andrew-healey commited on
Commit
5dbf0f2
·
verified ·
1 Parent(s): dbb4d87

Upload folder using huggingface_hub

Browse files
12_mini_head_one_mask_per_head_1_latent_mask_halved_lr/args.json ADDED
@@ -0,0 +1 @@
 
 
1
+ {"hellaswag": true, "attention_kind": "selective", "log_dir": "logs/repro_selective_pattern_rankings/12_mini_head_one_mask_per_head_1_latent_mask_halved_lr", "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": "repro_selective_pattern_rankings", "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, "protection_kind": "none", "leaky_relu_alpha": null, "leaky_relu_bias": null, "use_compile": true, "use_mini_model": false, "upload_to_hf": true, "seq_len": 256, "batch_size": 64, "total_batch_size": 131072, "protection_head_scaling_factor": 1.0, "protection_head_bias": 0.0, "n_sliced_masks": null, "n_latent_masks": 1, "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": 0.0015, "decay_lr": true, "readout_zero_init": false, "query_zero_init": false, "l1_loss": false, "debugpy": false, "key": "12_mini_head_one_mask_per_head_1_latent_mask_halved_lr"}
12_mini_head_one_mask_per_head_1_latent_mask_halved_lr/dataloader_04374.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:8ec6682852ed60a9d42cb4047300f65a04f87328c0ad7a4516be84d11b28f216
3
+ size 964
12_mini_head_one_mask_per_head_1_latent_mask_halved_lr/log2.txt ADDED
@@ -0,0 +1,529 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ max_steps: 4375
2
+ 0 val loss 11.2876
3
+ 0 val perplexity 79825.2969
4
+ 0 train 11.284595 (lr=2.0979e-06) (hash(x)=93356070)
5
+ 10 train 10.121101 (lr=2.3077e-05) (hash(x)=91253010)
6
+ 20 train 9.687342 (lr=4.4056e-05) (hash(x)=74934453)
7
+ 30 train 9.292213 (lr=6.5035e-05) (hash(x)=79278034)
8
+ 40 train 8.717637 (lr=8.6014e-05) (hash(x)=80107892)
9
+ 50 train 8.194167 (lr=1.0699e-04) (hash(x)=70745428)
10
+ 60 train 7.915437 (lr=1.2797e-04) (hash(x)=80077589)
11
+ 70 train 7.560098 (lr=1.4895e-04) (hash(x)=76213766)
12
+ 80 train 7.591424 (lr=1.6993e-04) (hash(x)=83218328)
13
+ 90 train 7.440104 (lr=1.9091e-04) (hash(x)=74855845)
14
+ 100 val loss 7.3731
15
+ 100 val perplexity 1592.4937
16
+ 100 train 7.218736 (lr=2.1189e-04) (hash(x)=82814902)
17
+ 110 train 7.161034 (lr=2.3287e-04) (hash(x)=84907741)
18
+ 120 train 7.028426 (lr=2.5385e-04) (hash(x)=82613223)
19
+ 130 train 6.889417 (lr=2.7483e-04) (hash(x)=83540876)
20
+ 140 train 6.844666 (lr=2.9580e-04) (hash(x)=75095216)
21
+ 150 train 6.895470 (lr=3.1678e-04) (hash(x)=97190944)
22
+ 160 train 6.798061 (lr=3.3776e-04) (hash(x)=82117809)
23
+ 170 train 6.698723 (lr=3.5874e-04) (hash(x)=70514724)
24
+ 180 train 6.618811 (lr=3.7972e-04) (hash(x)=81029624)
25
+ 190 train 6.541725 (lr=4.0070e-04) (hash(x)=78996841)
26
+ 200 val loss 6.4768
27
+ 200 val perplexity 649.9196
28
+ 200 train 6.484604 (lr=4.2168e-04) (hash(x)=79845097)
29
+ 210 train 6.358215 (lr=4.4266e-04) (hash(x)=78997683)
30
+ 220 train 6.273339 (lr=4.6364e-04) (hash(x)=74895865)
31
+ 230 train 6.443975 (lr=4.8462e-04) (hash(x)=80933276)
32
+ 240 train 6.127409 (lr=5.0559e-04) (hash(x)=77664606)
33
+ 250 train 6.166188 (lr=5.2657e-04) (hash(x)=75903930)
34
+ 260 train 6.074569 (lr=5.4755e-04) (hash(x)=82996853)
35
+ 270 train 6.006226 (lr=5.6853e-04) (hash(x)=73269568)
36
+ 280 train 5.942740 (lr=5.8951e-04) (hash(x)=82906427)
37
+ 290 train 5.863214 (lr=6.1049e-04) (hash(x)=71797895)
38
+ 300 val loss 5.9985
39
+ 300 val perplexity 402.8120
40
+ 300 train 5.720431 (lr=6.3147e-04) (hash(x)=85232249)
41
+ 310 train 5.674852 (lr=6.5245e-04) (hash(x)=75252489)
42
+ 320 train 5.750916 (lr=6.7343e-04) (hash(x)=74394644)
43
+ 330 train 5.840282 (lr=6.9441e-04) (hash(x)=82772910)
44
+ 340 train 5.859873 (lr=7.1538e-04) (hash(x)=81627464)
45
+ 350 train 5.852388 (lr=7.3636e-04) (hash(x)=84678053)
46
+ 360 train 5.801634 (lr=7.5734e-04) (hash(x)=81884128)
47
+ 370 train 5.782537 (lr=7.7832e-04) (hash(x)=76379242)
