andrew-healey commited on
Commit
3d3c68e
·
verified ·
1 Parent(s): 93407ea

Upload folder using huggingface_hub

Browse files
one_mask_per_head_2_latent_masks/args.json ADDED
@@ -0,0 +1 @@
 
 
1
+ {"hellaswag": true, "attention_kind": "selective", "log_dir": "logs/repro_selective_pattern_rankings/one_mask_per_head_2_latent_masks", "resume_checkpoint": null, "resume_optimizer": false, "add_a_head": false, "add_head_to_start": true, "new_head_init": "normal", "n_heads": 4, "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": 128, "total_batch_size": 131072, "protection_head_scaling_factor": 1.0, "protection_head_bias": 0.0, "n_sliced_masks": null, "n_latent_masks": 2, "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": 0.0011, "decay_lr": true, "readout_zero_init": false, "query_zero_init": false, "l1_loss": false, "debugpy": false, "key": "one_mask_per_head_2_latent_masks", "n_embd": 256}
one_mask_per_head_2_latent_masks/dataloader_04374.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:1e9422c936c4d62660261ba5451b4b72bd5055f233420bd7b0797ede3ccedd1e
3
+ size 964
one_mask_per_head_2_latent_masks/log2.txt ADDED
@@ -0,0 +1,529 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ max_steps: 4375
2
+ 0 val loss 11.2299
3
+ 0 val perplexity 75350.7344
4
+ 0 train 11.229395 (lr=1.5385e-06) (hash(x)=161979327)
5
+ 10 train 10.292687 (lr=1.6923e-05) (hash(x)=165941070)
6
+ 20 train 9.862397 (lr=3.2308e-05) (hash(x)=152918033)
7
+ 30 train 9.578296 (lr=4.7692e-05) (hash(x)=169931607)
8
+ 40 train 9.100039 (lr=6.3077e-05) (hash(x)=174342675)
9
+ 50 train 8.643604 (lr=7.8462e-05) (hash(x)=163872554)
10
+ 60 train 8.254272 (lr=9.3846e-05) (hash(x)=161534908)
11
+ 70 train 7.744008 (lr=1.0923e-04) (hash(x)=156776712)
12
+ 80 train 7.665780 (lr=1.2462e-04) (hash(x)=164996929)
13
+ 90 train 7.534584 (lr=1.4000e-04) (hash(x)=152962169)
14
+ 100 val loss 7.4798
15
+ 100 val perplexity 1771.9442
16
+ 100 train 7.339922 (lr=1.5538e-04) (hash(x)=147111701)
17
+ 110 train 7.293898 (lr=1.7077e-04) (hash(x)=167595270)
18
+ 120 train 7.162320 (lr=1.8615e-04) (hash(x)=152205426)
19
+ 130 train 7.011769 (lr=2.0154e-04) (hash(x)=158427335)
20
+ 140 train 6.972773 (lr=2.1692e-04) (hash(x)=174588384)
21
+ 150 train 7.037151 (lr=2.3231e-04) (hash(x)=157067270)
22
+ 160 train 6.919617 (lr=2.4769e-04) (hash(x)=161859595)
23
+ 170 train 6.813656 (lr=2.6308e-04) (hash(x)=162739537)
24
+ 180 train 6.739655 (lr=2.7846e-04) (hash(x)=170785304)
25
+ 190 train 6.634585 (lr=2.9385e-04) (hash(x)=168658088)
26
+ 200 val loss 6.5892
27
+ 200 val perplexity 727.2336
28
+ 200 train 6.590330 (lr=3.0923e-04) (hash(x)=150749086)
29
+ 210 train 6.456457 (lr=3.2462e-04) (hash(x)=148789576)
30
+ 220 train 6.374694 (lr=3.4000e-04) (hash(x)=153164408)
31
+ 230 train 6.550076 (lr=3.5538e-04) (hash(x)=184594366)
32
+ 240 train 6.265978 (lr=3.7077e-04) (hash(x)=157817319)
