr=1.9318e-03) (hash(x)=41952328)
291
+ 2310 train 3.665669 (lr=1.9205e-03) (hash(x)=39758123)
292
+ 2320 train 3.732141 (lr=1.9091e-03) (hash(x)=51089268)
293
+ 2330 train 3.818531 (lr=1.8977e-03) (hash(x)=39767618)
294
+ 2340 train 3.859157 (lr=1.8863e-03) (hash(x)=40409617)
295
+ 2350 train 3.700011 (lr=1.8749e-03) (hash(x)=40349634)
296
+ 2360 train 3.982908 (lr=1.8635e-03) (hash(x)=31841172)
297
+ 2370 train 3.759320 (lr=1.8520e-03) (hash(x)=42720539)
298
+ 2380 train 3.617237 (lr=1.8406e-03) (hash(x)=40998632)
299
+ 2390 train 3.720074 (lr=1.8291e-03) (hash(x)=40615413)
300
+ 2400 val loss 3.7214
301
+ 2400 val perplexity 41.3218
302
+ 2400 train 3.658755 (lr=1.8176e-03) (hash(x)=39373658)
303
+ 2410 train 3.696934 (lr=1.8061e-03) (hash(x)=35480858)
304
+ 2420 train 3.688976 (lr=1.7946e-03) (hash(x)=32241095)
305
+ 2430 train 3.669248 (lr=1.7830e-03) (hash(x)=36669715)
306
+ 2440 train 3.682950 (lr=1.7715e-03) (hash(x)=45768335)
307
+ 2450 train 3.581079 (lr=1.7600e-03) (hash(x)=39395055)
308
+ 2460 train 3.623884 (lr=1.7484e-03) (hash(x)=34899269)
309
+ 2470 train 3.766529 (lr=1.7368e-03) (hash(x)=43642420)
310
+ 2480 train 3.834340 (lr=1.7253e-03) (hash(x)=47099507)
311
+ 2490 train 3.784206 (lr=1.7137e-03) (hash(x)=38825558)
312
+ 2500 val loss 3.7033
313
+ 2500 val perplexity 40.5805
314
+ 2500 hella 0.2639
315
+ 2500 train 3.752115 (lr=1.7021e-03) (hash(x)=39833804)
316
+ 2510 train 3.824955 (lr=1.6906e-03) (hash(x)=35051654)
317
+ 2520 train 3.746531 (lr=1.6790e-03) (hash(x)=40291109)
318
+ 2530 train 3.659722 (lr=1.6674e-03) (hash(x)=36915768)
319
+ 2540 train 3.640140 (lr=1.6558e-03) (hash(x)=37016308)
320
+ 2550 train 3.598763 (lr=1.6442e-03) (hash(x)=43221777)
321
+ 2560 train 3.662602 (lr=1.6326e-03) (hash(x)=37233207)
322
+ 2570 train 3.530366 (lr=1.6210e-03) (hash(x)=37510626)
323
+ 2580 train 3.593141 (lr=1.6094e-03) (hash(x)=39624656)
324
+ 2590 train 3.443625 (lr=1.5979e-03) (hash(x)=37647501)
325
+ 2600 val loss 3.6826
326
+ 2600 val perplexity 39.7511
327
+ 2600 train 3.589912 (lr=1.5863e-03) (hash(x)=37200138)
328
+ 2610 train 3.607315 (lr=1.5747e-03) (hash(x)=43066970)
329
+ 2620 train 3.695782 (lr=1.5632e-03) (hash(x)=40503799)
330
+ 2630 train 3.644807 (lr=1.5516e-03) (hash(x)=45534660)
331
+ 2640 train 3.811368 (lr=1.5400e-03) (hash(x)=52358353)
332
+ 2650 train 3.653673 (lr=1.5285e-03) (hash(x)=41843599)
333
+ 2660 train 3.611670 (lr=1.5170e-03) (hash(x)=40519770)
334
+ 2670 train 3.692467 (lr=1.5054e-03) (hash(x)=37185602)
335
+ 2680 train 3.570742 (lr=1.4939e-03) (hash(x)=43805719)
336
+ 2690 train 3.606414 (lr=1.4824e-03) (hash(x)=39458250)
337
+ 2700 val loss 3.6569
338
+ 2700 val perplexity 38.7426
339
+ 2700 train 3.702733 (lr=1.4709e-03) (hash(x)=39149255)
340
+ 2710 train 3.543034 (lr=1.4594e-03) (hash(x)=43544861)
341
+ 2720 train 3.630728 (lr=1.4480e-03) (hash(x)=36858197)
342
+ 2730 train 3.550951 (lr=1.4365e-03) (hash(x)=37971712)