48
+ 380 train 5.580529 (lr=7.9930e-04) (hash(x)=98182875)
49
+ 390 train 5.565805 (lr=8.2028e-04) (hash(x)=79710436)
50
+ 400 val loss 5.6257
51
+ 400 val perplexity 277.4746
52
+ 400 train 5.573838 (lr=8.4126e-04) (hash(x)=79841071)
53
+ 410 train 5.505904 (lr=8.6224e-04) (hash(x)=75844151)
54
+ 420 train 5.500185 (lr=8.8322e-04) (hash(x)=73125036)
55
+ 430 train 5.544938 (lr=9.0420e-04) (hash(x)=84214858)
56
+ 440 train 5.430576 (lr=9.2517e-04) (hash(x)=80456994)
57
+ 450 train 5.402608 (lr=9.4615e-04) (hash(x)=76735962)
58
+ 460 train 5.328233 (lr=9.6713e-04) (hash(x)=81845446)
59
+ 470 train 5.165900 (lr=9.8811e-04) (hash(x)=76094689)
60
+ 480 train 5.273071 (lr=1.0091e-03) (hash(x)=83806686)
61
+ 490 train 5.097137 (lr=1.0301e-03) (hash(x)=84690227)
62
+ 500 val loss 5.4281
63
+ 500 val perplexity 227.7151
64
+ 500 train 5.176426 (lr=1.0510e-03) (hash(x)=71851938)
65
+ 510 train 5.261232 (lr=1.0720e-03) (hash(x)=77159346)
66
+ 520 train 5.328233 (lr=1.0930e-03) (hash(x)=80755753)
67
+ 530 train 5.422802 (lr=1.1140e-03) (hash(x)=75379679)
68
+ 540 train 5.334445 (lr=1.1350e-03) (hash(x)=82458619)
69
+ 550 train 5.315642 (lr=1.1559e-03) (hash(x)=86366396)
70
+ 560 train 5.314054 (lr=1.1769e-03) (hash(x)=85095044)
71
+ 570 train 5.294374 (lr=1.1979e-03) (hash(x)=78385159)
72
+ 580 train 5.227594 (lr=1.2189e-03) (hash(x)=79342394)
73
+ 590 train 5.208125 (lr=1.2399e-03) (hash(x)=70782192)
74
+ 600 val loss 5.2270
75
+ 600 val perplexity 186.2377
76
+ 600 train 5.206278 (lr=1.2608e-03) (hash(x)=76130353)
77
+ 610 train 5.133800 (lr=1.2818e-03) (hash(x)=74778440)
78
+ 620 train 5.108348 (lr=1.3028e-03) (hash(x)=79129709)
79
+ 630 train 5.011270 (lr=1.3238e-03) (hash(x)=76469962)
80
+ 640 train 5.014485 (lr=1.3448e-03) (hash(x)=78288049)
81
+ 650 train 4.972734 (lr=1.3657e-03) (hash(x)=76641388)
82
+ 660 train 4.898829 (lr=1.3867e-03) (hash(x)=75937906)
83
+ 670 train 4.890547 (lr=1.4077e-03) (hash(x)=74807157)
84
+ 680 train 4.800326 (lr=1.4287e-03) (hash(x)=77490144)
85
+ 690 train 4.824878 (lr=1.4497e-03) (hash(x)=72900124)
86
+ 700 val loss 5.1115
87
+ 700 val perplexity 165.9268
88
+ 700 train 4.798738 (lr=1.4706e-03) (hash(x)=73424218)
89
+ 710 train 5.078708 (lr=1.4916e-03) (hash(x)=86052345)
90
+ 720 train 5.056393 (lr=1.5000e-03) (hash(x)=77462613)
91
+ 730 train 5.066881 (lr=1.4999e-03) (hash(x)=79027471)
92
+ 740 train 5.025856 (lr=1.4998e-03) (hash(x)=78149992)
93
+ 750 train 5.001462 (lr=1.4997e-03) (hash(x)=89147499)
94
+ 760 train 4.996930 (lr=1.4995e-03) (hash(x)=74931538)
95
+ 770 train 4.953558 (lr=1.4992e-03) (hash(x)=89648204)
96
+ 780 train 4.984262 (lr=1.4989e-03) (hash(x)=83708293)
97
+ 790 train 4.998816 (lr=1.4986e-03) (hash(x)=84336512)
98
+ 800 val loss 4.9686
99
+ 800 val perplexity 143.8217
100
+ 800 train 4.890866 (lr=1.4982e-03) (hash(x)=75025285)
101
+ 810 train 4.809444 (lr=1.4978e-03) (hash(x)=79436536)
102
+ 820 train 4.792324 (lr=1.4973e-03) (hash(x)=90467390)
103
+ 830 train 4.848436 (lr=1.4967e-03) (hash(x)=77292020)
104
+ 840 train 4.755088 (lr=1.4961e-03) (hash(x)=75568927)
105
+ 850 train 4.803519 (lr=1.4955e-03) (hash(x)=79671781)
106
+ 860 train 4.873863 (lr=1.4948e-03) (hash(x)=84280943)
107
+ 870 train 4.850063 (lr=1.4940e-03) (hash(x)=80901294)
108
+ 880 train 4.977321 (lr=1.4932e-03) (hash(x)=81437584)
109
+ 890 train 4.807866 (lr=1.4924e-03) (hash(x)=73893778)
110
+ 900 val loss 4.8264
111
+ 900 val perplexity 124.7557
112
+ 900 train 4.753095 (lr=1.4915e-03) (hash(x)=74987794)
113
+ 910 train 4.762667 (lr=1.4906e-03) (hash(x)=93721374)
114
+ 920 train 4.753925 (lr=1.4896e-03) (hash(x)=79149678)
115
+ 930 train 4.702624 (lr=1.4885e-03) (hash(x)=83179387)
116
+ 940 train 4.700030 (lr=1.4875e-03) (hash(x)=89009978)
117
+ 950 train 4.686575 (lr=1.4863e-03) (hash(x)=76483735)
118
+ 960 train 4.620933 (lr=1.4851e-03) (hash(x)=78638850)
119
