andrew-healey commited on
Commit
61a2f13
·
verified ·
1 Parent(s): 8fc0cc1

Upload folder using huggingface_hub

Browse files
logs/fix_1_latent_mask/1_latent_mask_lr_35e-4_n_latent_masks_2_no_compile_seed_1340/args.json ADDED
@@ -0,0 +1 @@
 
 
1
+ {"hellaswag": true, "attention_kind": "selective", "log_dir": "logs/fix_1_latent_mask/1_latent_mask_lr_35e-4_n_latent_masks_2_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": 12, "n_embd": 264, "head_dim": 22, "protect_bos_token": true, "prevent_from_masking_myself": true, "max_steps": 4375, "warmup_steps": 250, "group": "fix_1_latent_mask", "use_wandb": true, "kill_self_after_run": false, "random_seed": 1340, "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, "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": 32, "total_batch_size": 131072, "protection_head_scaling_factor": 1.0, "protection_head_bias": 0.0, "n_sliced_masks": null, "n_latent_masks": 2, "init_latent_masks_to_identity": true, "latent_mask_scale": null, "latent_mask_sigmoid": false, "S_layernorm": false, "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": 0.0035, "decay_lr": true, "readout_zero_init": false, "query_zero_init": false, "l1_loss": false, "debugpy": false, "key": "1_latent_mask_lr_35e-4_n_latent_masks_2_no_compile"}
logs/fix_1_latent_mask/1_latent_mask_lr_35e-4_n_latent_masks_2_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/fix_1_latent_mask/1_latent_mask_lr_35e-4_n_latent_masks_2_no_compile_seed_1340/log2.txt ADDED
@@ -0,0 +1,548 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ max_steps: 4375
2
+ 0 val loss 11.2068
3
+ 0 val perplexity 73626.5078
4
+ 0 hella 0.2453
5
+ 0 train 11.205779 (lr=4.8951e-06) (hash(x)=45482580)
6
+ 10 train 9.812593 (lr=5.3846e-05) (hash(x)=38414730)
7
+ 20 train 9.237236 (lr=1.0280e-04) (hash(x)=39783917)
8
+ 30 train 8.427109 (lr=1.5175e-04) (hash(x)=38395733)
9
+ 40 train 7.774746 (lr=2.0070e-04) (hash(x)=37486707)
10
+ 50 train 7.614880 (lr=2.4965e-04) (hash(x)=39897505)
11
+ 60 train 7.530652 (lr=2.9860e-04) (hash(x)=40933473)
12
+ 70 train 7.508999 (lr=3.4755e-04) (hash(x)=37690746)
13
+ 80 train 7.367839 (lr=3.9650e-04) (hash(x)=38609656)
14
+ 90 train 7.127762 (lr=4.4545e-04) (hash(x)=38356571)
15
+ 100 val loss 7.0330
16
+ 100 val perplexity 1133.4333
17
+ 100 train 7.014496 (lr=4.9441e-04) (hash(x)=35980376)
18
+ 110 train 6.914314 (lr=5.4336e-04) (hash(x)=42011042)
19
+ 120 train 6.822062 (lr=5.9231e-04) (hash(x)=40266823)
20
+ 130 train 6.663940 (lr=6.4126e-04) (hash(x)=38645447)
21
+ 140 train 6.539555 (lr=6.9021e-04) (hash(x)=36059313)
22
+ 150 train 6.489724 (lr=7.3916e-04) (hash(x)=34161947)
23
+ 160 train 6.364343 (lr=7.8811e-04) (hash(x)=42263375)
24
+ 170 train 6.415509 (lr=8.3706e-04) (hash(x)=42957725)
25
+ 180 train 6.413583 (lr=8.8601e-04) (hash(x)=40198018)
26
+ 190 train 6.160376 (lr=9.3497e-04) (hash(x)=41666215)
27
+ 200 val loss 6.1414
28
+ 200 val perplexity 464.7236
29
+ 200 train 6.102100 (lr=9.8392e-04) (hash(x)=54060482)
30
+ 210 train 6.010913 (lr=1.0329e-03) (hash(x)=42332778)
31
+ 220 train 5.938154 (lr=1.0818e-03) (hash(x)=39551486)
32
+ 230 train 5.968668 (lr=1.1308e-03) (hash(x)=38462018)
33
+ 240 train 6.010826 (lr=1.1797e-03) (hash(x)=36591442)
34
+ 250 hella 0.2387
35
+ 250 train 5.957157 (lr=1.2287e-03) (hash(x)=40861237)
36
+ 260 train 5.900552 (lr=1.2776e-03) (hash(x)=41739369)
37
+ 270 train 5.766555 (lr=1.3266e-03) (hash(x)=39415360)
38
+ 280 train 5.745899 (lr=1.3755e-03) (hash(x)=44665934)
39
+ 290 train 5.696787 (lr=1.4245e-03) (hash(x)=37035579)
40
+ 300 val loss 5.6736
41
+ 300 val perplexity 291.0877
42
+ 300 train 5.696092 (lr=1.4734e-03) (hash(x)=38301011)
43
+ 310 train 5.530906 (lr=1.5224e-03) (hash(x)=42850980)
44
+ 320 train 5.612459 (lr=1.5713e-03) (hash(x)=37476222)
45
+ 330 train 5.519908 (lr=1.6203e-03) (hash(x)=53028205)
46
+ 340 train 5.536469 (lr=1.6692e-03) (hash(x)=41466008)
47
+ 350 train 5.441945 (lr=1.7182e-03) (hash(x)=37802865)
48
