andrew-healey commited on
Commit
b132d66
·
verified ·
1 Parent(s): 7395c41

Upload folder using huggingface_hub

Browse files
12_head_baseline_lr_20e-4_head_dim_22_fixed/args.json ADDED
@@ -0,0 +1 @@
 
 
1
+ {"hellaswag": true, "attention_kind": "selective", "log_dir": "logs/repro_selective_pattern_rankings/12_head_baseline_lr_20e-4_head_dim_22_fixed", "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": "none", "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": 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": 0.002, "decay_lr": true, "readout_zero_init": false, "query_zero_init": false, "l1_loss": false, "debugpy": false, "key": "12_head_baseline_lr_20e-4_fixed"}
12_head_baseline_lr_20e-4_head_dim_22_fixed/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_head_baseline_lr_20e-4_head_dim_22_fixed/log2.txt ADDED
@@ -0,0 +1,529 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ max_steps: 4375
2
+ 0 val loss 11.4043
3
+ 0 val perplexity 89706.7812
4
+ 0 train 11.388929 (lr=2.7972e-06) (hash(x)=93356070)
5
+ 10 train 10.155953 (lr=3.0769e-05) (hash(x)=91253010)
6
+ 20 train 9.680853 (lr=5.8741e-05) (hash(x)=74934453)
7
+ 30 train 9.160400 (lr=8.6713e-05) (hash(x)=79278034)
8
+ 40 train 8.461512 (lr=1.1469e-04) (hash(x)=80107892)
9
+ 50 train 7.943209 (lr=1.4266e-04) (hash(x)=70745428)
10
+ 60 train 7.786048 (lr=1.7063e-04) (hash(x)=80077589)
11
+ 70 train 7.507499 (lr=1.9860e-04) (hash(x)=76213766)
12
+ 80 train 7.522699 (lr=2.2657e-04) (hash(x)=83218328)
13
+ 90 train 7.342048 (lr=2.5455e-04) (hash(x)=74855845)
14
+ 100 val loss 7.2428
15
+ 100 val perplexity 1397.9952
16
+ 100 train 7.081751 (lr=2.8252e-04) (hash(x)=82814902)
17
+ 110 train 7.070141 (lr=3.1049e-04) (hash(x)=84907741)
18
+ 120 train 6.914880 (lr=3.3846e-04) (hash(x)=82613223)
19
+ 130 train 6.733666 (lr=3.6643e-04) (hash(x)=83540876)
20
+ 140 train 6.690533 (lr=3.9441e-04) (hash(x)=75095216)
21
+ 150 train 6.751195 (lr=4.2238e-04) (hash(x)=97190944)
22
+ 160 train 6.670516 (lr=4.5035e-04) (hash(x)=82117809)
23
+ 170 train 6.552846 (lr=4.7832e-04) (hash(x)=70514724)
24
+ 180 train 6.480225 (lr=5.0629e-04) (hash(x)=81029624)
25
+ 190 train 6.394446 (lr=5.3427e-04) (hash(x)=78996841)
26
+ 200 val loss 6.3445
27
+ 200 val perplexity 569.3679
28
+ 200 train 6.349635 (lr=5.6224e-04) (hash(x)=79845097)
29
+ 210 train 6.230053 (lr=5.9021e-04) (hash(x)=78997683)
30
+ 220 train 6.134311 (lr=6.1818e-04) (hash(x)=74895865)
31
+ 230 train 6.267432 (lr=6.4615e-04) (hash(x)=80933276)
32
+ 240 train 5.973943 (lr=6.7413e-04) (hash(x)=77664606)
33
+ 250 train 6.012483 (lr=7.0210e-04) (hash(x)=75903930)
34
+ 260 train 5.916834 (lr=7.3007e-04) (hash(x)=82996853)
35
+ 270 train 5.829728 (lr=7.5804e-04) (hash(x)=73269568)
36
+ 280 train 5.786141 (lr=7.8601e-04) (hash(x)=82906427)
37
+ 290 train 5.698073 (lr=8.1399e-04) (hash(x)=71797895)
38
+ 300 val loss 5.8610
39
+ 300 val perplexity 351.0898
40
+ 300 train 5.572039 (lr=8.4196e-04) (hash(x)=85232249)
41
+ 310 train 5.500593 (lr=8.6993e-04) (hash(x)=75252489)
42
+ 320 train 5.576862 (lr=8.9790e-04) (hash(x)=74394644)
43
+ 330 train 5.683979 (lr=9.2587e-04) (hash(x)=82772910)
44
+ 340 train 5.708167 (lr=9.5385e-04) (hash(x)=81627464)
45
+ 350 train 5.688528 (lr=9.8182e-04) (hash(x)=84678053)
46
+ 360 train 5.631006 (lr=1.0098e-03) (hash(x)=81884128)
47
+ 370 train 5.636319 (lr=1.0378e-03) (hash(x)=76379242)
48
+ 380 train 5.443344 (lr=1.0657e-03) (hash(x)=98182875)
49
+ 390 train 5.416781 (lr=1.0937e-03) (hash(x)=79710436)
50
+ 400 val loss 5.4776
51
+ 400 val perplexity 239.2821
52
+ 400 train 5.421921 (lr=1.1217e-03) (hash(x)=79841071)
53
+ 410 train 5.377325 (lr=1.1497e-03) (hash(x)=75844151)
54
+ 420 train 5.385082 (lr=1.1776e-03) (hash(x)=73125036)
55
+ 430 train 5.406912 (lr=1.2056e-03) (hash(x)=84214858)
56