343
+ 2740 train 3.582062 (lr=1.4251e-03) (hash(x)=42416578)
344
+ 2750 hella 0.2619
345
+ 2750 train 3.517304 (lr=1.4137e-03) (hash(x)=53554622)
346
+ 2760 train 3.616549 (lr=1.4023e-03) (hash(x)=40502545)
347
+ 2770 train 3.601926 (lr=1.3909e-03) (hash(x)=36086417)
348
+ 2780 train 3.639012 (lr=1.3795e-03) (hash(x)=36242982)
349
+ 2790 train 3.664913 (lr=1.3682e-03) (hash(x)=41433780)
350
+ 2800 val loss 3.6351
351
+ 2800 val perplexity 37.9071
352
+ 2800 train 3.601521 (lr=1.3569e-03) (hash(x)=48553484)
353
+ 2810 train 3.543077 (lr=1.3456e-03) (hash(x)=39928650)
354
+ 2820 train 3.651902 (lr=1.3343e-03) (hash(x)=37891724)
355
+ 2830 train 3.612575 (lr=1.3230e-03) (hash(x)=34353412)
356
+ 2840 train 3.635442 (lr=1.3118e-03) (hash(x)=38486611)
357
+ 2850 train 3.574555 (lr=1.3006e-03) (hash(x)=32706934)
358
+ 2860 train 3.512821 (lr=1.2894e-03) (hash(x)=38693681)
359
+ 2870 train 3.585625 (lr=1.2783e-03) (hash(x)=38430800)
360
+ 2880 train 3.549526 (lr=1.2671e-03) (hash(x)=40619559)
361
+ 2890 train 3.504555 (lr=1.2560e-03) (hash(x)=42741066)
362
+ 2900 val loss 3.6305
363
+ 2900 val perplexity 37.7334
364
+ 2900 train 3.768107 (lr=1.2450e-03) (hash(x)=38216091)
365
+ 2910 train 3.564208 (lr=1.2339e-03) (hash(x)=35447832)
366
+ 2920 train 3.684887 (lr=1.2229e-03) (hash(x)=39825190)
367
+ 2930 train 3.677660 (lr=1.2119e-03) (hash(x)=37789121)
368
+ 2940 train 3.580509 (lr=1.2010e-03) (hash(x)=44499116)
369
+ 2950 train 3.655101 (lr=1.1901e-03) (hash(x)=39669860)
370
+ 2960 train 3.588198 (lr=1.1792e-03) (hash(x)=51683741)
371
+ 2970 train 3.487049 (lr=1.1684e-03) (hash(x)=36399721)
372
+ 2980 train 3.570423 (lr=1.1576e-03) (hash(x)=42629700)
373
+ 2990 train 3.451481 (lr=1.1468e-03) (hash(x)=39263773)
374
+ 3000 val loss 3.6024
375
+ 3000 val perplexity 36.6861
376
+ 3000 hella 0.2612
377
+ 3000 train 3.505232 (lr=1.1361e-03) (hash(x)=37920485)
378
+ 3010 train 3.545671 (lr=1.1254e-03) (hash(x)=40835161)
379
+ 3020 train 3.477572 (lr=1.1147e-03) (hash(x)=38381159)
380
+ 3030 train 3.495310 (lr=1.1041e-03) (hash(x)=35846270)
381
+ 3040 train 3.856606 (lr=1.0935e-03) (hash(x)=47516567)
382
+ 3050 train 3.689031 (lr=1.0830e-03) (hash(x)=35600311)
383
+ 3060 train 3.727193 (lr=1.0725e-03) (hash(x)=37775318)
384
+ 3070 train 3.601333 (lr=1.0620e-03) (hash(x)=39881333)
385
+ 3080 train 3.603901 (lr=1.0516e-03) (hash(x)=37310168)
386
+ 3090 train 3.530572 (lr=1.0412e-03) (hash(x)=43669978)
387
+ 3100 val loss 3.5857
388
+ 3100 val perplexity 36.0775
389
+ 3100 train 3.956401 (lr=1.0309e-03) (hash(x)=46356797)
390
+ 3110 train 3.656170 (lr=1.0207e-03) (hash(x)=40814232)
391
+ 3120 train 3.489997 (lr=1.0104e-03) (hash(x)=42504837)
392
+ 3130 train 3.609744 (lr=1.0002e-03) (hash(x)=39693594)
393
+ 3140 train 3.504098 (lr=9.9011e-04) (hash(x)=40277645)
394
+ 3150 train 3.502033 (lr=9.8002e-04) (hash(x)=40988003)
395
+ 3160 train 3.624182 (lr=9.6999e-04) (hash(x)=38875266)