+ 360 train 5.395929 (lr=1.7671e-03) (hash(x)=42769282)
49
+ 370 train 5.420222 (lr=1.8161e-03) (hash(x)=39319256)
50
+ 380 train 5.396572 (lr=1.8650e-03) (hash(x)=42637402)
51
+ 390 train 5.372044 (lr=1.9140e-03) (hash(x)=31141514)
52
+ 400 val loss 5.3507
53
+ 400 val perplexity 210.7647
54
+ 400 train 5.428014 (lr=1.9629e-03) (hash(x)=38151157)
55
+ 410 train 5.364438 (lr=2.0119e-03) (hash(x)=46033439)
56
+ 420 train 5.348393 (lr=2.0608e-03) (hash(x)=41365246)
57
+ 430 train 5.328675 (lr=2.1098e-03) (hash(x)=42369184)
58
+ 440 train 5.337112 (lr=2.1587e-03) (hash(x)=42004840)
59
+ 450 train 5.244347 (lr=2.2077e-03) (hash(x)=37181172)
60
+ 460 train 5.330682 (lr=2.2566e-03) (hash(x)=31630797)
61
+ 470 train 5.197980 (lr=2.3056e-03) (hash(x)=42135747)
62
+ 480 train 5.116594 (lr=2.3545e-03) (hash(x)=30023651)
63
+ 490 train 5.201200 (lr=2.4035e-03) (hash(x)=40463476)
64
+ 500 val loss 5.1559
65
+ 500 val perplexity 173.4585
66
+ 500 hella 0.2393
67
+ 500 train 5.081638 (lr=2.4524e-03) (hash(x)=37089842)
68
+ 510 train 5.087924 (lr=2.5014e-03) (hash(x)=41095974)
69
+ 520 train 5.149768 (lr=2.5503e-03) (hash(x)=41525258)
70
+ 530 train 5.104909 (lr=2.5993e-03) (hash(x)=37697291)
71
+ 540 train 5.134089 (lr=2.6483e-03) (hash(x)=43459894)
72
+ 550 train 5.005734 (lr=2.6972e-03) (hash(x)=38993956)
73
+ 560 train 5.048613 (lr=2.7462e-03) (hash(x)=40677277)
74
+ 570 train 5.090444 (lr=2.7951e-03) (hash(x)=43103127)
75
+ 580 train 5.007751 (lr=2.8441e-03) (hash(x)=41871776)
76
+ 590 train 5.019064 (lr=2.8930e-03) (hash(x)=43230967)
77
+ 600 val loss 4.9280
78
+ 600 val perplexity 138.1039
79
+ 600 train 4.974318 (lr=2.9420e-03) (hash(x)=42226172)
80
+ 610 train 5.315032 (lr=2.9909e-03) (hash(x)=43007164)
81
+ 620 train 4.850182 (lr=3.0399e-03) (hash(x)=35752341)
82
+ 630 train 4.938048 (lr=3.0888e-03) (hash(x)=39453504)
83
+ 640 train 4.869149 (lr=3.1378e-03) (hash(x)=43635712)
84
+ 650 train 4.875855 (lr=3.1867e-03) (hash(x)=41852456)
85
+ 660 train 4.896130 (lr=3.2357e-03) (hash(x)=49125742)
86
+ 670 train 4.889990 (lr=3.2846e-03) (hash(x)=34886275)
87
+ 680 train 4.984262 (lr=3.3336e-03) (hash(x)=40546482)
88
+ 690 train 4.800055 (lr=3.3825e-03) (hash(x)=40850901)
89
+ 700 val loss 4.7828
90
+ 700 val perplexity 119.4384
91
+ 700 train 4.745400 (lr=3.4315e-03) (hash(x)=37426993)
92
+ 710 train 4.929419 (lr=3.4804e-03) (hash(x)=42679743)
93
+ 720 train 4.751391 (lr=3.5000e-03) (hash(x)=44856043)
94
+ 730 train 4.734615 (lr=3.4999e-03) (hash(x)=36457471)
95
+ 740 train 4.692545 (lr=3.4996e-03) (hash(x)=47203534)
96
+ 750 hella 0.2475
97
+ 750 train 4.721806 (lr=3.4993e-03) (hash(x)=35585657)
98
+ 760 train 4.628271 (lr=3.4988e-03) (hash(x)=37259264)
99
+ 770 train 4.545980 (lr=3.4982e-03) (hash(x)=44161997)
100
+ 780 train 4.492832 (lr=3.4975e-03) (hash(x)=41272886)
101
+ 790 train 4.493739 (lr=3.4967e-03) (hash(x)=43303662)
102
+ 800 val loss 4.6973
103
+ 800 val perplexity 109.6485
104
+ 800 train 4.408896 (lr=3.4958e-03) (hash(x)=39067231)
105
+ 810 train 4.639615 (lr=3.4948e-03) (hash(x)=44046732)
106
+ 820 train 4.618834 (lr=3.4936e-03) (hash(x)=39783962)
107
+ 830 train 4.681157 (lr=3.4923e-03) (hash(x)=40973974)
108
+ 840 train 4.656079 (lr=3.4909e-03) (hash(x)=41619968)
109
+ 850 train 4.708729 (lr=3.4894e-03) (hash(x)=36939960)
110
+ 860 train 4.589850 (lr=3.4878e-03) (hash(x)=42942377)
111
+ 870 train 4.593907 (lr=3.4861e-03) (hash(x)=39659455)
112
+ 880 train 4.610484 (lr=3.4842e-03) (hash(x)=40331986)
113
+ 890 train 4.522497 (lr=3.4823e-03) (hash(x)=38084814)
114
+ 900 val loss 4.5575
115
+ 900 val perplexity 95.3409
116
+ 900 train 4.550611 (lr=3.4802e-03) (hash(x)=30830367)
117
+ 910 train 4.464406 (lr=3.4780e-03) (hash(x)=39664356)
118
+ 920 train 4.472213 (lr=3.4757e-03) (hash(x)=39007775)
119
+ 930 train 4.490801 (lr=3.4733e-03) (hash(x)=39319254)
120
+ 940 train 4.535435 (lr=3.4707e-03) (hash(x)=36347051)
121
+ 950 train 4.500622 (lr=3.4681e-03) (hash(x)=39243577)
122