+ 440 train 5.300082 (lr=1.2336e-03) (hash(x)=80456994)
57
+ 450 train 5.283535 (lr=1.2615e-03) (hash(x)=76735962)
58
+ 460 train 5.190903 (lr=1.2895e-03) (hash(x)=81845446)
59
+ 470 train 5.008363 (lr=1.3175e-03) (hash(x)=76094689)
60
+ 480 train 5.147998 (lr=1.3455e-03) (hash(x)=83806686)
61
+ 490 train 4.955710 (lr=1.3734e-03) (hash(x)=84690227)
62
+ 500 val loss 5.2888
63
+ 500 val perplexity 198.0969
64
+ 500 train 5.027457 (lr=1.4014e-03) (hash(x)=71851938)
65
+ 510 train 5.145571 (lr=1.4294e-03) (hash(x)=77159346)
66
+ 520 train 5.195375 (lr=1.4573e-03) (hash(x)=80755753)
67
+ 530 train 5.294312 (lr=1.4853e-03) (hash(x)=75379679)
68
+ 540 train 5.214166 (lr=1.5133e-03) (hash(x)=82458619)
69
+ 550 train 5.179514 (lr=1.5413e-03) (hash(x)=86366396)
70
+ 560 train 5.220656 (lr=1.5692e-03) (hash(x)=85095044)
71
+ 570 train 5.167289 (lr=1.5972e-03) (hash(x)=78385159)
72
+ 580 train 5.109234 (lr=1.6252e-03) (hash(x)=79342394)
73
+ 590 train 5.137025 (lr=1.6531e-03) (hash(x)=70782192)
74
+ 600 val loss 5.1027
75
+ 600 val perplexity 164.4602
76
+ 600 train 5.081174 (lr=1.6811e-03) (hash(x)=76130353)
77
+ 610 train 5.012655 (lr=1.7091e-03) (hash(x)=74778440)
78
+ 620 train 4.976805 (lr=1.7371e-03) (hash(x)=79129709)
79
+ 630 train 4.886073 (lr=1.7650e-03) (hash(x)=76469962)
80
+ 640 train 4.874603 (lr=1.7930e-03) (hash(x)=78288049)
81
+ 650 train 4.827965 (lr=1.8210e-03) (hash(x)=76641388)
82
+ 660 train 4.780703 (lr=1.8490e-03) (hash(x)=75937906)
83
+ 670 train 4.763486 (lr=1.8769e-03) (hash(x)=74807157)
84
+ 680 train 4.645681 (lr=1.9049e-03) (hash(x)=77490144)
85
+ 690 train 4.672904 (lr=1.9329e-03) (hash(x)=72900124)
86
+ 700 val loss 4.9424
87
+ 700 val perplexity 140.0994
88
+ 700 train 4.631561 (lr=1.9608e-03) (hash(x)=73424218)
89
+ 710 train 4.930869 (lr=1.9888e-03) (hash(x)=86052345)
90
+ 720 train 4.908722 (lr=2.0000e-03) (hash(x)=77462613)
91
+ 730 train 4.898405 (lr=1.9999e-03) (hash(x)=79027471)
92
+ 740 train 4.861896 (lr=1.9998e-03) (hash(x)=78149992)
93
+ 750 train 4.800230 (lr=1.9996e-03) (hash(x)=89147499)
94
+ 760 train 4.842011 (lr=1.9993e-03) (hash(x)=74931538)
95
+ 770 train 4.778931 (lr=1.9990e-03) (hash(x)=89648204)
96
+ 780 train 4.836777 (lr=1.9986e-03) (hash(x)=83708293)
97
+ 790 train 4.859270 (lr=1.9981e-03) (hash(x)=84336512)
98
+ 800 val loss 4.8048
99
+ 800 val perplexity 122.0980
100
+ 800 train 4.723723 (lr=1.9976e-03) (hash(x)=75025285)
101
+ 810 train 4.665617 (lr=1.9970e-03) (hash(x)=79436536)
102
+ 820 train 4.662936 (lr=1.9963e-03) (hash(x)=90467390)
103
+ 830 train 4.672590 (lr=1.9956e-03) (hash(x)=77292020)
104
+ 840 train 4.603032 (lr=1.9948e-03) (hash(x)=75568927)
105
+ 850 train 4.627654 (lr=1.9940e-03) (hash(x)=79671781)
106
+ 860 train 4.719775 (lr=1.9930e-03) (hash(x)=84280943)
107
+ 870 train 4.695788 (lr=1.9920e-03) (hash(x)=80901294)
108
+ 880 train 4.819887 (lr=1.9910e-03) (hash(x)=81437584)
109
+ 890 train 4.670351 (lr=1.9899e-03) (hash(x)=73893778)
110
+ 900 val loss 4.7010
111
+ 900 val perplexity 110.0528
112
+ 900 train 4.632740 (lr=1.9887e-03) (hash(x)=74987794)
113
+ 910 train 4.637210 (lr=1.9874e-03) (hash(x)=93721374)
114
+ 920 train 4.620126 (lr=1.9861e-03) (hash(x)=79149678)
115
+ 930 train 4.575650 (lr=1.9847e-03) (hash(x)=83179387)
116
+ 940 train 4.579799 (lr=1.9833e-03) (hash(x)=89009978)
117
+ 950 train 4.569757 (lr=1.9818e-03) (hash(x)=76483735)
118
+ 960 train 4.495657 (lr=1.9802e-03) (hash(x)=78638850)
119
+ 970 train 4.532346 (lr=1.9785e-03) (hash(x)=90415206)
120
+ 980 train 4.470536 (lr=1.9768e-03) (hash(x)=90819195)
121
+ 990 train 4.596102 (lr=1.9750e-03) (hash(x)=85938852)
122
+ 1000 val loss 4.5775
123
+ 1000 val perplexity 97.2722
124
+ 1000 train 4.624080 (lr=1.9732e-03) (hash(x)=82436789)
125
+ 1010 train 4.689419 (lr=1.9713e-03) (hash(x)=63022149)
126
+ 1020 train 4.601723 (lr=1.9693e-03) (hash(x)=83501199)