33
+ 250 train 6.277492 (lr=3.8615e-04) (hash(x)=149451562)
34
+ 260 train 6.189072 (lr=4.0154e-04) (hash(x)=165738161)
35
+ 270 train 6.123448 (lr=4.1692e-04) (hash(x)=165724764)
36
+ 280 train 6.060798 (lr=4.3231e-04) (hash(x)=155414136)
37
+ 290 train 5.983246 (lr=4.4769e-04) (hash(x)=165090903)
38
+ 300 val loss 6.1278
39
+ 300 val perplexity 458.4379
40
+ 300 train 5.850461 (lr=4.6308e-04) (hash(x)=154445545)
41
+ 310 train 5.792864 (lr=4.7846e-04) (hash(x)=158949603)
42
+ 320 train 5.874228 (lr=4.9385e-04) (hash(x)=161412973)
43
+ 330 train 5.979318 (lr=5.0923e-04) (hash(x)=164572249)
44
+ 340 train 5.990306 (lr=5.2462e-04) (hash(x)=162202191)
45
+ 350 train 5.965785 (lr=5.4000e-04) (hash(x)=158135392)
46
+ 360 train 5.916399 (lr=5.5538e-04) (hash(x)=163752391)
47
+ 370 train 5.915674 (lr=5.7077e-04) (hash(x)=172758799)
48
+ 380 train 5.688145 (lr=5.8615e-04) (hash(x)=144569488)
49
+ 390 train 5.683249 (lr=6.0154e-04) (hash(x)=150429229)
50
+ 400 val loss 5.7391
51
+ 400 val perplexity 310.7847
52
+ 400 train 5.683249 (lr=6.1692e-04) (hash(x)=154261530)
53
+ 410 train 5.602164 (lr=6.3231e-04) (hash(x)=159172506)
54
+ 420 train 5.616530 (lr=6.4769e-04) (hash(x)=154419867)
55
+ 430 train 5.643787 (lr=6.6308e-04) (hash(x)=167337099)
56
+ 440 train 5.530875 (lr=6.7846e-04) (hash(x)=154535255)
57
+ 450 train 5.498389 (lr=6.9385e-04) (hash(x)=153077071)
58
+ 460 train 5.430981 (lr=7.0923e-04) (hash(x)=150853070)
59
+ 470 train 5.246686 (lr=7.2462e-04) (hash(x)=160701113)
60
+ 480 train 5.378659 (lr=7.4000e-04) (hash(x)=154690870)
61
+ 490 train 5.193591 (lr=7.5538e-04) (hash(x)=163834286)
62
+ 500 val loss 5.5172
63
+ 500 val perplexity 248.9455
64
+ 500 train 5.270822 (lr=7.7077e-04) (hash(x)=162651717)
65
+ 510 train 5.355933 (lr=7.8615e-04) (hash(x)=160448311)
66
+ 520 train 5.404992 (lr=8.0154e-04) (hash(x)=165741643)
67
+ 530 train 5.507198 (lr=8.1692e-04) (hash(x)=153201264)
68
+ 540 train 5.414214 (lr=8.3231e-04) (hash(x)=159027160)
69
+ 550 train 5.402275 (lr=8.4769e-04) (hash(x)=170769095)
70
+ 560 train 5.372064 (lr=8.6308e-04) (hash(x)=153606190)
71
+ 570 train 5.358816 (lr=8.7846e-04) (hash(x)=159708689)
72
+ 580 train 5.310168 (lr=8.9385e-04) (hash(x)=179108359)
73
+ 590 train 5.305269 (lr=9.0923e-04) (hash(x)=157134241)
74
+ 600 val loss 5.3106
75
+ 600 val perplexity 202.4816
76
+ 600 train 5.287784 (lr=9.2462e-04) (hash(x)=162178356)
77
+ 610 train 5.199407 (lr=9.4000e-04) (hash(x)=167646928)
78
+ 620 train 5.171767 (lr=9.5538e-04) (hash(x)=160979601)
79
+ 630 train 5.075449 (lr=9.7077e-04) (hash(x)=158349468)
80
+ 640 train 5.085411 (lr=9.8615e-04) (hash(x)=150370348)
81
+ 650 train 5.042173 (lr=1.0015e-03) (hash(x)=152810075)
82
+ 660 train 4.953032 (lr=1.0169e-03) (hash(x)=145786262)
83
+ 670 train 4.952306 (lr=1.0323e-03) (hash(x)=160779604)
84
+ 680 train 4.870088 (lr=1.0477e-03) (hash(x)=152343688)