+ 970 train 4.642137 (lr=1.4839e-03) (hash(x)=90415206)
120
+ 980 train 4.579130 (lr=1.4826e-03) (hash(x)=90819195)
121
+ 990 train 4.710921 (lr=1.4813e-03) (hash(x)=85938852)
122
+ 1000 val loss 4.6934
123
+ 1000 val perplexity 109.2292
124
+ 1000 train 4.742970 (lr=1.4799e-03) (hash(x)=82436789)
125
+ 1010 train 4.784894 (lr=1.4785e-03) (hash(x)=63022149)
126
+ 1020 train 4.706555 (lr=1.4770e-03) (hash(x)=83501199)
127
+ 1030 train 4.674661 (lr=1.4755e-03) (hash(x)=77093326)
128
+ 1040 train 4.559695 (lr=1.4739e-03) (hash(x)=73437559)
129
+ 1050 train 4.651922 (lr=1.4723e-03) (hash(x)=84550388)
130
+ 1060 train 4.646999 (lr=1.4706e-03) (hash(x)=86184566)
131
+ 1070 train 4.635979 (lr=1.4689e-03) (hash(x)=85644922)
132
+ 1080 train 4.575619 (lr=1.4671e-03) (hash(x)=86179801)
133
+ 1090 train 4.543534 (lr=1.4653e-03) (hash(x)=82868303)
134
+ 1100 val loss 4.6156
135
+ 1100 val perplexity 101.0505
136
+ 1100 train 4.645684 (lr=1.4635e-03) (hash(x)=96780388)
137
+ 1110 train 4.575282 (lr=1.4616e-03) (hash(x)=87223122)
138
+ 1120 train 4.546115 (lr=1.4596e-03) (hash(x)=77292786)
139
+ 1130 train 4.574295 (lr=1.4576e-03) (hash(x)=88761375)
140
+ 1140 train 4.480355 (lr=1.4556e-03) (hash(x)=80956468)
141
+ 1150 train 4.656193 (lr=1.4535e-03) (hash(x)=84725462)
142
+ 1160 train 4.627910 (lr=1.4514e-03) (hash(x)=76098113)
143
+ 1170 train 4.612670 (lr=1.4492e-03) (hash(x)=84387685)
144
+ 1180 train 4.553823 (lr=1.4469e-03) (hash(x)=76120321)
145
+ 1190 train 4.521862 (lr=1.4447e-03) (hash(x)=85974065)
146
+ 1200 val loss 4.5357
147
+ 1200 val perplexity 93.2844
148
+ 1200 train 4.418312 (lr=1.4423e-03) (hash(x)=79340644)
149
+ 1210 train 4.515040 (lr=1.4400e-03) (hash(x)=88345025)
150
+ 1220 train 4.427493 (lr=1.4376e-03) (hash(x)=86406230)
151
+ 1230 train 4.413098 (lr=1.4351e-03) (hash(x)=80608975)
152
+ 1240 train 4.462535 (lr=1.4326e-03) (hash(x)=81635225)
153
+ 1250 train 4.305313 (lr=1.4301e-03) (hash(x)=82126376)
154
+ 1260 train 4.476536 (lr=1.4275e-03) (hash(x)=83737972)
155
+ 1270 train 4.389083 (lr=1.4248e-03) (hash(x)=81132345)
156
+ 1280 train 4.478844 (lr=1.4222e-03) (hash(x)=85850404)
157
+ 1290 train 4.490558 (lr=1.4194e-03) (hash(x)=83490640)
158
+ 1300 val loss 4.4923
159
+ 1300 val perplexity 89.3224
160
+ 1300 train 4.462030 (lr=1.4167e-03) (hash(x)=77545187)
161
+ 1310 train 4.518206 (lr=1.4139e-03) (hash(x)=86412685)
162
+ 1320 train 4.399871 (lr=1.4110e-03) (hash(x)=86429640)
163
+ 1330 train 4.538294 (lr=1.4081e-03) (hash(x)=79645281)
164
+ 1340 train 4.497840 (lr=1.4052e-03) (hash(x)=82668541)
165
+ 1350 train 4.350512 (lr=1.4022e-03) (hash(x)=73627248)
166
+ 1360 train 4.449611 (lr=1.3992e-03) (hash(x)=74845133)
167
+ 1370 train 4.363082 (lr=1.3961e-03) (hash(x)=77850497)
168
+ 1380 train 4.385639 (lr=1.3930e-03) (hash(x)=76143954)
169
+ 1390 train 4.399740 (lr=1.3898e-03) (hash(x)=80675544)
170
+ 1400 val loss 4.4650
171
+ 1400 val perplexity 86.9220
172
+ 1400 train 4.373606 (lr=1.3866e-03) (hash(x)=76010938)
173
+ 1410 train 4.426188 (lr=1.3834e-03) (hash(x)=81260300)
174
+ 1420 train 4.532896 (lr=1.3801e-03) (hash(x)=83803629)
175
+ 1430 train 4.365252 (lr=1.3768e-03) (hash(x)=83840110)
176
+ 1440 train 4.559702 (lr=1.3735e-03) (hash(x)=97711831)
177
+ 1450 train 4.434718 (lr=1.3701e-03) (hash(x)=81110139)
178
+ 1460 train 4.354506 (lr=1.3666e-03) (hash(x)=90942839)
179
+ 1470 train 4.395035 (lr=1.3631e-03) (hash(x)=80721564)
180
+ 1480 train 4.485539 (lr=1.3596e-03) (hash(x)=77852059)
181
+ 1490 train 4.379970 (lr=1.3561e-03) (hash(x)=80011365)
182
+ 1500 val loss 4.4194
183
+ 1500 val perplexity 83.0453
184
+ 1500 train 4.403055 (lr=1.3525e-03) (hash(x)=84496142)
185
+ 1510 train 4.347849 (lr=1.3488e-03) (hash(x)=70186729)
186
+ 1520 train 4.368427 (lr=1.3452e-03) (hash(x)=74854227)
187
+ 1530 train 4.274542 (lr=1.3415e-03) (hash(x)=77468161)
188
+ 1540 train 4.292623 (lr=1.3377e-03) (hash(x)=87853059)
189