127
+ 1030 train 4.572693 (lr=1.9673e-03) (hash(x)=77093326)
128
+ 1040 train 4.469246 (lr=1.9652e-03) (hash(x)=73437559)
129
+ 1050 train 4.552911 (lr=1.9630e-03) (hash(x)=84550388)
130
+ 1060 train 4.554533 (lr=1.9608e-03) (hash(x)=86184566)
131
+ 1070 train 4.534015 (lr=1.9585e-03) (hash(x)=85644922)
132
+ 1080 train 4.474965 (lr=1.9562e-03) (hash(x)=86179801)
133
+ 1090 train 4.447537 (lr=1.9538e-03) (hash(x)=82868303)
134
+ 1100 val loss 4.5196
135
+ 1100 val perplexity 91.8007
136
+ 1100 train 4.546020 (lr=1.9513e-03) (hash(x)=96780388)
137
+ 1110 train 4.491750 (lr=1.9488e-03) (hash(x)=87223122)
138
+ 1120 train 4.454951 (lr=1.9462e-03) (hash(x)=77292786)
139
+ 1130 train 4.479125 (lr=1.9435e-03) (hash(x)=88761375)
140
+ 1140 train 4.384434 (lr=1.9408e-03) (hash(x)=80956468)
141
+ 1150 train 4.560179 (lr=1.9380e-03) (hash(x)=84725462)
142
+ 1160 train 4.551941 (lr=1.9351e-03) (hash(x)=76098113)
143
+ 1170 train 4.516565 (lr=1.9322e-03) (hash(x)=84387685)
144
+ 1180 train 4.465141 (lr=1.9293e-03) (hash(x)=76120321)
145
+ 1190 train 4.431442 (lr=1.9262e-03) (hash(x)=85974065)
146
+ 1200 val loss 4.4486
147
+ 1200 val perplexity 85.5059
148
+ 1200 train 4.338611 (lr=1.9231e-03) (hash(x)=79340644)
149
+ 1210 train 4.418822 (lr=1.9200e-03) (hash(x)=88345025)
150
+ 1220 train 4.331698 (lr=1.9168e-03) (hash(x)=86406230)
151
+ 1230 train 4.315969 (lr=1.9135e-03) (hash(x)=80608975)
152
+ 1240 train 4.377308 (lr=1.9102e-03) (hash(x)=81635225)
153
+ 1250 train 4.217971 (lr=1.9068e-03) (hash(x)=82126376)
154
+ 1260 train 4.391666 (lr=1.9033e-03) (hash(x)=83737972)
155
+ 1270 train 4.300951 (lr=1.8998e-03) (hash(x)=81132345)
156
+ 1280 train 4.389063 (lr=1.8962e-03) (hash(x)=85850404)
157
+ 1290 train 4.402140 (lr=1.8926e-03) (hash(x)=83490640)
158
+ 1300 val loss 4.4067
159
+ 1300 val perplexity 81.9997
160
+ 1300 train 4.374988 (lr=1.8889e-03) (hash(x)=77545187)
161
+ 1310 train 4.418741 (lr=1.8852e-03) (hash(x)=86412685)
162
+ 1320 train 4.318754 (lr=1.8813e-03) (hash(x)=86429640)
163
+ 1330 train 4.447455 (lr=1.8775e-03) (hash(x)=79645281)
164
+ 1340 train 4.408598 (lr=1.8736e-03) (hash(x)=82668541)
165
+ 1350 train 4.274465 (lr=1.8696e-03) (hash(x)=73627248)
166
+ 1360 train 4.367060 (lr=1.8656e-03) (hash(x)=74845133)
167
+ 1370 train 4.279519 (lr=1.8615e-03) (hash(x)=77850497)
168
+ 1380 train 4.306496 (lr=1.8573e-03) (hash(x)=76143954)
169
+ 1390 train 4.314542 (lr=1.8531e-03) (hash(x)=80675544)
170
+ 1400 val loss 4.3810
171
+ 1400 val perplexity 79.9188
172
+ 1400 train 4.290009 (lr=1.8489e-03) (hash(x)=76010938)
173
+ 1410 train 4.338096 (lr=1.8445e-03) (hash(x)=81260300)
174
+ 1420 train 4.451114 (lr=1.8402e-03) (hash(x)=83803629)
175
+ 1430 train 4.280384 (lr=1.8358e-03) (hash(x)=83840110)
176
+ 1440 train 4.469792 (lr=1.8313e-03) (hash(x)=97711831)
177
+ 1450 train 4.345677 (lr=1.8268e-03) (hash(x)=81110139)
178
+ 1460 train 4.271200 (lr=1.8222e-03) (hash(x)=90942839)
179
+ 1470 train 4.312171 (lr=1.8175e-03) (hash(x)=80721564)
180
+ 1480 train 4.398525 (lr=1.8128e-03) (hash(x)=77852059)
181
+ 1490 train 4.299018 (lr=1.8081e-03) (hash(x)=80011365)
182
+ 1500 val loss 4.3376
183
+ 1500 val perplexity 76.5224
184
+ 1500 train 4.329418 (lr=1.8033e-03) (hash(x)=84496142)
185
+ 1510 train 4.259302 (lr=1.7985e-03) (hash(x)=70186729)
186
+ 1520 train 4.282751 (lr=1.7936e-03) (hash(x)=74854227)
187
+ 1530 train 4.193608 (lr=1.7886e-03) (hash(x)=77468161)
188
+ 1540 train 4.211213 (lr=1.7836e-03) (hash(x)=87853059)
189
+ 1550 train 4.143783 (lr=1.7786e-03) (hash(x)=71225436)
190
+ 1560 train 4.246489 (lr=1.7735e-03) (hash(x)=84072783)
191
+ 1570 train 4.292968 (lr=1.7683e-03) (hash(x)=72723098)
192
+ 1580 train 4.310395 (lr=1.7631e-03) (hash(x)=90409866)
193
+ 1590 train 4.386286 (lr=1.7579e-03) (hash(x)=87481378)
194
+ 1600 val loss 4.3100