396
+ 3170 train 3.574172 (lr=9.6000e-04) (hash(x)=45232173)
397
+ 3180 train 3.650594 (lr=9.5007e-04) (hash(x)=39213336)
398
+ 3190 train 3.701874 (lr=9.4019e-04) (hash(x)=42118576)
399
+ 3200 val loss 3.5655
400
+ 3200 val perplexity 35.3559
401
+ 3200 train 3.556341 (lr=9.3036e-04) (hash(x)=32884223)
402
+ 3210 train 3.603194 (lr=9.2058e-04) (hash(x)=41276800)
403
+ 3220 train 3.656444 (lr=9.1085e-04) (hash(x)=40284461)
404
+ 3230 train 3.638096 (lr=9.0118e-04) (hash(x)=40566734)
405
+ 3240 train 3.626228 (lr=8.9157e-04) (hash(x)=36484570)
406
+ 3250 hella 0.2601
407
+ 3250 train 3.593519 (lr=8.8201e-04) (hash(x)=41642338)
408
+ 3260 train 3.546887 (lr=8.7251e-04) (hash(x)=43883570)
409
+ 3270 train 3.505651 (lr=8.6307e-04) (hash(x)=40432560)
410
+ 3280 train 3.474246 (lr=8.5368e-04) (hash(x)=38002717)
411
+ 3290 train 3.418109 (lr=8.4435e-04) (hash(x)=41926004)
412
+ 3300 val loss 3.5568
413
+ 3300 val perplexity 35.0509
414
+ 3300 train 3.568847 (lr=8.3508e-04) (hash(x)=46015509)
415
+ 3310 train 3.602935 (lr=8.2588e-04) (hash(x)=40112249)
416
+ 3320 train 3.559838 (lr=8.1673e-04) (hash(x)=49162296)
417
+ 3330 train 3.651586 (lr=8.0764e-04) (hash(x)=46606969)
418
+ 3340 train 3.648204 (lr=7.9862e-04) (hash(x)=31128992)
419
+ 3350 train 3.606328 (lr=7.8966e-04) (hash(x)=41232534)
420
+ 3360 train 3.467052 (lr=7.8076e-04) (hash(x)=41599699)
421
+ 3370 train 3.649995 (lr=7.7192e-04) (hash(x)=40885280)
422
+ 3380 train 3.540782 (lr=7.6315e-04) (hash(x)=37169148)
423
+ 3390 train 3.482176 (lr=7.5445e-04) (hash(x)=46581889)
424
+ 3400 val loss 3.5402
425
+ 3400 val perplexity 34.4746
426
+ 3400 train 3.500167 (lr=7.4581e-04) (hash(x)=37612074)
427
+ 3410 train 3.575910 (lr=7.3724e-04) (hash(x)=38763316)
428
+ 3420 train 3.411633 (lr=7.2874e-04) (hash(x)=40562379)
429
+ 3430 train 3.503099 (lr=7.2030e-04) (hash(x)=44724867)
430
+ 3440 train 3.690131 (lr=7.1193e-04) (hash(x)=39386624)
431
+ 3450 train 3.626503 (lr=7.0363e-04) (hash(x)=38652923)
432
+ 3460 train 3.636847 (lr=6.9541e-04) (hash(x)=29587379)
433
+ 3470 train 3.530133 (lr=6.8725e-04) (hash(x)=38228776)
434
+ 3480 train 3.535330 (lr=6.7916e-04) (hash(x)=30076039)
435
+ 3490 train 3.581557 (lr=6.7114e-04) (hash(x)=38398908)
436
+ 3500 val loss 3.5226
437
+ 3500 val perplexity 33.8731
438
+ 3500 hella 0.2617
439
+ 3500 train 3.591099 (lr=6.6320e-04) (hash(x)=39259918)
440
+ 3510 train 3.485363 (lr=6.5533e-04) (hash(x)=40664091)
441
+ 3520 train 3.532675 (lr=6.4753e-04) (hash(x)=40309647)
442
+ 3530 train 3.495113 (lr=6.3981e-04) (hash(x)=38424801)
443
+ 3540 train 3.416564 (lr=6.3216e-04) (hash(x)=51322307)
444
+ 3550 train 3.502070 (lr=6.2458e-04) (hash(x)=38192628)
445
+ 3560 train 3.518896 (lr=6.1708e-04) (hash(x)=41563952)
446
+ 3570 train 3.502570 (lr=6.0966e-04) (hash(x)=39508843)
447
+ 3580 train 3.547942 (lr=6.0231e-04) (hash(x)=41260225)
448
+ 3590 train 3.503125 (lr=5.9504e-04) (hash(x)=49098107)