+ 1550 train 4.240846 (lr=1.3339e-03) (hash(x)=71225436)
190
+ 1560 train 4.329825 (lr=1.3301e-03) (hash(x)=84072783)
191
+ 1570 train 4.387258 (lr=1.3262e-03) (hash(x)=72723098)
192
+ 1580 train 4.395650 (lr=1.3223e-03) (hash(x)=90409866)
193
+ 1590 train 4.469543 (lr=1.3184e-03) (hash(x)=87481378)
194
+ 1600 val loss 4.3891
195
+ 1600 val perplexity 80.5672
196
+ 1600 train 4.442002 (lr=1.3144e-03) (hash(x)=77643862)
197
+ 1610 train 4.343353 (lr=1.3104e-03) (hash(x)=88862575)
198
+ 1620 train 4.344193 (lr=1.3064e-03) (hash(x)=84612581)
199
+ 1630 train 4.383558 (lr=1.3023e-03) (hash(x)=87075989)
200
+ 1640 train 4.271748 (lr=1.2982e-03) (hash(x)=88277361)
201
+ 1650 train 4.386574 (lr=1.2940e-03) (hash(x)=78750236)
202
+ 1660 train 4.313765 (lr=1.2898e-03) (hash(x)=82604581)
203
+ 1670 train 4.151987 (lr=1.2856e-03) (hash(x)=68482265)
204
+ 1680 train 4.273175 (lr=1.2814e-03) (hash(x)=75088835)
205
+ 1690 train 4.215827 (lr=1.2771e-03) (hash(x)=66060989)
206
+ 1700 val loss 4.3945
207
+ 1700 val perplexity 81.0013
208
+ 1700 train 4.390934 (lr=1.2728e-03) (hash(x)=79986754)
209
+ 1710 train 4.368111 (lr=1.2684e-03) (hash(x)=83657930)
210
+ 1720 train 4.318359 (lr=1.2640e-03) (hash(x)=81754135)
211
+ 1730 train 4.396428 (lr=1.2596e-03) (hash(x)=78904427)
212
+ 1740 train 4.304997 (lr=1.2552e-03) (hash(x)=85920177)
213
+ 1750 train 4.352421 (lr=1.2507e-03) (hash(x)=86573211)
214
+ 1760 train 4.306359 (lr=1.2462e-03) (hash(x)=81737128)
215
+ 1770 train 4.302860 (lr=1.2416e-03) (hash(x)=77400968)
216
+ 1780 train 4.307379 (lr=1.2371e-03) (hash(x)=73545497)
217
+ 1790 train 4.557861 (lr=1.2325e-03) (hash(x)=71641943)
218
+ 1800 val loss 4.3455
219
+ 1800 val perplexity 77.1299
220
+ 1800 train 4.404317 (lr=1.2278e-03) (hash(x)=87819781)
221
+ 1810 train 4.263799 (lr=1.2232e-03) (hash(x)=86870770)
222
+ 1820 train 4.275720 (lr=1.2185e-03) (hash(x)=82522211)
223
+ 1830 train 4.353501 (lr=1.2137e-03) (hash(x)=61947437)
224
+ 1840 train 4.237915 (lr=1.2090e-03) (hash(x)=79865406)
225
+ 1850 train 4.212577 (lr=1.2042e-03) (hash(x)=79828721)
226
+ 1860 train 4.354757 (lr=1.1994e-03) (hash(x)=80869571)
227
+ 1870 train 4.334395 (lr=1.1946e-03) (hash(x)=73780971)
228
+ 1880 train 4.301668 (lr=1.1897e-03) (hash(x)=79249549)
229
+ 1890 train 4.352936 (lr=1.1848e-03) (hash(x)=81041904)
230
+ 1900 val loss 4.3246
231
+ 1900 val perplexity 75.5348
232
+ 1900 train 4.416948 (lr=1.1799e-03) (hash(x)=82456430)
233
+ 1910 train 4.185815 (lr=1.1750e-03) (hash(x)=82222135)
234
+ 1920 train 4.227049 (lr=1.1700e-03) (hash(x)=70033249)
235
+ 1930 train 4.209605 (lr=1.1650e-03) (hash(x)=72887360)
236
+ 1940 train 4.258805 (lr=1.1600e-03) (hash(x)=83251100)
237
+ 1950 train 4.300284 (lr=1.1549e-03) (hash(x)=79660266)
238
+ 1960 train 4.181851 (lr=1.1499e-03) (hash(x)=78919068)
239
+ 1970 train 4.274166 (lr=1.1448e-03) (hash(x)=94690431)
240
+ 1980 train 4.113525 (lr=1.1397e-03) (hash(x)=88153756)
241
+ 1990 train 4.167477 (lr=1.1345e-03) (hash(x)=77195688)
242
+ 2000 val loss 4.3038
243
+ 2000 val perplexity 73.9823
244
+ 2000 train 4.272132 (lr=1.1294e-03) (hash(x)=81308591)
245
+ 2010 train 4.315694 (lr=1.1242e-03) (hash(x)=80205479)
246
+ 2020 train 4.232471 (lr=1.1190e-03) (hash(x)=81008704)
247
+ 2030 train 4.298613 (lr=1.1138e-03) (hash(x)=82914358)
248
+ 2040 train 4.306230 (lr=1.1085e-03) (hash(x)=88064399)
249
+ 2050 train 4.311451 (lr=1.1032e-03) (hash(x)=82889390)
250
+ 2060 train 4.387699 (lr=1.0980e-03) (hash(x)=80913554)
251
+ 2070 train 4.232565 (lr=1.0926e-03) (hash(x)=78121791)
252
+ 2080 train 4.361650 (lr=1.0873e-03) (hash(x)=75356657)
253
+ 2090 train 4.185152 (lr=1.0820e-03) (hash(x)=74048412)
254
+ 2100 val loss 4.2817
255
+ 2100 val perplexity 72.3637
256
+ 2100 train 4.107095 (lr=1.0766e-03) (hash(x)=68928225)
257
+ 2110 train 4.157742 (lr=1.0712e-03) (hash(x)=93734745)
258
+ 2120 train 4.150732 (lr=1.0658e-03) (hash(x)=72878164)