195
+ 1600 val perplexity 74.4422
196
+ 1600 train 4.365993 (lr=1.7526e-03) (hash(x)=77643862)
197
+ 1610 train 4.256927 (lr=1.7472e-03) (hash(x)=88862575)
198
+ 1620 train 4.271255 (lr=1.7418e-03) (hash(x)=84612581)
199
+ 1630 train 4.302596 (lr=1.7364e-03) (hash(x)=87075989)
200
+ 1640 train 4.195437 (lr=1.7309e-03) (hash(x)=88277361)
201
+ 1650 train 4.311106 (lr=1.7254e-03) (hash(x)=78750236)
202
+ 1660 train 4.227193 (lr=1.7198e-03) (hash(x)=82604581)
203
+ 1670 train 4.066800 (lr=1.7142e-03) (hash(x)=68482265)
204
+ 1680 train 4.188039 (lr=1.7085e-03) (hash(x)=75088835)
205
+ 1690 train 4.137449 (lr=1.7028e-03) (hash(x)=66060989)
206
+ 1700 val loss 4.3105
207
+ 1700 val perplexity 74.4799
208
+ 1700 train 4.303092 (lr=1.6970e-03) (hash(x)=79986754)
209
+ 1710 train 4.292951 (lr=1.6912e-03) (hash(x)=83657930)
210
+ 1720 train 4.233716 (lr=1.6854e-03) (hash(x)=81754135)
211
+ 1730 train 4.317089 (lr=1.6795e-03) (hash(x)=78904427)
212
+ 1740 train 4.214405 (lr=1.6736e-03) (hash(x)=85920177)
213
+ 1750 train 4.268301 (lr=1.6676e-03) (hash(x)=86573211)
214
+ 1760 train 4.226496 (lr=1.6616e-03) (hash(x)=81737128)
215
+ 1770 train 4.218390 (lr=1.6555e-03) (hash(x)=77400968)
216
+ 1780 train 4.229881 (lr=1.6494e-03) (hash(x)=73545497)
217
+ 1790 train 4.482973 (lr=1.6433e-03) (hash(x)=71641943)
218
+ 1800 val loss 4.2666
219
+ 1800 val perplexity 71.2754
220
+ 1800 train 4.330650 (lr=1.6371e-03) (hash(x)=87819781)
221
+ 1810 train 4.188914 (lr=1.6309e-03) (hash(x)=86870770)
222
+ 1820 train 4.197684 (lr=1.6246e-03) (hash(x)=82522211)
223
+ 1830 train 4.282329 (lr=1.6183e-03) (hash(x)=61947437)
224
+ 1840 train 4.163430 (lr=1.6120e-03) (hash(x)=79865406)
225
+ 1850 train 4.122688 (lr=1.6056e-03) (hash(x)=79828721)
226
+ 1860 train 4.270079 (lr=1.5992e-03) (hash(x)=80869571)
227
+ 1870 train 4.255038 (lr=1.5928e-03) (hash(x)=73780971)
228
+ 1880 train 4.214435 (lr=1.5863e-03) (hash(x)=79249549)
229
+ 1890 train 4.268536 (lr=1.5798e-03) (hash(x)=81041904)
230
+ 1900 val loss 4.2431
231
+ 1900 val perplexity 69.6259
232
+ 1900 train 4.336476 (lr=1.5732e-03) (hash(x)=82456430)
233
+ 1910 train 4.109306 (lr=1.5666e-03) (hash(x)=82222135)
234
+ 1920 train 4.143790 (lr=1.5600e-03) (hash(x)=70033249)
235
+ 1930 train 4.130184 (lr=1.5533e-03) (hash(x)=72887360)
236
+ 1940 train 4.179301 (lr=1.5467e-03) (hash(x)=83251100)
237
+ 1950 train 4.212892 (lr=1.5399e-03) (hash(x)=79660266)
238
+ 1960 train 4.096307 (lr=1.5332e-03) (hash(x)=78919068)
239
+ 1970 train 4.198797 (lr=1.5264e-03) (hash(x)=94690431)
240
+ 1980 train 4.037574 (lr=1.5196e-03) (hash(x)=88153756)
241
+ 1990 train 4.085520 (lr=1.5127e-03) (hash(x)=77195688)
242
+ 2000 val loss 4.2188
243
+ 2000 val perplexity 67.9518
244
+ 2000 train 4.182022 (lr=1.5058e-03) (hash(x)=81308591)
245
+ 2010 train 4.235046 (lr=1.4989e-03) (hash(x)=80205479)
246
+ 2020 train 4.153543 (lr=1.4920e-03) (hash(x)=81008704)
247
+ 2030 train 4.215045 (lr=1.4850e-03) (hash(x)=82914358)
248
+ 2040 train 4.234581 (lr=1.4780e-03) (hash(x)=88064399)
249
+ 2050 train 4.232649 (lr=1.4710e-03) (hash(x)=82889390)
250
+ 2060 train 4.307598 (lr=1.4639e-03) (hash(x)=80913554)
251
+ 2070 train 4.143919 (lr=1.4569e-03) (hash(x)=78121791)
252
+ 2080 train 4.275077 (lr=1.4498e-03) (hash(x)=75356657)
253
+ 2090 train 4.099466 (lr=1.4426e-03) (hash(x)=74048412)
254
+ 2100 val loss 4.1962
255
+ 2100 val perplexity 66.4306
256
+ 2100 train 4.019866 (lr=1.4355e-03) (hash(x)=68928225)
257
+ 2110 train 4.074094 (lr=1.4283e-03) (hash(x)=93734745)
258
+ 2120 train 4.071678 (lr=1.4211e-03) (hash(x)=72878164)
259
+ 2130 train 4.196694 (lr=1.4139e-03) (hash(x)=71508226)
260
+ 2140 train 4.163470 (lr=1.4066e-03) (hash(x)=84168671)
261
+ 2150 train 4.093198 (lr=1.3993e-03) (hash(x)=74673639)
262
+ 2160 train 4.239716 (lr=1.3920e-03) (hash(x)=75470031)