+ 960 train 4.524755 (lr=3.4653e-03) (hash(x)=45603839)
123
+ 970 train 4.330237 (lr=3.4624e-03) (hash(x)=40036075)
124
+ 980 train 4.384459 (lr=3.4594e-03) (hash(x)=43068524)
125
+ 990 train 4.365358 (lr=3.4563e-03) (hash(x)=33798472)
126
+ 1000 val loss 4.5283
127
+ 1000 val perplexity 92.6008
128
+ 1000 hella 0.2504
129
+ 1000 train 4.736703 (lr=3.4531e-03) (hash(x)=37728665)
130
+ 1010 train 4.596229 (lr=3.4498e-03) (hash(x)=43678636)
131
+ 1020 train 4.621488 (lr=3.4463e-03) (hash(x)=48464376)
132
+ 1030 train 4.586704 (lr=3.4428e-03) (hash(x)=43771461)
133
+ 1040 train 4.443330 (lr=3.4391e-03) (hash(x)=44643085)
134
+ 1050 train 4.527156 (lr=3.4353e-03) (hash(x)=39201716)
135
+ 1060 train 4.435090 (lr=3.4314e-03) (hash(x)=36396343)
136
+ 1070 train 4.431265 (lr=3.4274e-03) (hash(x)=37504111)
137
+ 1080 train 4.460018 (lr=3.4233e-03) (hash(x)=37454537)
138
+ 1090 train 4.461719 (lr=3.4191e-03) (hash(x)=36888183)
139
+ 1100 val loss 4.4496
140
+ 1100 val perplexity 85.5910
141
+ 1100 train 4.398065 (lr=3.4148e-03) (hash(x)=39549045)
142
+ 1110 train 4.594175 (lr=3.4103e-03) (hash(x)=56818420)
143
+ 1120 train 4.401610 (lr=3.4058e-03) (hash(x)=36573357)
144
+ 1130 train 4.336138 (lr=3.4011e-03) (hash(x)=38540878)
145
+ 1140 train 4.329080 (lr=3.3964e-03) (hash(x)=38645859)
146
+ 1150 train 4.538975 (lr=3.3915e-03) (hash(x)=41972503)
147
+ 1160 train 4.153083 (lr=3.3865e-03) (hash(x)=38151868)
148
+ 1170 train 4.187085 (lr=3.3814e-03) (hash(x)=42375436)
149
+ 1180 train 4.173806 (lr=3.3762e-03) (hash(x)=42868028)
150
+ 1190 train 4.421863 (lr=3.3709e-03) (hash(x)=41060868)
151
+ 1200 val loss 4.4096
152
+ 1200 val perplexity 82.2333
153
+ 1200 train 4.424465 (lr=3.3655e-03) (hash(x)=40852814)
154
+ 1210 train 4.407389 (lr=3.3600e-03) (hash(x)=43365554)
155
+ 1220 train 4.349785 (lr=3.3543e-03) (hash(x)=39121134)
156
+ 1230 train 4.434183 (lr=3.3486e-03) (hash(x)=40565426)
157
+ 1240 train 4.494906 (lr=3.3428e-03) (hash(x)=43378926)
158
+ 1250 hella 0.2556
159
+ 1250 train 4.356088 (lr=3.3368e-03) (hash(x)=35227381)
160
+ 1260 train 4.450854 (lr=3.3308e-03) (hash(x)=41267226)
161
+ 1270 train 4.357942 (lr=3.3246e-03) (hash(x)=31395565)
162
+ 1280 train 4.475410 (lr=3.3184e-03) (hash(x)=41470963)
163
+ 1290 train 4.285267 (lr=3.3120e-03) (hash(x)=40729366)
164
+ 1300 val loss 4.3517
165
+ 1300 val perplexity 77.6083
166
+ 1300 train 4.324142 (lr=3.3056e-03) (hash(x)=39916663)
167
+ 1310 train 4.372407 (lr=3.2990e-03) (hash(x)=42179962)
168
+ 1320 train 4.239473 (lr=3.2924e-03) (hash(x)=44991932)
169
+ 1330 train 4.204738 (lr=3.2856e-03) (hash(x)=38925911)
170
+ 1340 train 4.157549 (lr=3.2787e-03) (hash(x)=38565560)
171
+ 1350 train 4.203199 (lr=3.2718e-03) (hash(x)=43384902)
172
+ 1360 train 4.153557 (lr=3.2647e-03) (hash(x)=45422261)
173
+ 1370 train 4.134592 (lr=3.2576e-03) (hash(x)=40167325)
174
+ 1380 train 4.293466 (lr=3.2503e-03) (hash(x)=51258162)
175
+ 1390 train 4.351717 (lr=3.2430e-03) (hash(x)=43229106)
176
+ 1400 val loss 4.3073
177
+ 1400 val perplexity 74.2422
178
+ 1400 train 4.256281 (lr=3.2355e-03) (hash(x)=42537494)
179
+ 1410 train 4.562027 (lr=3.2280e-03) (hash(x)=51236490)
180
+ 1420 train 4.409715 (lr=3.2203e-03) (hash(x)=45243347)
181
+ 1430 train 4.243119 (lr=3.2126e-03) (hash(x)=37658566)
182
+ 1440 train 4.306640 (lr=3.2047e-03) (hash(x)=38856106)
183
+ 1450 train 4.376338 (lr=3.1968e-03) (hash(x)=36649488)
184
+ 1460 train 4.374071 (lr=3.1888e-03) (hash(x)=40445407)
185
+ 1470 train 4.311240 (lr=3.1807e-03) (hash(x)=39916387)
186
+ 1480 train 4.275204 (lr=3.1725e-03) (hash(x)=40934555)
187
+ 1490 train 4.239572 (lr=3.1642e-03) (hash(x)=42276527)
188
+ 1500 val loss 4.2939
189
+ 1500 val perplexity 73.2508
190
+ 1500 hella 0.2533
191
+ 1500 train 4.231426 (lr=3.1558e-03) (hash(x)=37973883)
192
+ 1510 train 4.291758 (lr=3.1473e-03) (hash(x)=38929286)
193
+ 1520 train 4.115152 (lr=3.1387e-03) (hash(x)=36855837)
194
+ 1530 train 4.358483 (lr=3.1301e-03) (hash(x)=42146431)
195