259
+ 2130 train 4.280064 (lr=1.0604e-03) (hash(x)=71508226)
260
+ 2140 train 4.242207 (lr=1.0550e-03) (hash(x)=84168671)
261
+ 2150 train 4.179171 (lr=1.0495e-03) (hash(x)=74673639)
262
+ 2160 train 4.325368 (lr=1.0440e-03) (hash(x)=75470031)
263
+ 2170 train 4.284217 (lr=1.0385e-03) (hash(x)=74307890)
264
+ 2180 train 4.391325 (lr=1.0330e-03) (hash(x)=77214245)
265
+ 2190 train 4.178585 (lr=1.0275e-03) (hash(x)=88628359)
266
+ 2200 val loss 4.2648
267
+ 2200 val perplexity 71.1485
268
+ 2200 train 4.202308 (lr=1.0220e-03) (hash(x)=74779126)
269
+ 2210 train 4.232403 (lr=1.0164e-03) (hash(x)=79817976)
270
+ 2220 train 4.188246 (lr=1.0109e-03) (hash(x)=75258996)
271
+ 2230 train 4.161020 (lr=1.0053e-03) (hash(x)=88164047)
272
+ 2240 train 4.143344 (lr=9.9970e-04) (hash(x)=78580686)
273
+ 2250 train 4.085512 (lr=9.9410e-04) (hash(x)=76510617)
274
+ 2260 train 4.165926 (lr=9.8848e-04) (hash(x)=76043966)
275
+ 2270 train 4.139482 (lr=9.8286e-04) (hash(x)=73062098)
276
+ 2280 train 4.368248 (lr=9.7722e-04) (hash(x)=81885909)
277
+ 2290 train 4.295055 (lr=9.7157e-04) (hash(x)=81481741)
278
+ 2300 val loss 4.2408
279
+ 2300 val perplexity 69.4627
280
+ 2300 train 4.160977 (lr=9.6591e-04) (hash(x)=82104275)
281
+ 2310 train 4.210849 (lr=9.6024e-04) (hash(x)=82137309)
282
+ 2320 train 4.193648 (lr=9.5455e-04) (hash(x)=86741167)
283
+ 2330 train 4.210475 (lr=9.4886e-04) (hash(x)=84796263)
284
+ 2340 train 4.172543 (lr=9.4316e-04) (hash(x)=169966529)
285
+ 2350 train 4.050961 (lr=9.3745e-04) (hash(x)=80006061)
286
+ 2360 train 4.115921 (lr=9.3174e-04) (hash(x)=71090999)
287
+ 2370 train 4.240656 (lr=9.2601e-04) (hash(x)=86641850)
288
+ 2380 train 4.138200 (lr=9.2028e-04) (hash(x)=84573253)
289
+ 2390 train 4.168169 (lr=9.1454e-04) (hash(x)=74483764)
290
+ 2400 val loss 4.2350
291
+ 2400 val perplexity 69.0619
292
+ 2400 train 4.143218 (lr=9.0880e-04) (hash(x)=78327659)
293
+ 2410 train 4.209142 (lr=9.0304e-04) (hash(x)=77273627)
294
+ 2420 train 4.069314 (lr=8.9729e-04) (hash(x)=76938049)
295
+ 2430 train 4.082942 (lr=8.9152e-04) (hash(x)=98449442)
296
+ 2440 train 3.937424 (lr=8.8575e-04) (hash(x)=78157797)
297
+ 2450 train 3.980594 (lr=8.7998e-04) (hash(x)=80637582)
298
+ 2460 train 3.977829 (lr=8.7420e-04) (hash(x)=80023854)
299
+ 2470 train 4.033025 (lr=8.6842e-04) (hash(x)=86678884)
300
+ 2480 train 4.256807 (lr=8.6264e-04) (hash(x)=84086469)
301
+ 2490 train 4.271472 (lr=8.5685e-04) (hash(x)=83131141)
302
+ 2500 val loss 4.2121
303
+ 2500 val perplexity 67.4955
304
+ 2500 train 4.196254 (lr=8.5107e-04) (hash(x)=82583497)
305
+ 2510 train 4.249238 (lr=8.4528e-04) (hash(x)=79164326)
306
+ 2520 train 4.160680 (lr=8.3948e-04) (hash(x)=76456503)
307
+ 2530 train 4.215426 (lr=8.3369e-04) (hash(x)=82001228)
308
+ 2540 train 4.253792 (lr=8.2790e-04) (hash(x)=75615595)
309
+ 2550 train 4.201610 (lr=8.2210e-04) (hash(x)=78397869)
310
+ 2560 train 4.205235 (lr=8.1631e-04) (hash(x)=75547032)
311
+ 2570 train 4.110967 (lr=8.1052e-04) (hash(x)=89201025)
312
+ 2580 train 4.101588 (lr=8.0472e-04) (hash(x)=89856704)
313
+ 2590 train 4.104813 (lr=7.9893e-04) (hash(x)=82175682)
314
+ 2600 val loss 4.1966
315
+ 2600 val perplexity 66.4626
316
+ 2600 train 4.119221 (lr=7.9315e-04) (hash(x)=78312826)
317
+ 2610 train 4.105620 (lr=7.8736e-04) (hash(x)=77066588)
318
+ 2620 train 3.948243 (lr=7.8158e-04) (hash(x)=78666061)
319
+ 2630 train 3.947694 (lr=7.7580e-04) (hash(x)=93762143)
320
+ 2640 train 3.970228 (lr=7.7002e-04) (hash(x)=83191587)
321
+ 2650 train 3.884363 (lr=7.6425e-04) (hash(x)=87169585)
322
+ 2660 train 4.106050 (lr=7.5848e-04) (hash(x)=86426388)
323
+ 2670 train 4.241434 (lr=7.5271e-04) (hash(x)=76692638)
324
+ 2680 train 4.161576 (lr=7.4696e-04) (hash(x)=77446063)
325
+ 2690 train 4.161252 (lr=7.4120e-04) (hash(x)=79809050)
326
+ 2700 val loss 4.1928
327
+ 2700 val perplexity 66.2078
328
+ 2700 train 4.088799 (lr=7.3546e-04) (hash(x)=83116823)