263
+ 2170 train 4.199570 (lr=1.3847e-03) (hash(x)=74307890)
264
+ 2180 train 4.318952 (lr=1.3774e-03) (hash(x)=77214245)
265
+ 2190 train 4.099724 (lr=1.3700e-03) (hash(x)=88628359)
266
+ 2200 val loss 4.1772
267
+ 2200 val perplexity 65.1828
268
+ 2200 train 4.116773 (lr=1.3626e-03) (hash(x)=74779126)
269
+ 2210 train 4.152806 (lr=1.3552e-03) (hash(x)=79817976)
270
+ 2220 train 4.108828 (lr=1.3478e-03) (hash(x)=75258996)
271
+ 2230 train 4.077463 (lr=1.3404e-03) (hash(x)=88164047)
272
+ 2240 train 4.064227 (lr=1.3329e-03) (hash(x)=78580686)
273
+ 2250 train 3.999498 (lr=1.3255e-03) (hash(x)=76510617)
274
+ 2260 train 4.077482 (lr=1.3180e-03) (hash(x)=76043966)
275
+ 2270 train 4.057310 (lr=1.3105e-03) (hash(x)=73062098)
276
+ 2280 train 4.285351 (lr=1.3030e-03) (hash(x)=81885909)
277
+ 2290 train 4.221522 (lr=1.2954e-03) (hash(x)=81481741)
278
+ 2300 val loss 4.1555
279
+ 2300 val perplexity 63.7840
280
+ 2300 train 4.074477 (lr=1.2879e-03) (hash(x)=82104275)
281
+ 2310 train 4.122712 (lr=1.2803e-03) (hash(x)=82137309)
282
+ 2320 train 4.100186 (lr=1.2727e-03) (hash(x)=86741167)
283
+ 2330 train 4.117412 (lr=1.2652e-03) (hash(x)=84796263)
284
+ 2340 train 4.084436 (lr=1.2576e-03) (hash(x)=169966529)
285
+ 2350 train 3.965243 (lr=1.2499e-03) (hash(x)=80006061)
286
+ 2360 train 4.025638 (lr=1.2423e-03) (hash(x)=71090999)
287
+ 2370 train 4.149453 (lr=1.2347e-03) (hash(x)=86641850)
288
+ 2380 train 4.049049 (lr=1.2270e-03) (hash(x)=84573253)
289
+ 2390 train 4.081619 (lr=1.2194e-03) (hash(x)=74483764)
290
+ 2400 val loss 4.1536
291
+ 2400 val perplexity 63.6642
292
+ 2400 train 4.058445 (lr=1.2117e-03) (hash(x)=78327659)
293
+ 2410 train 4.127326 (lr=1.2041e-03) (hash(x)=77273627)
294
+ 2420 train 3.988265 (lr=1.1964e-03) (hash(x)=76938049)
295
+ 2430 train 3.988312 (lr=1.1887e-03) (hash(x)=98449442)
296
+ 2440 train 3.855947 (lr=1.1810e-03) (hash(x)=78157797)
297
+ 2450 train 3.894413 (lr=1.1733e-03) (hash(x)=80637582)
298
+ 2460 train 3.881385 (lr=1.1656e-03) (hash(x)=80023854)
299
+ 2470 train 3.940961 (lr=1.1579e-03) (hash(x)=86678884)
300
+ 2480 train 4.161708 (lr=1.1502e-03) (hash(x)=84086469)
301
+ 2490 train 4.185285 (lr=1.1425e-03) (hash(x)=83131141)
302
+ 2500 val loss 4.1253
303
+ 2500 val perplexity 61.8873
304
+ 2500 train 4.101366 (lr=1.1348e-03) (hash(x)=82583497)
305
+ 2510 train 4.162744 (lr=1.1270e-03) (hash(x)=79164326)
306
+ 2520 train 4.072135 (lr=1.1193e-03) (hash(x)=76456503)
307
+ 2530 train 4.130395 (lr=1.1116e-03) (hash(x)=82001228)
308
+ 2540 train 4.164486 (lr=1.1039e-03) (hash(x)=75615595)
309
+ 2550 train 4.109352 (lr=1.0961e-03) (hash(x)=78397869)
310
+ 2560 train 4.114742 (lr=1.0884e-03) (hash(x)=75547032)
311
+ 2570 train 4.022499 (lr=1.0807e-03) (hash(x)=89201025)
312
+ 2580 train 4.010750 (lr=1.0730e-03) (hash(x)=89856704)
313
+ 2590 train 4.018825 (lr=1.0652e-03) (hash(x)=82175682)
314
+ 2600 val loss 4.1078
315
+ 2600 val perplexity 60.8114
316
+ 2600 train 4.029406 (lr=1.0575e-03) (hash(x)=78312826)
317
+ 2610 train 4.021497 (lr=1.0498e-03) (hash(x)=77066588)
318
+ 2620 train 3.857282 (lr=1.0421e-03) (hash(x)=78666061)
319
+ 2630 train 3.864325 (lr=1.0344e-03) (hash(x)=93762143)
320
+ 2640 train 3.875328 (lr=1.0267e-03) (hash(x)=83191587)
321
+ 2650 train 3.796257 (lr=1.0190e-03) (hash(x)=87169585)
322
+ 2660 train 4.013497 (lr=1.0113e-03) (hash(x)=86426388)
323
+ 2670 train 4.153199 (lr=1.0036e-03) (hash(x)=76692638)
324
+ 2680 train 4.069827 (lr=9.9594e-04) (hash(x)=77446063)
325
+ 2690 train 4.066907 (lr=9.8827e-04) (hash(x)=79809050)
326
+ 2700 val loss 4.1018
327
+ 2700 val perplexity 60.4508
328
+ 2700 train 3.993732 (lr=9.8061e-04) (hash(x)=83116823)
329
+ 2710 train 4.107031 (lr=9.7296e-04) (hash(x)=75622148)
330
+ 2720 train 4.036642 (lr=9.6532e-04) (hash(x)=80690512)
331
+ 2730 train 4.105757 (lr=9.5768e-04) (hash(x)=78009984)