+ 1540 train 4.304111 (lr=3.1213e-03) (hash(x)=41898411)
196
+ 1550 train 4.211119 (lr=3.1125e-03) (hash(x)=44906272)
197
+ 1560 train 4.241699 (lr=3.1036e-03) (hash(x)=40414353)
198
+ 1570 train 4.269977 (lr=3.0945e-03) (hash(x)=38723190)
199
+ 1580 train 4.186987 (lr=3.0855e-03) (hash(x)=39065271)
200
+ 1590 train 4.181960 (lr=3.0763e-03) (hash(x)=40000886)
201
+ 1600 val loss 4.2684
202
+ 1600 val perplexity 71.4057
203
+ 1600 train 4.101627 (lr=3.0670e-03) (hash(x)=37898571)
204
+ 1610 train 4.210349 (lr=3.0576e-03) (hash(x)=40447863)
205
+ 1620 train 4.174629 (lr=3.0482e-03) (hash(x)=40119318)
206
+ 1630 train 4.107739 (lr=3.0387e-03) (hash(x)=40664074)
207
+ 1640 train 4.071666 (lr=3.0291e-03) (hash(x)=37404741)
208
+ 1650 train 4.232249 (lr=3.0194e-03) (hash(x)=39214928)
209
+ 1660 train 4.274699 (lr=3.0096e-03) (hash(x)=35005019)
210
+ 1670 train 4.313293 (lr=2.9998e-03) (hash(x)=37675832)
211
+ 1680 train 4.252539 (lr=2.9899e-03) (hash(x)=40489680)
212
+ 1690 train 4.164642 (lr=2.9799e-03) (hash(x)=45537879)
213
+ 1700 val loss 4.2327
214
+ 1700 val perplexity 68.9011
215
+ 1700 train 4.145432 (lr=2.9698e-03) (hash(x)=38866100)
216
+ 1710 train 4.217449 (lr=2.9597e-03) (hash(x)=44726254)
217
+ 1720 train 4.185666 (lr=2.9494e-03) (hash(x)=36917792)
218
+ 1730 train 4.172013 (lr=2.9391e-03) (hash(x)=42814805)
219
+ 1740 train 4.248007 (lr=2.9287e-03) (hash(x)=51812216)
220
+ 1750 hella 0.2528
221
+ 1750 train 4.114487 (lr=2.9183e-03) (hash(x)=33549014)
222
+ 1760 train 4.129948 (lr=2.9078e-03) (hash(x)=40354215)
223
+ 1770 train 4.150578 (lr=2.8972e-03) (hash(x)=40718606)
224
+ 1780 train 4.084775 (lr=2.8865e-03) (hash(x)=38274164)
225
+ 1790 train 4.263676 (lr=2.8757e-03) (hash(x)=42531471)
226
+ 1800 val loss 4.2129
227
+ 1800 val perplexity 67.5501
228
+ 1800 train 4.191330 (lr=2.8649e-03) (hash(x)=35616519)
229
+ 1810 train 4.245029 (lr=2.8540e-03) (hash(x)=33803118)
230
+ 1820 train 4.244155 (lr=2.8431e-03) (hash(x)=36973525)
231
+ 1830 train 4.184198 (lr=2.8321e-03) (hash(x)=38162549)
232
+ 1840 train 4.153059 (lr=2.8210e-03) (hash(x)=33937159)
233
+ 1850 train 4.265369 (lr=2.8098e-03) (hash(x)=39887546)
234
+ 1860 train 4.091496 (lr=2.7986e-03) (hash(x)=37818525)
235
+ 1870 train 4.148082 (lr=2.7873e-03) (hash(x)=37250478)
236
+ 1880 train 4.039495 (lr=2.7760e-03) (hash(x)=36412167)
237
+ 1890 train 4.149627 (lr=2.7646e-03) (hash(x)=40261189)
238
+ 1900 val loss 4.2079
239
+ 1900 val perplexity 67.2125
240
+ 1900 train 4.105598 (lr=2.7531e-03) (hash(x)=38654303)
241
+ 1910 train 4.079849 (lr=2.7416e-03) (hash(x)=37929515)
242
+ 1920 train 4.258565 (lr=2.7300e-03) (hash(x)=45580146)
243
+ 1930 train 4.203111 (lr=2.7183e-03) (hash(x)=32386330)
244
+ 1940 train 4.238146 (lr=2.7066e-03) (hash(x)=36331864)
245
+ 1950 train 4.216351 (lr=2.6949e-03) (hash(x)=41181727)
246
+ 1960 train 4.047747 (lr=2.6831e-03) (hash(x)=42705152)
247
+ 1970 train 4.158383 (lr=2.6712e-03) (hash(x)=40251511)
248
+ 1980 train 4.076075 (lr=2.6592e-03) (hash(x)=37525551)
249
+ 1990 train 4.149635 (lr=2.6473e-03) (hash(x)=41849618)
250
+ 2000 val loss 4.1764
251
+ 2000 val perplexity 65.1337
252
+ 2000 hella 0.2555
253
+ 2000 train 4.153386 (lr=2.6352e-03) (hash(x)=37642582)
254
+ 2010 train 4.155840 (lr=2.6231e-03) (hash(x)=42131121)
255
+ 2020 train 4.035791 (lr=2.6110e-03) (hash(x)=39000209)
256
+ 2030 train 4.124985 (lr=2.5988e-03) (hash(x)=43641355)
257
+ 2040 train 4.015590 (lr=2.5865e-03) (hash(x)=47910507)
258
+ 2050 train 4.018216 (lr=2.5742e-03) (hash(x)=36670359)
259
+ 2060 train 4.214338 (lr=2.5619e-03) (hash(x)=36477755)
260
+ 2070 train 4.306351 (lr=2.5495e-03) (hash(x)=43262487)
261
+ 2080 train 4.118608 (lr=2.5371e-03) (hash(x)=49546029)
262
+ 2090 train 4.250965 (lr=2.5246e-03) (hash(x)=39924731)
263
+ 2100 val loss 4.1613
264
+ 2100 val perplexity 64.1524
265
+ 2100 train 4.111554 (lr=2.5121e-03) (hash(x)=39921304)
266
+ 2110 train 4.063148 (lr=2.4995e-03) (hash(x)=38152788)
267
+ 2120 train 4.112677 (lr=2.4869e-03) (hash(x)=37977911)