329
+ 2710 train 4.201213 (lr=7.2972e-04) (hash(x)=75622148)
330
+ 2720 train 4.128744 (lr=7.2399e-04) (hash(x)=80690512)
331
+ 2730 train 4.198697 (lr=7.1826e-04) (hash(x)=78009984)
332
+ 2740 train 4.151336 (lr=7.1255e-04) (hash(x)=74205488)
333
+ 2750 train 4.153377 (lr=7.0684e-04) (hash(x)=91013332)
334
+ 2760 train 4.079216 (lr=7.0114e-04) (hash(x)=76954961)
335
+ 2770 train 4.079950 (lr=6.9545e-04) (hash(x)=79390317)
336
+ 2780 train 3.949634 (lr=6.8976e-04) (hash(x)=70168783)
337
+ 2790 train 4.113686 (lr=6.8409e-04) (hash(x)=76028417)
338
+ 2800 val loss 4.1722
339
+ 2800 val perplexity 64.8558
340
+ 2800 train 3.942339 (lr=6.7843e-04) (hash(x)=77656050)
341
+ 2810 train 3.991017 (lr=6.7278e-04) (hash(x)=94248216)
342
+ 2820 train 4.055126 (lr=6.6714e-04) (hash(x)=78305078)
343
+ 2830 train 3.969747 (lr=6.6152e-04) (hash(x)=79948848)
344
+ 2840 train 3.841772 (lr=6.5590e-04) (hash(x)=85341024)
345
+ 2850 train 4.145309 (lr=6.5030e-04) (hash(x)=78735170)
346
+ 2860 train 4.390820 (lr=6.4471e-04) (hash(x)=71616419)
347
+ 2870 train 4.167860 (lr=6.3913e-04) (hash(x)=78656517)
348
+ 2880 train 4.179040 (lr=6.3357e-04) (hash(x)=80073987)
349
+ 2890 train 4.105946 (lr=6.2802e-04) (hash(x)=76894809)
350
+ 2900 val loss 4.1507
351
+ 2900 val perplexity 63.4758
352
+ 2900 train 4.089047 (lr=6.2248e-04) (hash(x)=80499838)
353
+ 2910 train 4.134538 (lr=6.1696e-04) (hash(x)=72673354)
354
+ 2920 train 4.113180 (lr=6.1146e-04) (hash(x)=84265768)
355
+ 2930 train 4.144595 (lr=6.0597e-04) (hash(x)=79612060)
356
+ 2940 train 4.004796 (lr=6.0050e-04) (hash(x)=74970087)
357
+ 2950 train 4.156507 (lr=5.9504e-04) (hash(x)=84166818)
358
+ 2960 train 4.143835 (lr=5.8961e-04) (hash(x)=89410221)
359
+ 2970 train 4.084720 (lr=5.8418e-04) (hash(x)=75672566)
360
+ 2980 train 3.947438 (lr=5.7878e-04) (hash(x)=81760314)
361
+ 2990 train 3.922724 (lr=5.7339e-04) (hash(x)=80605200)
362
+ 3000 val loss 4.1521
363
+ 3000 val perplexity 63.5671
364
+ 3000 train 3.909436 (lr=5.6803e-04) (hash(x)=83804735)
365
+ 3010 train 3.843107 (lr=5.6268e-04) (hash(x)=77015303)
366
+ 3020 train 3.911096 (lr=5.5735e-04) (hash(x)=81464523)
367
+ 3030 train 4.102357 (lr=5.5204e-04) (hash(x)=80168230)
368
+ 3040 train 4.049564 (lr=5.4675e-04) (hash(x)=84817006)
369
+ 3050 train 4.158447 (lr=5.4148e-04) (hash(x)=71601811)
370
+ 3060 train 4.073994 (lr=5.3624e-04) (hash(x)=85499733)
371
+ 3070 train 4.127561 (lr=5.3101e-04) (hash(x)=78670408)
372
+ 3080 train 4.021262 (lr=5.2581e-04) (hash(x)=77120468)
373
+ 3090 train 4.129619 (lr=5.2062e-04) (hash(x)=77927426)
374
+ 3100 val loss 4.1320
375
+ 3100 val perplexity 62.3038
376
+ 3100 train 4.142215 (lr=5.1546e-04) (hash(x)=83998606)
377
+ 3110 train 4.043350 (lr=5.1033e-04) (hash(x)=77990218)
378
+ 3120 train 3.956764 (lr=5.0521e-04) (hash(x)=81623970)
379
+ 3130 train 3.930210 (lr=5.0012e-04) (hash(x)=75014781)
380
+ 3140 train 3.984820 (lr=4.9505e-04) (hash(x)=72591250)
381
+ 3150 train 4.000894 (lr=4.9001e-04) (hash(x)=81421847)
382
+ 3160 train 4.049676 (lr=4.8499e-04) (hash(x)=75112631)
383
+ 3170 train 4.148863 (lr=4.8000e-04) (hash(x)=87518033)
384
+ 3180 train 3.981578 (lr=4.7503e-04) (hash(x)=78648348)
385
+ 3190 train 4.112570 (lr=4.7009e-04) (hash(x)=78270029)
386
+ 3200 val loss 4.1136
387
+ 3200 val perplexity 61.1637
388
+ 3200 train 4.185241 (lr=4.6518e-04) (hash(x)=83380714)
389
+ 3210 train 4.365189 (lr=4.6029e-04) (hash(x)=67804991)
390
+ 3220 train 4.099415 (lr=4.5543e-04) (hash(x)=76968804)
391
+ 3230 train 4.034735 (lr=4.5059e-04) (hash(x)=75798670)
392
+ 3240 train 4.107241 (lr=4.4579e-04) (hash(x)=83192811)
393
+ 3250 train 4.048763 (lr=4.4101e-04) (hash(x)=81659789)
394
+ 3260 train 4.081182 (lr=4.3626e-04) (hash(x)=80622502)
395
+ 3270 train 4.055993 (lr=4.3153e-04) (hash(x)=85436511)
396
+ 3280 train 4.001787 (lr=4.2684e-04) (hash(x)=80589180)
397