332
+ 2740 train 4.056344 (lr=9.5006e-04) (hash(x)=74205488)
333
+ 2750 train 4.061750 (lr=9.4245e-04) (hash(x)=91013332)
334
+ 2760 train 3.986669 (lr=9.3485e-04) (hash(x)=76954961)
335
+ 2770 train 3.993824 (lr=9.2726e-04) (hash(x)=79390317)
336
+ 2780 train 3.861348 (lr=9.1969e-04) (hash(x)=70168783)
337
+ 2790 train 4.028719 (lr=9.1212e-04) (hash(x)=76028417)
338
+ 2800 val loss 4.0823
339
+ 2800 val perplexity 59.2845
340
+ 2800 train 3.858847 (lr=9.0458e-04) (hash(x)=77656050)
341
+ 2810 train 3.901048 (lr=8.9704e-04) (hash(x)=94248216)
342
+ 2820 train 3.961937 (lr=8.8952e-04) (hash(x)=78305078)
343
+ 2830 train 3.873692 (lr=8.8202e-04) (hash(x)=79948848)
344
+ 2840 train 3.743119 (lr=8.7453e-04) (hash(x)=85341024)
345
+ 2850 train 4.048859 (lr=8.6706e-04) (hash(x)=78735170)
346
+ 2860 train 4.297523 (lr=8.5961e-04) (hash(x)=71616419)
347
+ 2870 train 4.071503 (lr=8.5217e-04) (hash(x)=78656517)
348
+ 2880 train 4.081413 (lr=8.4476e-04) (hash(x)=80073987)
349
+ 2890 train 4.013014 (lr=8.3736e-04) (hash(x)=76894809)
350
+ 2900 val loss 4.0561
351
+ 2900 val perplexity 57.7509
352
+ 2900 train 3.990272 (lr=8.2998e-04) (hash(x)=80499838)
353
+ 2910 train 4.046641 (lr=8.2262e-04) (hash(x)=72673354)
354
+ 2920 train 4.026279 (lr=8.1528e-04) (hash(x)=84265768)
355
+ 2930 train 4.057053 (lr=8.0796e-04) (hash(x)=79612060)
356
+ 2940 train 3.908737 (lr=8.0067e-04) (hash(x)=74970087)
357
+ 2950 train 4.062082 (lr=7.9339e-04) (hash(x)=84166818)
358
+ 2960 train 4.051934 (lr=7.8614e-04) (hash(x)=89410221)
359
+ 2970 train 3.991116 (lr=7.7891e-04) (hash(x)=75672566)
360
+ 2980 train 3.845129 (lr=7.7171e-04) (hash(x)=81760314)
361
+ 2990 train 3.823239 (lr=7.6453e-04) (hash(x)=80605200)
362
+ 3000 val loss 4.0562
363
+ 3000 val perplexity 57.7550
364
+ 3000 train 3.815274 (lr=7.5737e-04) (hash(x)=83804735)
365
+ 3010 train 3.746502 (lr=7.5024e-04) (hash(x)=77015303)
366
+ 3020 train 3.814465 (lr=7.4313e-04) (hash(x)=81464523)
367
+ 3030 train 4.013062 (lr=7.3606e-04) (hash(x)=80168230)
368
+ 3040 train 3.950656 (lr=7.2900e-04) (hash(x)=84817006)
369
+ 3050 train 4.058958 (lr=7.2198e-04) (hash(x)=71601811)
370
+ 3060 train 3.979730 (lr=7.1498e-04) (hash(x)=85499733)
371
+ 3070 train 4.037491 (lr=7.0801e-04) (hash(x)=78670408)
372
+ 3080 train 3.926682 (lr=7.0107e-04) (hash(x)=77120468)
373
+ 3090 train 4.033261 (lr=6.9416e-04) (hash(x)=77927426)
374
+ 3100 val loss 4.0353
375
+ 3100 val perplexity 56.5615
376
+ 3100 train 4.050038 (lr=6.8728e-04) (hash(x)=83998606)
377
+ 3110 train 3.947989 (lr=6.8043e-04) (hash(x)=77990218)
378
+ 3120 train 3.865227 (lr=6.7362e-04) (hash(x)=81623970)
379
+ 3130 train 3.836588 (lr=6.6683e-04) (hash(x)=75014781)
380
+ 3140 train 3.892474 (lr=6.6007e-04) (hash(x)=72591250)
381
+ 3150 train 3.906770 (lr=6.5335e-04) (hash(x)=81421847)
382
+ 3160 train 3.956095 (lr=6.4666e-04) (hash(x)=75112631)
383
+ 3170 train 4.053905 (lr=6.4000e-04) (hash(x)=87518033)
384
+ 3180 train 3.894997 (lr=6.3338e-04) (hash(x)=78648348)
385
+ 3190 train 4.017542 (lr=6.2679e-04) (hash(x)=78270029)
386
+ 3200 val loss 4.0160
387
+ 3200 val perplexity 55.4812
388
+ 3200 train 4.091595 (lr=6.2024e-04) (hash(x)=83380714)
389
+ 3210 train 4.267375 (lr=6.1372e-04) (hash(x)=67804991)
390
+ 3220 train 4.000564 (lr=6.0724e-04) (hash(x)=76968804)
391
+ 3230 train 3.937478 (lr=6.0079e-04) (hash(x)=75798670)
392
+ 3240 train 4.011580 (lr=5.9438e-04) (hash(x)=83192811)
393
+ 3250 train 3.947131 (lr=5.8801e-04) (hash(x)=81659789)
394
+ 3260 train 3.991312 (lr=5.8167e-04) (hash(x)=80622502)
395
+ 3270 train 3.959703 (lr=5.7538e-04) (hash(x)=85436511)
396
+ 3280 train 3.903218 (lr=5.6912e-04) (hash(x)=80589180)
397
+ 3290 train 4.043356 (lr=5.6290e-04) (hash(x)=85747193)
398
+ 3300 val loss 4.0113
399
+ 3300 val perplexity 55.2159
400