268
+ 2130 train 4.074975 (lr=2.4743e-03) (hash(x)=34748760)
269
+ 2140 train 4.125198 (lr=2.4616e-03) (hash(x)=40509369)
270
+ 2150 train 4.099833 (lr=2.4488e-03) (hash(x)=37654262)
271
+ 2160 train 4.053321 (lr=2.4361e-03) (hash(x)=38139543)
272
+ 2170 train 3.955775 (lr=2.4233e-03) (hash(x)=42501806)
273
+ 2180 train 4.115121 (lr=2.4104e-03) (hash(x)=40085092)
274
+ 2190 train 4.056927 (lr=2.3975e-03) (hash(x)=50480193)
275
+ 2200 val loss 4.1441
276
+ 2200 val perplexity 63.0582
277
+ 2200 train 4.146177 (lr=2.3846e-03) (hash(x)=40604084)
278
+ 2210 train 4.254169 (lr=2.3717e-03) (hash(x)=41555823)
279
+ 2220 train 4.070233 (lr=2.3587e-03) (hash(x)=50441765)
280
+ 2230 train 4.173196 (lr=2.3457e-03) (hash(x)=39796580)
281
+ 2240 train 4.197507 (lr=2.3326e-03) (hash(x)=44127022)
282
+ 2250 hella 0.2531
283
+ 2250 train 4.097971 (lr=2.3196e-03) (hash(x)=37026826)
284
+ 2260 train 4.136424 (lr=2.3065e-03) (hash(x)=42133839)
285
+ 2270 train 4.093678 (lr=2.2933e-03) (hash(x)=38500664)
286
+ 2280 train 4.080942 (lr=2.2802e-03) (hash(x)=40538661)
287
+ 2290 train 4.090720 (lr=2.2670e-03) (hash(x)=51509210)
288
+ 2300 val loss 4.1194
289
+ 2300 val perplexity 61.5196
290
+ 2300 train 4.093284 (lr=2.2538e-03) (hash(x)=41952328)
291
+ 2310 train 4.016378 (lr=2.2405e-03) (hash(x)=39758123)
292
+ 2320 train 4.132858 (lr=2.2273e-03) (hash(x)=51089268)
293
+ 2330 train 4.192353 (lr=2.2140e-03) (hash(x)=39767618)
294
+ 2340 train 4.238276 (lr=2.2007e-03) (hash(x)=40409617)
295
+ 2350 train 4.067914 (lr=2.1874e-03) (hash(x)=40349634)
296
+ 2360 train 4.345475 (lr=2.1741e-03) (hash(x)=31841172)
297
+ 2370 train 4.137584 (lr=2.1607e-03) (hash(x)=42720539)
298
+ 2380 train 4.003537 (lr=2.1473e-03) (hash(x)=40998632)
299
+ 2390 train 4.108484 (lr=2.1339e-03) (hash(x)=40615413)
300
+ 2400 val loss 4.1024
301
+ 2400 val perplexity 60.4858
302
+ 2400 train 4.020265 (lr=2.1205e-03) (hash(x)=39373658)
303
+ 2410 train 4.063818 (lr=2.1071e-03) (hash(x)=35480858)
304
+ 2420 train 4.057462 (lr=2.0937e-03) (hash(x)=32241095)
305
+ 2430 train 4.046044 (lr=2.0802e-03) (hash(x)=36669715)
306
+ 2440 train 4.047603 (lr=2.0668e-03) (hash(x)=45768335)
307
+ 2450 train 3.952236 (lr=2.0533e-03) (hash(x)=39395055)
308
+ 2460 train 3.980482 (lr=2.0398e-03) (hash(x)=34899269)
309
+ 2470 train 4.168346 (lr=2.0263e-03) (hash(x)=43642420)
310
+ 2480 train 4.276309 (lr=2.0128e-03) (hash(x)=47099507)
311
+ 2490 train 4.147141 (lr=1.9993e-03) (hash(x)=38825558)
312
+ 2500 val loss 4.0890
313
+ 2500 val perplexity 59.6827
314
+ 2500 hella 0.2519
315
+ 2500 train 4.119624 (lr=1.9858e-03) (hash(x)=39833804)
316
+ 2510 train 4.215500 (lr=1.9723e-03) (hash(x)=35051654)
317
+ 2520 train 4.113863 (lr=1.9588e-03) (hash(x)=40291109)
318
+ 2530 train 4.022317 (lr=1.9453e-03) (hash(x)=36915768)
319
+ 2540 train 4.004189 (lr=1.9318e-03) (hash(x)=37016308)
320
+ 2550 train 3.967799 (lr=1.9182e-03) (hash(x)=43221777)
321
+ 2560 train 4.014973 (lr=1.9047e-03) (hash(x)=37233207)
322
+ 2570 train 3.914779 (lr=1.8912e-03) (hash(x)=37510626)
323
+ 2580 train 3.964463 (lr=1.8777e-03) (hash(x)=39624656)
324
+ 2590 train 3.823573 (lr=1.8642e-03) (hash(x)=37647501)
325
+ 2600 val loss 4.0787
326
+ 2600 val perplexity 59.0689
327
+ 2600 train 3.957325 (lr=1.8507e-03) (hash(x)=37200138)
328
+ 2610 train 3.966778 (lr=1.8372e-03) (hash(x)=43066970)
329
+ 2620 train 4.106297 (lr=1.8237e-03) (hash(x)=40503799)
330
+ 2630 train 4.045425 (lr=1.8102e-03) (hash(x)=45534660)
331
+ 2640 train 4.260882 (lr=1.7967e-03) (hash(x)=52358353)
332
+ 2650 train 4.040200 (lr=1.7832e-03) (hash(x)=41843599)
333
+ 2660 train 4.021623 (lr=1.7698e-03) (hash(x)=40519770)
334
+ 2670 train 4.100499 (lr=1.7563e-03) (hash(x)=37185602)
335
+ 2680 train 3.967336 (lr=1.7429e-03) (hash(x)=43805719)
336
+ 2690 train 3.976134 (lr=1.7295e-03) (hash(x)=39458250)
337
+ 2700 val loss 4.0539
338
+ 2700 val perplexity 57.6235
339
+ 2700 train 4.076973 (lr=1.7161e-03) (hash(x)=39149255)