+ 3290 train 4.142225 (lr=4.2218e-04) (hash(x)=85747193)
398
+ 3300 val loss 4.1081
399
+ 3300 val perplexity 60.8290
400
+ 3300 train 4.002334 (lr=4.1754e-04) (hash(x)=74635692)
401
+ 3310 train 3.989581 (lr=4.1294e-04) (hash(x)=82045455)
402
+ 3320 train 3.865089 (lr=4.0836e-04) (hash(x)=73249173)
403
+ 3330 train 3.821280 (lr=4.0382e-04) (hash(x)=71492338)
404
+ 3340 train 4.158810 (lr=3.9931e-04) (hash(x)=76374071)
405
+ 3350 train 4.121274 (lr=3.9483e-04) (hash(x)=80474064)
406
+ 3360 train 4.038265 (lr=3.9038e-04) (hash(x)=84390892)
407
+ 3370 train 4.087763 (lr=3.8596e-04) (hash(x)=83399949)
408
+ 3380 train 4.131597 (lr=3.8158e-04) (hash(x)=77648059)
409
+ 3390 train 4.104954 (lr=3.7723e-04) (hash(x)=80152701)
410
+ 3400 val loss 4.0885
411
+ 3400 val perplexity 59.6481
412
+ 3400 train 4.151741 (lr=3.7291e-04) (hash(x)=81952545)
413
+ 3410 train 4.114931 (lr=3.6862e-04) (hash(x)=80908993)
414
+ 3420 train 4.152986 (lr=3.6437e-04) (hash(x)=79610037)
415
+ 3430 train 4.022924 (lr=3.6015e-04) (hash(x)=87624382)
416
+ 3440 train 4.078052 (lr=3.5597e-04) (hash(x)=82336381)
417
+ 3450 train 4.054916 (lr=3.5182e-04) (hash(x)=87687835)
418
+ 3460 train 4.055671 (lr=3.4770e-04) (hash(x)=76376135)
419
+ 3470 train 3.967362 (lr=3.4362e-04) (hash(x)=79331391)
420
+ 3480 train 3.880614 (lr=3.3958e-04) (hash(x)=107002681)
421
+ 3490 train 3.955560 (lr=3.3557e-04) (hash(x)=77465514)
422
+ 3500 val loss 4.0854
423
+ 3500 val perplexity 59.4684
424
+ 3500 train 3.920421 (lr=3.3160e-04) (hash(x)=88237229)
425
+ 3510 train 3.990474 (lr=3.2766e-04) (hash(x)=81011739)
426
+ 3520 train 4.042212 (lr=3.2377e-04) (hash(x)=64643427)
427
+ 3530 train 4.301577 (lr=3.1990e-04) (hash(x)=78029539)
428
+ 3540 train 4.150437 (lr=3.1608e-04) (hash(x)=83188968)
429
+ 3550 train 4.168146 (lr=3.1229e-04) (hash(x)=86104185)
430
+ 3560 train 4.090512 (lr=3.0854e-04) (hash(x)=81993629)
431
+ 3570 train 4.066632 (lr=3.0483e-04) (hash(x)=72719368)
432
+ 3580 train 4.078227 (lr=3.0116e-04) (hash(x)=90091487)
433
+ 3590 train 3.963018 (lr=2.9752e-04) (hash(x)=77393152)
434
+ 3600 val loss 4.0676
435
+ 3600 val perplexity 58.4164
436
+ 3600 train 4.070786 (lr=2.9393e-04) (hash(x)=73103504)
437
+ 3610 train 4.073221 (lr=2.9037e-04) (hash(x)=84429400)
438
+ 3620 train 4.048222 (lr=2.8685e-04) (hash(x)=77021795)
439
+ 3630 train 3.998981 (lr=2.8337e-04) (hash(x)=92830605)
440
+ 3640 train 4.194371 (lr=2.7994e-04) (hash(x)=78313175)
441
+ 3650 train 3.993938 (lr=2.7654e-04) (hash(x)=85395549)
442
+ 3660 train 3.793330 (lr=2.7318e-04) (hash(x)=78114459)
443
+ 3670 train 3.919963 (lr=2.6987e-04) (hash(x)=74968316)
444
+ 3680 train 3.848638 (lr=2.6659e-04) (hash(x)=73358737)
445
+ 3690 train 3.953099 (lr=2.6336e-04) (hash(x)=76399442)
446
+ 3700 val loss 4.0699
447
+ 3700 val perplexity 58.5501
448
+ 3700 train 3.953761 (lr=2.6016e-04) (hash(x)=74039273)
449
+ 3710 train 4.130785 (lr=2.5701e-04) (hash(x)=79299680)
450
+ 3720 train 4.050108 (lr=2.5390e-04) (hash(x)=94701498)
451
+ 3730 train 4.123174 (lr=2.5083e-04) (hash(x)=75352071)
452
+ 3740 train 4.003727 (lr=2.4781e-04) (hash(x)=77909487)
453
+ 3750 train 4.104887 (lr=2.4483e-04) (hash(x)=73986730)
454
+ 3760 train 4.062451 (lr=2.4189e-04) (hash(x)=79325763)
455
+ 3770 train 4.113098 (lr=2.3899e-04) (hash(x)=72457818)
456
+ 3780 train 4.249071 (lr=2.3614e-04) (hash(x)=71775590)
457
+ 3790 train 4.131860 (lr=2.3333e-04) (hash(x)=82638943)
458
+ 3800 val loss 4.0532
459
+ 3800 val perplexity 57.5820
460
+ 3800 train 4.102822 (lr=2.3056e-04) (hash(x)=79965893)
461
+ 3810 train 4.040622 (lr=2.2784e-04) (hash(x)=72598235)
462
+ 3820 train 4.020828 (lr=2.2516e-04) (hash(x)=83113889)
463
+ 3830 train 4.083835 (lr=2.2252e-04) (hash(x)=74434590)
464
+ 3840 train 4.001180 (lr=2.1993e-04) (hash(x)=82860348)
465
+ 3850 train 3.935466 (lr=2.1739e-04) (hash(x)=78067565)
466
+ 3860 train 4.006458 (lr=2.1488e-04) (hash(x)=82592498)