+ 3300 train 3.909591 (lr=5.5672e-04) (hash(x)=74635692)
401
+ 3310 train 3.896422 (lr=5.5058e-04) (hash(x)=82045455)
402
+ 3320 train 3.766707 (lr=5.4449e-04) (hash(x)=73249173)
403
+ 3330 train 3.727118 (lr=5.3843e-04) (hash(x)=71492338)
404
+ 3340 train 4.062700 (lr=5.3241e-04) (hash(x)=76374071)
405
+ 3350 train 4.030070 (lr=5.2644e-04) (hash(x)=80474064)
406
+ 3360 train 3.943725 (lr=5.2051e-04) (hash(x)=84390892)
407
+ 3370 train 3.985867 (lr=5.1462e-04) (hash(x)=83399949)
408
+ 3380 train 4.038242 (lr=5.0877e-04) (hash(x)=77648059)
409
+ 3390 train 4.008017 (lr=5.0297e-04) (hash(x)=80152701)
410
+ 3400 val loss 3.9922
411
+ 3400 val perplexity 54.1716
412
+ 3400 train 4.060925 (lr=4.9721e-04) (hash(x)=81952545)
413
+ 3410 train 4.018839 (lr=4.9149e-04) (hash(x)=80908993)
414
+ 3420 train 4.061543 (lr=4.8582e-04) (hash(x)=79610037)
415
+ 3430 train 3.923720 (lr=4.8020e-04) (hash(x)=87624382)
416
+ 3440 train 3.983872 (lr=4.7462e-04) (hash(x)=82336381)
417
+ 3450 train 3.961577 (lr=4.6909e-04) (hash(x)=87687835)
418
+ 3460 train 3.963325 (lr=4.6360e-04) (hash(x)=76376135)
419
+ 3470 train 3.874464 (lr=4.5816e-04) (hash(x)=79331391)
420
+ 3480 train 3.786614 (lr=4.5277e-04) (hash(x)=107002681)
421
+ 3490 train 3.864139 (lr=4.4743e-04) (hash(x)=77465514)
422
+ 3500 val loss 3.9882
423
+ 3500 val perplexity 53.9590
424
+ 3500 train 3.825139 (lr=4.4213e-04) (hash(x)=88237229)
425
+ 3510 train 3.892308 (lr=4.3689e-04) (hash(x)=81011739)
426
+ 3520 train 3.945078 (lr=4.3169e-04) (hash(x)=64643427)
427
+ 3530 train 4.217670 (lr=4.2654e-04) (hash(x)=78029539)
428
+ 3540 train 4.051165 (lr=4.2144e-04) (hash(x)=83188968)
429
+ 3550 train 4.070911 (lr=4.1639e-04) (hash(x)=86104185)
430
+ 3560 train 3.989864 (lr=4.1139e-04) (hash(x)=81993629)
431
+ 3570 train 3.969471 (lr=4.0644e-04) (hash(x)=72719368)
432
+ 3580 train 3.978530 (lr=4.0154e-04) (hash(x)=90091487)
433
+ 3590 train 3.865051 (lr=3.9670e-04) (hash(x)=77393152)
434
+ 3600 val loss 3.9702
435
+ 3600 val perplexity 52.9961
436
+ 3600 train 3.975209 (lr=3.9190e-04) (hash(x)=73103504)
437
+ 3610 train 3.977831 (lr=3.8716e-04) (hash(x)=84429400)
438
+ 3620 train 3.953152 (lr=3.8247e-04) (hash(x)=77021795)
439
+ 3630 train 3.907960 (lr=3.7783e-04) (hash(x)=92830605)
440
+ 3640 train 4.103097 (lr=3.7325e-04) (hash(x)=78313175)
441
+ 3650 train 3.900068 (lr=3.6872e-04) (hash(x)=85395549)
442
+ 3660 train 3.694602 (lr=3.6424e-04) (hash(x)=78114459)
443
+ 3670 train 3.823733 (lr=3.5982e-04) (hash(x)=74968316)
444
+ 3680 train 3.751409 (lr=3.5545e-04) (hash(x)=73358737)
445
+ 3690 train 3.850452 (lr=3.5114e-04) (hash(x)=76399442)
446
+ 3700 val loss 3.9721
447
+ 3700 val perplexity 53.0935
448
+ 3700 train 3.859210 (lr=3.4688e-04) (hash(x)=74039273)
449
+ 3710 train 4.038212 (lr=3.4268e-04) (hash(x)=79299680)
450
+ 3720 train 3.943767 (lr=3.3854e-04) (hash(x)=94701498)
451
+ 3730 train 4.029913 (lr=3.3445e-04) (hash(x)=75352071)
452
+ 3740 train 3.905077 (lr=3.3041e-04) (hash(x)=77909487)
453
+ 3750 train 4.009928 (lr=3.2644e-04) (hash(x)=73986730)
454
+ 3760 train 3.963302 (lr=3.2252e-04) (hash(x)=79325763)
455
+ 3770 train 4.015776 (lr=3.1865e-04) (hash(x)=72457818)
456
+ 3780 train 4.149921 (lr=3.1485e-04) (hash(x)=71775590)
457
+ 3790 train 4.036805 (lr=3.1110e-04) (hash(x)=82638943)
458
+ 3800 val loss 3.9552
459
+ 3800 val perplexity 52.2058
460
+ 3800 train 4.009172 (lr=3.0741e-04) (hash(x)=79965893)
461
+ 3810 train 3.945343 (lr=3.0378e-04) (hash(x)=72598235)
462
+ 3820 train 3.926299 (lr=3.0021e-04) (hash(x)=83113889)
463
+ 3830 train 3.985319 (lr=2.9670e-04) (hash(x)=74434590)
464
+ 3840 train 3.903815 (lr=2.9324e-04) (hash(x)=82860348)
465
+ 3850 train 3.843144 (lr=2.8985e-04) (hash(x)=78067565)
466
+ 3860 train 3.909347 (lr=2.8651e-04) (hash(x)=82592498)
467
+ 3870 train 3.841512 (lr=2.8324e-04) (hash(x)=81820733)