85
+ 690 train 4.901697 (lr=1.0631e-03) (hash(x)=166507600)
86
+ 700 val loss 5.1852
87
+ 700 val perplexity 178.6020
88
+ 700 train 4.870753 (lr=1.0785e-03) (hash(x)=162623700)
89
+ 710 train 5.141260 (lr=1.0938e-03) (hash(x)=162293403)
90
+ 720 train 5.110381 (lr=1.1000e-03) (hash(x)=159564657)
91
+ 730 train 5.113923 (lr=1.1000e-03) (hash(x)=154209107)
92
+ 740 train 5.084786 (lr=1.0999e-03) (hash(x)=156485709)
93
+ 750 train 5.054815 (lr=1.0998e-03) (hash(x)=160834702)
94
+ 760 train 5.053354 (lr=1.0996e-03) (hash(x)=152240275)
95
+ 770 train 5.040745 (lr=1.0994e-03) (hash(x)=168903075)
96
+ 780 train 5.093730 (lr=1.0992e-03) (hash(x)=166843738)
97
+ 790 train 5.130337 (lr=1.0990e-03) (hash(x)=166358104)
98
+ 800 val loss 5.0086
99
+ 800 val perplexity 149.6900
100
+ 800 train 4.940433 (lr=1.0987e-03) (hash(x)=153477747)
101
+ 810 train 4.864687 (lr=1.0984e-03) (hash(x)=157954385)
102
+ 820 train 4.866491 (lr=1.0980e-03) (hash(x)=174383505)
103
+ 830 train 4.884716 (lr=1.0976e-03) (hash(x)=145226181)
104
+ 840 train 4.865328 (lr=1.0972e-03) (hash(x)=157665743)
105
+ 850 train 4.851415 (lr=1.0967e-03) (hash(x)=158565836)
106
+ 860 train 4.852178 (lr=1.0962e-03) (hash(x)=171168105)
107
+ 870 train 5.256691 (lr=1.0956e-03) (hash(x)=159017777)
108
+ 880 train 5.024911 (lr=1.0950e-03) (hash(x)=168156768)
109
+ 890 train 4.930644 (lr=1.0944e-03) (hash(x)=145645245)
110
+ 900 val loss 4.8546
111
+ 900 val perplexity 128.3249
112
+ 900 train 4.735715 (lr=1.0938e-03) (hash(x)=151245521)
113
+ 910 train 4.767969 (lr=1.0931e-03) (hash(x)=182614672)
114
+ 920 train 4.756940 (lr=1.0924e-03) (hash(x)=167377432)
115
+ 930 train 4.673415 (lr=1.0916e-03) (hash(x)=171347228)
116
+ 940 train 4.701856 (lr=1.0908e-03) (hash(x)=171588755)
117
+ 950 train 4.784557 (lr=1.0900e-03) (hash(x)=159560168)
118
+ 960 train 4.668504 (lr=1.0891e-03) (hash(x)=171801357)
119
+ 970 train 4.670782 (lr=1.0882e-03) (hash(x)=172513884)
120
+ 980 train 4.692821 (lr=1.0872e-03) (hash(x)=170659685)
121
+ 990 train 4.712802 (lr=1.0863e-03) (hash(x)=165436358)
122
+ 1000 val loss 4.7261
123
+ 1000 val perplexity 112.8574
124
+ 1000 train 4.778811 (lr=1.0853e-03) (hash(x)=167684690)
125
+ 1010 train 4.745741 (lr=1.0842e-03) (hash(x)=127225492)
126
+ 1020 train 4.741680 (lr=1.0831e-03) (hash(x)=167575323)
127
+ 1030 train 4.663941 (lr=1.0820e-03) (hash(x)=161266302)
128
+ 1040 train 4.637856 (lr=1.0809e-03) (hash(x)=155607661)
129
+ 1050 train 4.679067 (lr=1.0797e-03) (hash(x)=165685359)
130
+ 1060 train 4.539914 (lr=1.0785e-03) (hash(x)=160417824)
131
+ 1070 train 4.677355 (lr=1.0772e-03) (hash(x)=173313882)
132
+ 1080 train 4.639106 (lr=1.0759e-03) (hash(x)=165485047)
133
+ 1090 train 4.499640 (lr=1.0746e-03) (hash(x)=159848305)
134
+ 1100 val loss 4.6453
135
+ 1100 val perplexity 104.0981
136
+ 1100 train 4.666270 (lr=1.0732e-03) (hash(x)=170890601)