340
+ 2710 train 3.922950 (lr=1.7027e-03) (hash(x)=43544861)
341
+ 2720 train 4.025866 (lr=1.6893e-03) (hash(x)=36858197)
342
+ 2730 train 3.944071 (lr=1.6759e-03) (hash(x)=37971712)
343
+ 2740 train 3.971146 (lr=1.6626e-03) (hash(x)=42416578)
344
+ 2750 hella 0.2561
345
+ 2750 train 3.916256 (lr=1.6493e-03) (hash(x)=53554622)
346
+ 2760 train 4.054587 (lr=1.6360e-03) (hash(x)=40502545)
347
+ 2770 train 4.018095 (lr=1.6227e-03) (hash(x)=36086417)
348
+ 2780 train 4.049534 (lr=1.6095e-03) (hash(x)=36242982)
349
+ 2790 train 4.120588 (lr=1.5962e-03) (hash(x)=41433780)
350
+ 2800 val loss 4.0345
351
+ 2800 val perplexity 56.5161
352
+ 2800 train 3.990044 (lr=1.5830e-03) (hash(x)=48553484)
353
+ 2810 train 3.933424 (lr=1.5698e-03) (hash(x)=39928650)
354
+ 2820 train 4.037273 (lr=1.5567e-03) (hash(x)=37891724)
355
+ 2830 train 4.011147 (lr=1.5435e-03) (hash(x)=34353412)
356
+ 2840 train 4.033251 (lr=1.5304e-03) (hash(x)=38486611)
357
+ 2850 train 3.954027 (lr=1.5174e-03) (hash(x)=32706934)
358
+ 2860 train 3.903740 (lr=1.5043e-03) (hash(x)=38693681)
359
+ 2870 train 3.968284 (lr=1.4913e-03) (hash(x)=38430800)
360
+ 2880 train 3.943277 (lr=1.4783e-03) (hash(x)=40619559)
361
+ 2890 train 3.921525 (lr=1.4654e-03) (hash(x)=42741066)
362
+ 2900 val loss 4.0360
363
+ 2900 val perplexity 56.6005
364
+ 2900 train 4.168953 (lr=1.4525e-03) (hash(x)=38216091)
365
+ 2910 train 3.971304 (lr=1.4396e-03) (hash(x)=35447832)
366
+ 2920 train 4.081658 (lr=1.4267e-03) (hash(x)=39825190)
367
+ 2930 train 4.101611 (lr=1.4139e-03) (hash(x)=37789121)
368
+ 2940 train 3.978005 (lr=1.4012e-03) (hash(x)=44499116)
369
+ 2950 train 4.057795 (lr=1.3884e-03) (hash(x)=39669860)
370
+ 2960 train 4.000934 (lr=1.3757e-03) (hash(x)=51683741)
371
+ 2970 train 3.864923 (lr=1.3631e-03) (hash(x)=36399721)
372
+ 2980 train 3.965534 (lr=1.3505e-03) (hash(x)=42629700)
373
+ 2990 train 3.847225 (lr=1.3379e-03) (hash(x)=39263773)
374
+ 3000 val loss 4.0119
375
+ 3000 val perplexity 55.2505
376
+ 3000 hella 0.2598
377
+ 3000 train 3.888361 (lr=1.3254e-03) (hash(x)=37920485)
378
+ 3010 train 3.934680 (lr=1.3129e-03) (hash(x)=40835161)
379
+ 3020 train 3.902698 (lr=1.3005e-03) (hash(x)=38381159)
380
+ 3030 train 3.894580 (lr=1.2881e-03) (hash(x)=35846270)
381
+ 3040 train 4.292038 (lr=1.2758e-03) (hash(x)=47516567)
382
+ 3050 train 4.119843 (lr=1.2635e-03) (hash(x)=35600311)
383
+ 3060 train 4.152466 (lr=1.2512e-03) (hash(x)=37775318)
384
+ 3070 train 4.010235 (lr=1.2390e-03) (hash(x)=39881333)
385
+ 3080 train 4.013050 (lr=1.2269e-03) (hash(x)=37310168)
386
+ 3090 train 3.943904 (lr=1.2148e-03) (hash(x)=43669978)
387
+ 3100 val loss 3.9993
388
+ 3100 val perplexity 54.5573
389
+ 3100 train 4.403669 (lr=1.2027e-03) (hash(x)=46356797)
390
+ 3110 train 4.043657 (lr=1.1908e-03) (hash(x)=40814232)
391
+ 3120 train 3.880990 (lr=1.1788e-03) (hash(x)=42504837)
392
+ 3130 train 4.019946 (lr=1.1669e-03) (hash(x)=39693594)
393
+ 3140 train 3.924436 (lr=1.1551e-03) (hash(x)=40277645)
394
+ 3150 train 3.933050 (lr=1.1434e-03) (hash(x)=40988003)
395
+ 3160 train 4.052250 (lr=1.1317e-03) (hash(x)=38875266)
396
+ 3170 train 4.003973 (lr=1.1200e-03) (hash(x)=45232173)
397
+ 3180 train 4.058941 (lr=1.1084e-03) (hash(x)=39213336)
398
+ 3190 train 4.117190 (lr=1.0969e-03) (hash(x)=42118576)
399
+ 3200 val loss 3.9781
400
+ 3200 val perplexity 53.4169
401
+ 3200 train 3.976986 (lr=1.0854e-03) (hash(x)=32884223)
402
+ 3210 train 4.020449 (lr=1.0740e-03) (hash(x)=41276800)
403
+ 3220 train 4.065345 (lr=1.0627e-03) (hash(x)=40284461)
404
+ 3230 train 4.046233 (lr=1.0514e-03) (hash(x)=40566734)
405
+ 3240 train 4.035906 (lr=1.0402e-03) (hash(x)=36484570)
406
+ 3250 hella 0.2558
407
+ 3250 train 3.995875 (lr=1.0290e-03) (hash(x)=41642338)
408
+ 3260 train 3.962234 (lr=1.0179e-03) (hash(x)=43883570)
409
+ 3270 train 3.902210 (lr=1.0069e-03) (hash(x)=40432560)
410
+ 3280 train 3.870857 (lr=9.9596e-04) (hash(x)=38002717)
411
+ 3290 train 3.818748 (lr=9.8508e-04) (hash(x)=41926004)