467
+ 3870 train 3.938233 (lr=2.1243e-04) (hash(x)=81820733)
468
+ 3880 train 3.957686 (lr=2.1002e-04) (hash(x)=87709040)
469
+ 3890 train 4.002182 (lr=2.0765e-04) (hash(x)=70379093)
470
+ 3900 val loss 4.0471
471
+ 3900 val perplexity 57.2291
472
+ 3900 train 4.125259 (lr=2.0533e-04) (hash(x)=76597431)
473
+ 3910 train 4.010365 (lr=2.0306e-04) (hash(x)=90490716)
474
+ 3920 train 4.112679 (lr=2.0083e-04) (hash(x)=81970659)
475
+ 3930 train 4.097351 (lr=1.9865e-04) (hash(x)=81496334)
476
+ 3940 train 4.071361 (lr=1.9651e-04) (hash(x)=75717605)
477
+ 3950 train 4.020153 (lr=1.9442e-04) (hash(x)=82749357)
478
+ 3960 train 4.046355 (lr=1.9238e-04) (hash(x)=89355157)
479
+ 3970 train 4.008465 (lr=1.9038e-04) (hash(x)=78980403)
480
+ 3980 train 3.977622 (lr=1.8843e-04) (hash(x)=76627217)
481
+ 3990 train 3.976734 (lr=1.8652e-04) (hash(x)=72412879)
482
+ 4000 val loss 4.0427
483
+ 4000 val perplexity 56.9802
484
+ 4000 train 3.990011 (lr=1.8467e-04) (hash(x)=83018142)
485
+ 4010 train 4.000363 (lr=1.8286e-04) (hash(x)=81272436)
486
+ 4020 train 3.934374 (lr=1.8110e-04) (hash(x)=85497482)
487
+ 4030 train 4.010159 (lr=1.7938e-04) (hash(x)=85241734)
488
+ 4040 train 3.978693 (lr=1.7771e-04) (hash(x)=77925307)
489
+ 4050 train 3.970547 (lr=1.7610e-04) (hash(x)=84826179)
490
+ 4060 train 4.039222 (lr=1.7452e-04) (hash(x)=83606764)
491
+ 4070 train 3.947942 (lr=1.7300e-04) (hash(x)=80567590)
492
+ 4080 train 3.994195 (lr=1.7152e-04) (hash(x)=76860998)
493
+ 4090 train 4.032157 (lr=1.7010e-04) (hash(x)=74902328)
494
+ 4100 val loss 4.0322
495
+ 4100 val perplexity 56.3868
496
+ 4100 train 4.074505 (lr=1.6872e-04) (hash(x)=82832041)
497
+ 4110 train 3.956889 (lr=1.6739e-04) (hash(x)=79143262)
498
+ 4120 train 3.977154 (lr=1.6610e-04) (hash(x)=77038149)
499
+ 4130 train 4.165465 (lr=1.6487e-04) (hash(x)=86339074)
500
+ 4140 train 4.231615 (lr=1.6369e-04) (hash(x)=76686216)
501
+ 4150 train 4.006955 (lr=1.6255e-04) (hash(x)=70522682)
502
+ 4160 train 3.966703 (lr=1.6146e-04) (hash(x)=90958555)
503
+ 4170 train 3.998724 (lr=1.6042e-04) (hash(x)=91463532)
504
+ 4180 train 3.983728 (lr=1.5943e-04) (hash(x)=81959329)
505
+ 4190 train 3.947684 (lr=1.5849e-04) (hash(x)=83146752)
506
+ 4200 val loss 4.0345
507
+ 4200 val perplexity 56.5142
508
+ 4200 train 3.927823 (lr=1.5760e-04) (hash(x)=78361715)
509
+ 4210 train 4.076822 (lr=1.5676e-04) (hash(x)=87364889)
510
+ 4220 train 3.926729 (lr=1.5597e-04) (hash(x)=70465156)
511
+ 4230 train 4.085226 (lr=1.5522e-04) (hash(x)=84524081)
512
+ 4240 train 4.041117 (lr=1.5453e-04) (hash(x)=77824868)
513
+ 4250 train 4.053056 (lr=1.5388e-04) (hash(x)=81710711)
514
+ 4260 train 4.028132 (lr=1.5329e-04) (hash(x)=76362728)
515
+ 4270 train 4.016611 (lr=1.5274e-04) (hash(x)=83115208)
516
+ 4280 train 3.932349 (lr=1.5224e-04) (hash(x)=87218314)
517
+ 4290 train 4.088923 (lr=1.5180e-04) (hash(x)=74582673)
518
+ 4300 val loss 4.0294
519
+ 4300 val perplexity 56.2287
520
+ 4300 train 3.971585 (lr=1.5140e-04) (hash(x)=77379615)
521
+ 4310 train 3.944555 (lr=1.5105e-04) (hash(x)=78669579)
522
+ 4320 train 4.064184 (lr=1.5075e-04) (hash(x)=83066608)
523
+ 4330 train 4.027389 (lr=1.5050e-04) (hash(x)=83037340)
524
+ 4340 train 4.064446 (lr=1.5030e-04) (hash(x)=82849771)
525
+ 4350 train 4.001035 (lr=1.5016e-04) (hash(x)=76693985)
526
+ 4360 train 4.088914 (lr=1.5006e-04) (hash(x)=77745394)
527
+ 4370 train 4.052715 (lr=1.5001e-04) (hash(x)=79954388)
528
+ 4374 val loss 4.0237
529
+ 4374 val perplexity 55.9062
12_mini_head_one_mask_per_head_1_latent_mask_halved_lr/model_04374.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:90f9ad6bd1bb899942de2335b938a180b28b675cbfb93c61ed97b6f8cf90c07f
3
+ size 97706354
12_mini_head_one_mask_per_head_1_latent_mask_halved_lr/optimizer_04374.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:2f854081622ea057b0b1f0f2a92637572db850e9e25ed7340c48995ba6168014
3
+ size 189135414