449
+ 3600 val loss 3.5104
450
+ 3600 val perplexity 33.4628
451
+ 3600 train 3.482378 (lr=5.8785e-04) (hash(x)=41194370)
452
+ 3610 train 3.640460 (lr=5.8074e-04) (hash(x)=40399152)
453
+ 3620 train 3.568552 (lr=5.7370e-04) (hash(x)=41625018)
454
+ 3630 train 3.629507 (lr=5.6675e-04) (hash(x)=44077942)
455
+ 3640 train 3.605045 (lr=5.5987e-04) (hash(x)=42427512)
456
+ 3650 train 3.562643 (lr=5.5308e-04) (hash(x)=40067454)
457
+ 3660 train 3.526577 (lr=5.4636e-04) (hash(x)=37969892)
458
+ 3670 train 3.484229 (lr=5.3973e-04) (hash(x)=43670246)
459
+ 3680 train 3.596122 (lr=5.3318e-04) (hash(x)=37031732)
460
+ 3690 train 3.484931 (lr=5.2671e-04) (hash(x)=39237187)
461
+ 3700 val loss 3.5013
462
+ 3700 val perplexity 33.1577
463
+ 3700 train 3.497904 (lr=5.2033e-04) (hash(x)=39521416)
464
+ 3710 train 3.397075 (lr=5.1402e-04) (hash(x)=41147425)
465
+ 3720 train 3.397339 (lr=5.0780e-04) (hash(x)=40204658)
466
+ 3730 train 3.484844 (lr=5.0167e-04) (hash(x)=39420983)
467
+ 3740 train 3.506073 (lr=4.9562e-04) (hash(x)=51155740)
468
+ 3750 hella 0.2648
469
+ 3750 train 3.492072 (lr=4.8965e-04) (hash(x)=41654187)
470
+ 3760 train 3.461140 (lr=4.8377e-04) (hash(x)=40598339)
471
+ 3770 train 3.423937 (lr=4.7798e-04) (hash(x)=44950116)
472
+ 3780 train 3.500422 (lr=4.7227e-04) (hash(x)=40660326)
473
+ 3790 train 3.552405 (lr=4.6665e-04) (hash(x)=37805851)
474
+ 3800 val loss 3.4836
475
+ 3800 val perplexity 32.5783
476
+ 3800 train 3.514779 (lr=4.6112e-04) (hash(x)=38064443)
477
+ 3810 train 3.565730 (lr=4.5567e-04) (hash(x)=39161498)
478
+ 3820 train 3.444922 (lr=4.5031e-04) (hash(x)=40626649)
479
+ 3830 train 3.503605 (lr=4.4504e-04) (hash(x)=36894771)
480
+ 3840 train 3.449375 (lr=4.3986e-04) (hash(x)=33672141)
481
+ 3850 train 3.411830 (lr=4.3477e-04) (hash(x)=41751813)
482
+ 3860 train 3.271091 (lr=4.2977e-04) (hash(x)=33784172)
483
+ 3870 train 3.216483 (lr=4.2486e-04) (hash(x)=43110493)
484
+ 3880 train 3.306143 (lr=4.2004e-04) (hash(x)=36585111)
485
+ 3890 train 3.328751 (lr=4.1530e-04) (hash(x)=40134264)
486
+ 3900 val loss 3.4861
487
+ 3900 val perplexity 32.6579
488
+ 3900 train 3.288305 (lr=4.1066e-04) (hash(x)=39815215)
489
+ 3910 train 3.407058 (lr=4.0611e-04) (hash(x)=42081557)
490
+ 3920 train 3.328791 (lr=4.0166e-04) (hash(x)=48184973)
491
+ 3930 train 3.580671 (lr=3.9729e-04) (hash(x)=42416681)
492
+ 3940 train 3.487303 (lr=3.9302e-04) (hash(x)=38587062)
493
+ 3950 train 3.512002 (lr=3.8884e-04) (hash(x)=37296617)
494
+ 3960 train 3.451003 (lr=3.8475e-04) (hash(x)=37959330)
495
+ 3970 train 3.517230 (lr=3.8076e-04) (hash(x)=35175634)
496
+ 3980 train 3.531309 (lr=3.7685e-04) (hash(x)=37219362)
497
+ 3990 train 3.551440 (lr=3.7305e-04) (hash(x)=38906132)
498
+ 4000 val loss 3.4666
499
+ 4000 val perplexity 32.0283
500
+ 4000 hella 0.2639
501
+ 4000 train 3.465343 (lr=3.6933e-04) (hash(x)=39940517)
502
+ 4010 train 3.516365 (lr=3.6572e-04) (hash(x)=39952437)