412
+ 3300 val loss 3.9703
413
+ 3300 val perplexity 52.9979
414
+ 3300 train 3.973510 (lr=9.7426e-04) (hash(x)=46015509)
415
+ 3310 train 4.011311 (lr=9.6352e-04) (hash(x)=40112249)
416
+ 3320 train 3.984941 (lr=9.5285e-04) (hash(x)=49162296)
417
+ 3330 train 4.045941 (lr=9.4225e-04) (hash(x)=46606969)
418
+ 3340 train 4.059370 (lr=9.3172e-04) (hash(x)=31128992)
419
+ 3350 train 4.034096 (lr=9.2127e-04) (hash(x)=41232534)
420
+ 3360 train 3.878155 (lr=9.1088e-04) (hash(x)=41599699)
421
+ 3370 train 4.050038 (lr=9.0058e-04) (hash(x)=40885280)
422
+ 3380 train 3.967752 (lr=8.9035e-04) (hash(x)=37169148)
423
+ 3390 train 3.870433 (lr=8.8019e-04) (hash(x)=46581889)
424
+ 3400 val loss 3.9576
425
+ 3400 val perplexity 52.3329
426
+ 3400 train 3.924501 (lr=8.7012e-04) (hash(x)=37612074)
427
+ 3410 train 3.996443 (lr=8.6012e-04) (hash(x)=38763316)
428
+ 3420 train 3.815307 (lr=8.5019e-04) (hash(x)=40562379)
429
+ 3430 train 3.920310 (lr=8.4035e-04) (hash(x)=44724867)
430
+ 3440 train 4.124741 (lr=8.3059e-04) (hash(x)=39386624)
431
+ 3450 train 4.072584 (lr=8.2091e-04) (hash(x)=38652923)
432
+ 3460 train 4.054446 (lr=8.1131e-04) (hash(x)=29587379)
433
+ 3470 train 3.958468 (lr=8.0179e-04) (hash(x)=38228776)
434
+ 3480 train 3.936267 (lr=7.9235e-04) (hash(x)=30076039)
435
+ 3490 train 3.998910 (lr=7.8300e-04) (hash(x)=38398908)
436
+ 3500 val loss 3.9443
437
+ 3500 val perplexity 51.6394
438
+ 3500 hella 0.2570
439
+ 3500 train 4.011887 (lr=7.7373e-04) (hash(x)=39259918)
440
+ 3510 train 3.910538 (lr=7.6455e-04) (hash(x)=40664091)
441
+ 3520 train 3.942012 (lr=7.5545e-04) (hash(x)=40309647)
442
+ 3530 train 3.908617 (lr=7.4644e-04) (hash(x)=38424801)
443
+ 3540 train 3.832606 (lr=7.3752e-04) (hash(x)=51322307)
444
+ 3550 train 3.899303 (lr=7.2868e-04) (hash(x)=38192628)
445
+ 3560 train 3.949500 (lr=7.1993e-04) (hash(x)=41563952)
446
+ 3570 train 3.898190 (lr=7.1127e-04) (hash(x)=39508843)
447
+ 3580 train 3.969242 (lr=7.0270e-04) (hash(x)=41260225)
448
+ 3590 train 3.951468 (lr=6.9422e-04) (hash(x)=49098107)
449
+ 3600 val loss 3.9321
450
+ 3600 val perplexity 51.0136
451
+ 3600 train 3.938220 (lr=6.8583e-04) (hash(x)=41194370)
452
+ 3610 train 4.055422 (lr=6.7753e-04) (hash(x)=40399152)
453
+ 3620 train 4.017025 (lr=6.6932e-04) (hash(x)=41625018)
454
+ 3630 train 4.039500 (lr=6.6121e-04) (hash(x)=44077942)
455
+ 3640 train 4.036861 (lr=6.5319e-04) (hash(x)=42427512)
456
+ 3650 train 4.017832 (lr=6.4526e-04) (hash(x)=40067454)
457
+ 3660 train 3.943225 (lr=6.3743e-04) (hash(x)=37969892)
458
+ 3670 train 3.916439 (lr=6.2969e-04) (hash(x)=43670246)
459
+ 3680 train 3.994433 (lr=6.2204e-04) (hash(x)=37031732)
460
+ 3690 train 3.869253 (lr=6.1450e-04) (hash(x)=39237187)
461
+ 3700 val loss 3.9265
462
+ 3700 val perplexity 50.7266
463
+ 3700 train 3.911662 (lr=6.0705e-04) (hash(x)=39521416)
464
+ 3710 train 3.807154 (lr=5.9969e-04) (hash(x)=41147425)
465
+ 3720 train 3.799486 (lr=5.9244e-04) (hash(x)=40204658)
466
+ 3730 train 3.894923 (lr=5.8528e-04) (hash(x)=39420983)
467
+ 3740 train 3.942589 (lr=5.7822e-04) (hash(x)=51155740)
468
+ 3750 hella 0.2560
469
+ 3750 train 3.926321 (lr=5.7126e-04) (hash(x)=41654187)
470
+ 3760 train 3.884883 (lr=5.6440e-04) (hash(x)=40598339)
471
+ 3770 train 3.849714 (lr=5.5764e-04) (hash(x)=44950116)
472
+ 3780 train 3.911105 (lr=5.5098e-04) (hash(x)=40660326)
473
+ 3790 train 3.956152 (lr=5.4443e-04) (hash(x)=37805851)
474
+ 3800 val loss 3.9090
475
+ 3800 val perplexity 49.8471
476
+ 3800 train 3.925261 (lr=5.3797e-04) (hash(x)=38064443)
477
+ 3810 train 3.996564 (lr=5.3162e-04) (hash(x)=39161498)
478
+ 3820 train 3.846784 (lr=5.2537e-04) (hash(x)=40626649)
479
+ 3830 train 3.915901 (lr=5.1922e-04) (hash(x)=36894771)
480
+ 3840 train 3.859013 (lr=5.1317e-04) (hash(x)=33672141)
481
+ 3850 train 3.820592 (lr=5.0723e-04) (hash(x)=41751813)
482
+ 3860 train 3.694681 (lr=5.0140e-04) (hash(x)=33784172)
483
+ 3870 train 3.619635 (lr=4.9567e-04) (hash(x)=43110493)