137
+ 1110 train 4.586980 (lr=1.0718e-03) (hash(x)=172689983)
138
+ 1120 train 4.618576 (lr=1.0704e-03) (hash(x)=154762859)
139
+ 1130 train 4.540084 (lr=1.0689e-03) (hash(x)=168196871)
140
+ 1140 train 4.574792 (lr=1.0674e-03) (hash(x)=156715577)
141
+ 1150 train 4.695165 (lr=1.0659e-03) (hash(x)=166381788)
142
+ 1160 train 4.624650 (lr=1.0643e-03) (hash(x)=162095130)
143
+ 1170 train 4.606727 (lr=1.0627e-03) (hash(x)=166094750)
144
+ 1180 train 4.537101 (lr=1.0611e-03) (hash(x)=154965753)
145
+ 1190 train 4.519797 (lr=1.0594e-03) (hash(x)=172766447)
146
+ 1200 val loss 4.5657
147
+ 1200 val perplexity 96.1325
148
+ 1200 train 4.438736 (lr=1.0577e-03) (hash(x)=161185334)
149
+ 1210 train 4.621973 (lr=1.0560e-03) (hash(x)=167224144)
150
+ 1220 train 4.449998 (lr=1.0542e-03) (hash(x)=167252866)
151
+ 1230 train 4.457823 (lr=1.0524e-03) (hash(x)=155856924)
152
+ 1240 train 4.450935 (lr=1.0506e-03) (hash(x)=155956446)
153
+ 1250 train 4.363693 (lr=1.0487e-03) (hash(x)=140794796)
154
+ 1260 train 4.474868 (lr=1.0468e-03) (hash(x)=166863869)
155
+ 1270 train 4.538322 (lr=1.0449e-03) (hash(x)=166021892)
156
+ 1280 train 4.546858 (lr=1.0429e-03) (hash(x)=170503648)
157
+ 1290 train 4.461845 (lr=1.0409e-03) (hash(x)=164012719)
158
+ 1300 val loss 4.5231
159
+ 1300 val perplexity 92.1231
160
+ 1300 train 4.529904 (lr=1.0389e-03) (hash(x)=156519834)
161
+ 1310 train 4.473710 (lr=1.0368e-03) (hash(x)=165146652)
162
+ 1320 train 4.496887 (lr=1.0347e-03) (hash(x)=174391990)
163
+ 1330 train 4.502999 (lr=1.0326e-03) (hash(x)=161131201)
164
+ 1340 train 4.474665 (lr=1.0305e-03) (hash(x)=162240912)
165
+ 1350 train 4.393739 (lr=1.0283e-03) (hash(x)=152462931)
166
+ 1360 train 4.403590 (lr=1.0261e-03) (hash(x)=150443278)
167
+ 1370 train 4.337045 (lr=1.0238e-03) (hash(x)=153721711)
168
+ 1380 train 4.483485 (lr=1.0215e-03) (hash(x)=158070628)
169
+ 1390 train 4.504027 (lr=1.0192e-03) (hash(x)=163318827)
170
+ 1400 val loss 4.4887
171
+ 1400 val perplexity 89.0075
172
+ 1400 train 4.411917 (lr=1.0169e-03) (hash(x)=154210336)
173
+ 1410 train 4.456656 (lr=1.0145e-03) (hash(x)=162393933)
174
+ 1420 train 4.468651 (lr=1.0121e-03) (hash(x)=156659308)
175
+ 1430 train 4.433508 (lr=1.0097e-03) (hash(x)=171243439)
176
+ 1440 train 4.574506 (lr=1.0072e-03) (hash(x)=178459521)
177
+ 1450 train 4.417112 (lr=1.0047e-03) (hash(x)=159166350)
178
+ 1460 train 4.340501 (lr=1.0022e-03) (hash(x)=173554569)
179
+ 1470 train 4.435624 (lr=9.9964e-04) (hash(x)=158362521)
180
+ 1480 train 4.394901 (lr=9.9706e-04) (hash(x)=165899207)
181
+ 1490 train 4.428458 (lr=9.9445e-04) (hash(x)=156470939)
182
+ 1500 val loss 4.4378
183
+ 1500 val perplexity 84.5896
184
+ 1500 train 4.400634 (lr=9.9182e-04) (hash(x)=164197780)
185
+ 1510 train 4.292336 (lr=9.8915e-04) (hash(x)=146709896)
186
+ 1520 train 4.306647 (lr=9.8646e-04) (hash(x)=151304838)