468
+ 3880 train 3.858000 (lr=2.8002e-04) (hash(x)=87709040)
469
+ 3890 train 3.898998 (lr=2.7687e-04) (hash(x)=70379093)
470
+ 3900 val loss 3.9491
471
+ 3900 val perplexity 51.8884
472
+ 3900 train 4.025016 (lr=2.7378e-04) (hash(x)=76597431)
473
+ 3910 train 3.910314 (lr=2.7074e-04) (hash(x)=90490716)
474
+ 3920 train 4.014737 (lr=2.6777e-04) (hash(x)=81970659)
475
+ 3930 train 3.995528 (lr=2.6486e-04) (hash(x)=81496334)
476
+ 3940 train 3.971812 (lr=2.6201e-04) (hash(x)=75717605)
477
+ 3950 train 3.920965 (lr=2.5923e-04) (hash(x)=82749357)
478
+ 3960 train 3.948948 (lr=2.5650e-04) (hash(x)=89355157)
479
+ 3970 train 3.907092 (lr=2.5384e-04) (hash(x)=78980403)
480
+ 3980 train 3.879369 (lr=2.5124e-04) (hash(x)=76627217)
481
+ 3990 train 3.882000 (lr=2.4870e-04) (hash(x)=72412879)
482
+ 4000 val loss 3.9439
483
+ 4000 val perplexity 51.6200
484
+ 4000 train 3.892717 (lr=2.4622e-04) (hash(x)=83018142)
485
+ 4010 train 3.909440 (lr=2.4381e-04) (hash(x)=81272436)
486
+ 4020 train 3.838274 (lr=2.4146e-04) (hash(x)=85497482)
487
+ 4030 train 3.908234 (lr=2.3918e-04) (hash(x)=85241734)
488
+ 4040 train 3.873814 (lr=2.3695e-04) (hash(x)=77925307)
489
+ 4050 train 3.875640 (lr=2.3479e-04) (hash(x)=84826179)
490
+ 4060 train 3.935628 (lr=2.3270e-04) (hash(x)=83606764)
491
+ 4070 train 3.844050 (lr=2.3067e-04) (hash(x)=80567590)
492
+ 4080 train 3.899173 (lr=2.2870e-04) (hash(x)=76860998)
493
+ 4090 train 3.932312 (lr=2.2680e-04) (hash(x)=74902328)
494
+ 4100 val loss 3.9328
495
+ 4100 val perplexity 51.0477
496
+ 4100 train 3.979079 (lr=2.2496e-04) (hash(x)=82832041)
497
+ 4110 train 3.859890 (lr=2.2318e-04) (hash(x)=79143262)
498
+ 4120 train 3.879714 (lr=2.2147e-04) (hash(x)=77038149)
499
+ 4130 train 4.065175 (lr=2.1983e-04) (hash(x)=86339074)
500
+ 4140 train 4.135539 (lr=2.1825e-04) (hash(x)=76686216)
501
+ 4150 train 3.908681 (lr=2.1673e-04) (hash(x)=70522682)
502
+ 4160 train 3.870486 (lr=2.1528e-04) (hash(x)=90958555)
503
+ 4170 train 3.894336 (lr=2.1390e-04) (hash(x)=91463532)
504
+ 4180 train 3.884353 (lr=2.1258e-04) (hash(x)=81959329)
505
+ 4190 train 3.843055 (lr=2.1132e-04) (hash(x)=83146752)
506
+ 4200 val loss 3.9373
507
+ 4200 val perplexity 51.2792
508
+ 4200 train 3.831522 (lr=2.1013e-04) (hash(x)=78361715)
509
+ 4210 train 3.976437 (lr=2.0901e-04) (hash(x)=87364889)
510
+ 4220 train 3.825133 (lr=2.0795e-04) (hash(x)=70465156)
511
+ 4230 train 3.984936 (lr=2.0696e-04) (hash(x)=84524081)
512
+ 4240 train 3.938857 (lr=2.0604e-04) (hash(x)=77824868)
513
+ 4250 train 3.951124 (lr=2.0518e-04) (hash(x)=81710711)
514
+ 4260 train 3.930253 (lr=2.0438e-04) (hash(x)=76362728)
515
+ 4270 train 3.922884 (lr=2.0365e-04) (hash(x)=83115208)
516
+ 4280 train 3.839184 (lr=2.0299e-04) (hash(x)=87218314)
517
+ 4290 train 4.000162 (lr=2.0239e-04) (hash(x)=74582673)
518
+ 4300 val loss 3.9298
519
+ 4300 val perplexity 50.8975
520
+ 4300 train 3.869942 (lr=2.0186e-04) (hash(x)=77379615)
521
+ 4310 train 3.847957 (lr=2.0140e-04) (hash(x)=78669579)
522
+ 4320 train 3.963618 (lr=2.0100e-04) (hash(x)=83066608)
523
+ 4330 train 3.925366 (lr=2.0067e-04) (hash(x)=83037340)
524
+ 4340 train 3.964192 (lr=2.0041e-04) (hash(x)=82849771)
525
+ 4350 train 3.903562 (lr=2.0021e-04) (hash(x)=76693985)
526
+ 4360 train 3.990994 (lr=2.0007e-04) (hash(x)=77745394)
527
+ 4370 train 3.952344 (lr=2.0001e-04) (hash(x)=79954388)
528
+ 4374 val loss 3.9237
529
+ 4374 val perplexity 50.5891
12_head_baseline_lr_20e-4_head_dim_22_fixed/model_04374.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:00ae6c3f3fa1627c4a8f1fdb2e05767ac2ceed9d9e7236d5542f0e44e66c0313
3
+ size 96858242
12_head_baseline_lr_20e-4_head_dim_22_fixed/optimizer_04374.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:f1976acf5118fc88a50db290492b4a73158894461d6764293991a0e618ca490e
3
+ size 187435910