484
+ 3880 train 3.719017 (lr=4.9004e-04) (hash(x)=36585111)
485
+ 3890 train 3.726070 (lr=4.8452e-04) (hash(x)=40134264)
486
+ 3900 val loss 3.9159
487
+ 3900 val perplexity 50.1961
488
+ 3900 train 3.718529 (lr=4.7911e-04) (hash(x)=39815215)
489
+ 3910 train 3.859548 (lr=4.7380e-04) (hash(x)=42081557)
490
+ 3920 train 3.794125 (lr=4.6860e-04) (hash(x)=48184973)
491
+ 3930 train 3.991539 (lr=4.6351e-04) (hash(x)=42416681)
492
+ 3940 train 3.913430 (lr=4.5852e-04) (hash(x)=38587062)
493
+ 3950 train 3.910361 (lr=4.5364e-04) (hash(x)=37296617)
494
+ 3960 train 3.888490 (lr=4.4888e-04) (hash(x)=37959330)
495
+ 3970 train 3.914019 (lr=4.4422e-04) (hash(x)=35175634)
496
+ 3980 train 3.963458 (lr=4.3966e-04) (hash(x)=37219362)
497
+ 3990 train 3.983428 (lr=4.3522e-04) (hash(x)=38906132)
498
+ 4000 val loss 3.8941
499
+ 4000 val perplexity 49.1117
500
+ 4000 hella 0.2567
501
+ 4000 train 3.908096 (lr=4.3089e-04) (hash(x)=39940517)
502
+ 4010 train 3.915962 (lr=4.2667e-04) (hash(x)=39952437)
503
+ 4020 train 3.832983 (lr=4.2256e-04) (hash(x)=42000930)
504
+ 4030 train 3.849427 (lr=4.1856e-04) (hash(x)=41460672)
505
+ 4040 train 3.592105 (lr=4.1467e-04) (hash(x)=43702329)
506
+ 4050 train 3.689124 (lr=4.1089e-04) (hash(x)=41740121)
507
+ 4060 train 3.684111 (lr=4.0722e-04) (hash(x)=37066349)
508
+ 4070 train 3.526400 (lr=4.0367e-04) (hash(x)=38359869)
509
+ 4080 train 3.773262 (lr=4.0022e-04) (hash(x)=35131370)
510
+ 4090 train 3.832202 (lr=3.9689e-04) (hash(x)=39918811)
511
+ 4100 val loss 3.8927
512
+ 4100 val perplexity 49.0439
513
+ 4100 train 3.989994 (lr=3.9368e-04) (hash(x)=47036374)
514
+ 4110 train 3.915630 (lr=3.9057e-04) (hash(x)=37657523)
515
+ 4120 train 3.960364 (lr=3.8758e-04) (hash(x)=37218703)
516
+ 4130 train 3.971655 (lr=3.8470e-04) (hash(x)=41350513)
517
+ 4140 train 3.996273 (lr=3.8193e-04) (hash(x)=38283785)
518
+ 4150 train 3.978489 (lr=3.7928e-04) (hash(x)=32927892)
519
+ 4160 train 3.857933 (lr=3.7674e-04) (hash(x)=42011933)
520
+ 4170 train 3.881077 (lr=3.7432e-04) (hash(x)=41854594)
521
+ 4180 train 3.886825 (lr=3.7201e-04) (hash(x)=36737064)
522
+ 4190 train 3.800586 (lr=3.6982e-04) (hash(x)=36326176)
523
+ 4200 val loss 3.8849
524
+ 4200 val perplexity 48.6614
525
+ 4200 train 3.870943 (lr=3.6774e-04) (hash(x)=39106683)
526
+ 4210 train 3.921710 (lr=3.6577e-04) (hash(x)=39759977)
527
+ 4220 train 3.663461 (lr=3.6392e-04) (hash(x)=43666151)
528
+ 4230 train 3.610104 (lr=3.6218e-04) (hash(x)=38154367)
529
+ 4240 train 3.722485 (lr=3.6056e-04) (hash(x)=38284168)
530
+ 4250 hella 0.2564
531
+ 4250 train 3.654472 (lr=3.5906e-04) (hash(x)=38888879)
532
+ 4260 train 3.569697 (lr=3.5767e-04) (hash(x)=36846417)
533
+ 4270 train 3.979475 (lr=3.5639e-04) (hash(x)=43125375)
534
+ 4280 train 3.932252 (lr=3.5523e-04) (hash(x)=43556412)
535
+ 4290 train 3.864168 (lr=3.5419e-04) (hash(x)=48598302)
536
+ 4300 val loss 3.8794
537
+ 4300 val perplexity 48.3958
538
+ 4300 train 3.843756 (lr=3.5326e-04) (hash(x)=42764385)
539
+ 4310 train 3.901860 (lr=3.5245e-04) (hash(x)=41190028)
540
+ 4320 train 3.868364 (lr=3.5175e-04) (hash(x)=35767770)
541
+ 4330 train 3.944481 (lr=3.5117e-04) (hash(x)=36760476)
542
+ 4340 train 3.884115 (lr=3.5071e-04) (hash(x)=40618921)
543
+ 4350 train 3.988392 (lr=3.5036e-04) (hash(x)=32300164)
544
+ 4360 train 3.844056 (lr=3.5013e-04) (hash(x)=41400085)
545
+ 4370 train 3.880195 (lr=3.5001e-04) (hash(x)=40251828)
546
+ 4374 val loss 3.8796
547
+ 4374 val perplexity 48.4068
548
+ 4374 hella 0.2544
logs/fix_1_latent_mask/1_latent_mask_lr_35e-4_n_latent_masks_2_no_compile_seed_1340/model_04374.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:7a42764165513d110d8eb8c2a47874b1d7ccd0e91e81464c518755b8c9f7a48f
3
+ size 97707314
logs/fix_1_latent_mask/1_latent_mask_lr_35e-4_n_latent_masks_2_no_compile_seed_1340/optimizer_04374.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:9e41a3495de7956e4a342e578b2c9a43e732f5843b4f13363e3e6ae49a1562bd
3
+ size 189136950