187
+ 1530 train 4.348554 (lr=9.8374e-04) (hash(x)=163187333)
188
+ 1540 train 4.303465 (lr=9.8099e-04) (hash(x)=153630537)
189
+ 1550 train 4.311085 (lr=9.7821e-04) (hash(x)=157299103)
190
+ 1560 train 4.223326 (lr=9.7540e-04) (hash(x)=153786477)
191
+ 1570 train 4.548748 (lr=9.7257e-04) (hash(x)=155037209)
192
+ 1580 train 4.422492 (lr=9.6971e-04) (hash(x)=146157308)
193
+ 1590 train 4.463831 (lr=9.6683e-04) (hash(x)=147823317)
194
+ 1600 val loss 4.4094
195
+ 1600 val perplexity 82.2161
196
+ 1600 train 4.446334 (lr=9.6391e-04) (hash(x)=169140159)
197
+ 1610 train 4.235511 (lr=9.6097e-04) (hash(x)=157811493)
198
+ 1620 train 4.329185 (lr=9.5801e-04) (hash(x)=168940851)
199
+ 1630 train 4.312659 (lr=9.5502e-04) (hash(x)=151149213)
200
+ 1640 train 4.411238 (lr=9.5200e-04) (hash(x)=161822009)
201
+ 1650 train 4.823948 (lr=9.4896e-04) (hash(x)=151056313)
202
+ 1660 train 4.334769 (lr=9.4589e-04) (hash(x)=156920306)
203
+ 1670 train 4.247509 (lr=9.4280e-04) (hash(x)=159695815)
204
+ 1680 train 4.261425 (lr=9.3968e-04) (hash(x)=160232907)
205
+ 1690 train 4.252219 (lr=9.3654e-04) (hash(x)=158099357)
206
+ 1700 val loss 4.4067
207
+ 1700 val perplexity 81.9955
208
+ 1700 train 4.308255 (lr=9.3337e-04) (hash(x)=164127381)
209
+ 1710 train 4.479624 (lr=9.3018e-04) (hash(x)=156592040)
210
+ 1720 train 4.404734 (lr=9.2696e-04) (hash(x)=160220265)
211
+ 1730 train 4.345665 (lr=9.2372e-04) (hash(x)=156162204)
212
+ 1740 train 4.335474 (lr=9.2046e-04) (hash(x)=152767893)
213
+ 1750 train 4.320283 (lr=9.1717e-04) (hash(x)=162707306)
214
+ 1760 train 4.341174 (lr=9.1386e-04) (hash(x)=154472004)
215
+ 1770 train 4.199225 (lr=9.1053e-04) (hash(x)=160374234)
216
+ 1780 train 4.355536 (lr=9.0718e-04) (hash(x)=159709225)
217
+ 1790 train 4.387345 (lr=9.0380e-04) (hash(x)=152095944)
218
+ 1800 val loss 4.3564
219
+ 1800 val perplexity 77.9739
220
+ 1800 train 4.367451 (lr=9.0040e-04) (hash(x)=157920255)
221
+ 1810 train 4.116975 (lr=8.9698e-04) (hash(x)=167237680)
222
+ 1820 train 4.225972 (lr=8.9354e-04) (hash(x)=161255774)
223
+ 1830 train 4.248567 (lr=8.9008e-04) (hash(x)=154275481)
224
+ 1840 train 4.199273 (lr=8.8660e-04) (hash(x)=152398178)
225
+ 1850 train 4.329013 (lr=8.8309e-04) (hash(x)=147646022)
226
+ 1860 train 4.330511 (lr=8.7957e-04) (hash(x)=158895559)
227
+ 1870 train 4.267236 (lr=8.7602e-04) (hash(x)=162742598)
228
+ 1880 train 4.263573 (lr=8.7245e-04) (hash(x)=169911029)
229
+ 1890 train 4.298686 (lr=8.6887e-04) (hash(x)=172068270)
230
+ 1900 val loss 4.3563
231
+ 1900 val perplexity 77.9674
232
+ 1900 train 4.391531 (lr=8.6527e-04) (hash(x)=159975457)
233
+ 1910 train 4.367131 (lr=8.6164e-04) (hash(x)=178775667)
234
+ 1920 train 4.244201 (lr=8.5800e-04) (hash(x)=151180831)
235
+ 1930 train 4.268777 (lr=8.5434e-04) (hash(x)=154239412)
236
+ 1940 train 4.273088 (lr=8.5066e-04) (hash(x)=139144465)
237