503
+ 4020 train 3.414320 (lr=3.6219e-04) (hash(x)=42000930)
504
+ 4030 train 3.447486 (lr=3.5876e-04) (hash(x)=41460672)
505
+ 4040 train 3.170191 (lr=3.5543e-04) (hash(x)=43702329)
506
+ 4050 train 3.271285 (lr=3.5219e-04) (hash(x)=41740121)
507
+ 4060 train 3.261325 (lr=3.4905e-04) (hash(x)=37066349)
508
+ 4070 train 3.094980 (lr=3.4600e-04) (hash(x)=38359869)
509
+ 4080 train 3.372556 (lr=3.4305e-04) (hash(x)=35131370)
510
+ 4090 train 3.390840 (lr=3.4019e-04) (hash(x)=39918811)
511
+ 4100 val loss 3.4636
512
+ 4100 val perplexity 31.9308
513
+ 4100 train 3.544402 (lr=3.3744e-04) (hash(x)=47036374)
514
+ 4110 train 3.484908 (lr=3.3477e-04) (hash(x)=37657523)
515
+ 4120 train 3.530331 (lr=3.3221e-04) (hash(x)=37218703)
516
+ 4130 train 3.530121 (lr=3.2974e-04) (hash(x)=41350513)
517
+ 4140 train 3.556936 (lr=3.2737e-04) (hash(x)=38283785)
518
+ 4150 train 3.551051 (lr=3.2510e-04) (hash(x)=32927892)
519
+ 4160 train 3.445873 (lr=3.2292e-04) (hash(x)=42011933)
520
+ 4170 train 3.457908 (lr=3.2085e-04) (hash(x)=41854594)
521
+ 4180 train 3.454015 (lr=3.1887e-04) (hash(x)=36737064)
522
+ 4190 train 3.395487 (lr=3.1699e-04) (hash(x)=36326176)
523
+ 4200 val loss 3.4548
524
+ 4200 val perplexity 31.6510
525
+ 4200 train 3.426733 (lr=3.1520e-04) (hash(x)=39106683)
526
+ 4210 train 3.520210 (lr=3.1352e-04) (hash(x)=39759977)
527
+ 4220 train 3.222758 (lr=3.1193e-04) (hash(x)=43666151)
528
+ 4230 train 3.177455 (lr=3.1044e-04) (hash(x)=38154367)
529
+ 4240 train 3.306004 (lr=3.0905e-04) (hash(x)=38284168)
530
+ 4250 hella 0.2666
531
+ 4250 train 3.241930 (lr=3.0776e-04) (hash(x)=38888879)
532
+ 4260 train 3.157991 (lr=3.0657e-04) (hash(x)=36846417)
533
+ 4270 train 3.492629 (lr=3.0548e-04) (hash(x)=43125375)
534
+ 4280 train 3.500942 (lr=3.0449e-04) (hash(x)=43556412)
535
+ 4290 train 3.429575 (lr=3.0359e-04) (hash(x)=48598302)
536
+ 4300 val loss 3.4489
537
+ 4300 val perplexity 31.4651
538
+ 4300 train 3.410349 (lr=3.0280e-04) (hash(x)=42764385)
539
+ 4310 train 3.448474 (lr=3.0210e-04) (hash(x)=41190028)
540
+ 4320 train 3.451319 (lr=3.0150e-04) (hash(x)=35767770)
541
+ 4330 train 3.522011 (lr=3.0101e-04) (hash(x)=36760476)
542
+ 4340 train 3.441943 (lr=3.0061e-04) (hash(x)=40618921)
543
+ 4350 train 3.567061 (lr=3.0031e-04) (hash(x)=32300164)
544
+ 4360 train 3.431387 (lr=3.0011e-04) (hash(x)=41400085)
545
+ 4370 train 3.456936 (lr=3.0001e-04) (hash(x)=40251828)
546
+ 4374 val loss 3.4479
547
+ 4374 val perplexity 31.4330
548
+ 4374 hella 0.2669
logs/64_heads_bump_comparison/13_heads_no_compile_seed_1340/model_04374.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:e9fdfadf934b0720762a18b075ab87a49b537ae47a99eb1bc9684da181ab12d5
3
+ size 557599490
logs/64_heads_bump_comparison/13_heads_no_compile_seed_1340/optimizer_04374.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:24b65907d6b6dc65fa4cdbf0f2a19424ffd15d83cc4649a5af0baced835476af
3
+ size 1014545926