+ 1950 train 4.231500 (lr=8.4696e-04) (hash(x)=194482413)
238
+ 1960 train 4.253242 (lr=8.4324e-04) (hash(x)=155812344)
239
+ 1970 train 4.251190 (lr=8.3951e-04) (hash(x)=157577031)
240
+ 1980 train 4.281312 (lr=8.3576e-04) (hash(x)=152045027)
241
+ 1990 train 4.186634 (lr=8.3199e-04) (hash(x)=141742221)
242
+ 2000 val loss 4.3079
243
+ 2000 val perplexity 74.2841
244
+ 2000 train 4.316084 (lr=8.2821e-04) (hash(x)=160555812)
245
+ 2010 train 4.231083 (lr=8.2441e-04) (hash(x)=160371335)
246
+ 2020 train 4.241691 (lr=8.2059e-04) (hash(x)=167890767)
247
+ 2030 train 4.309393 (lr=8.1676e-04) (hash(x)=165127715)
248
+ 2040 train 4.250916 (lr=8.1291e-04) (hash(x)=151587302)
249
+ 2050 train 4.272454 (lr=8.0905e-04) (hash(x)=172347989)
250
+ 2060 train 4.350060 (lr=8.0517e-04) (hash(x)=161301919)
251
+ 2070 train 4.285516 (lr=8.0128e-04) (hash(x)=161472934)
252
+ 2080 train 4.366663 (lr=7.9737e-04) (hash(x)=159777322)
253
+ 2090 train 4.278683 (lr=7.9345e-04) (hash(x)=151149770)
254
+ 2100 val loss 4.2812
255
+ 2100 val perplexity 72.3285
256
+ 2100 train 4.108731 (lr=7.8951e-04) (hash(x)=160283520)
257
+ 2110 train 4.282444 (lr=7.8556e-04) (hash(x)=154222025)
258
+ 2120 train 4.222589 (lr=7.8160e-04) (hash(x)=155638178)
259
+ 2130 train 4.139166 (lr=7.7762e-04) (hash(x)=149125384)
260
+ 2140 train 4.189806 (lr=7.7363e-04) (hash(x)=159496074)
261
+ 2150 train 4.279702 (lr=7.6963e-04) (hash(x)=162019154)
262
+ 2160 train 4.427162 (lr=7.6562e-04) (hash(x)=155466928)
263
+ 2170 train 4.308986 (lr=7.6160e-04) (hash(x)=173469994)
264
+ 2180 train 4.415012 (lr=7.5756e-04) (hash(x)=163448753)
265
+ 2190 train 4.250129 (lr=7.5351e-04) (hash(x)=166595123)
266
+ 2200 val loss 4.2615
267
+ 2200 val perplexity 70.9132
268
+ 2200 train 4.131884 (lr=7.4945e-04) (hash(x)=165958875)
269
+ 2210 train 4.265624 (lr=7.4538e-04) (hash(x)=165019039)
270
+ 2220 train 4.251698 (lr=7.4130e-04) (hash(x)=167325225)
271
+ 2230 train 4.084932 (lr=7.3721e-04) (hash(x)=154287439)
272
+ 2240 train 4.149313 (lr=7.3312e-04) (hash(x)=158565817)
273
+ 2250 train 4.116706 (lr=7.2901e-04) (hash(x)=160592835)
274
+ 2260 train 4.156415 (lr=7.2489e-04) (hash(x)=144591648)
275
+ 2270 train 4.029102 (lr=7.2076e-04) (hash(x)=150486876)
276
+ 2280 train 4.228423 (lr=7.1663e-04) (hash(x)=159194788)
277
+ 2290 train 4.302792 (lr=7.1248e-04) (hash(x)=164315345)
278
+ 2300 val loss 4.2362
279
+ 2300 val perplexity 69.1447
280
+ 2300 train 4.230641 (lr=7.0833e-04) (hash(x)=170470665)
281
+ 2310 train 4.193524 (lr=7.0417e-04) (hash(x)=176539068)
282
+ 2320 train 4.174839 (lr=7.0001e-04) (hash(x)=161008474)
283
+ 2330 train 4.245769 (lr=6.9583e-04) (hash(x)=168474380)
284
+ 2340 train 4.243580 (lr=6.9165e-04) (hash(x)=156072131)
285
+ 2350 train 4.332134 (lr=6.8747e-04) (hash(x)=165127087)
286
+ 2360 train 4.328112 (lr=6.8327e-04) (hash(x)=159462801)
287
+ 2370 train 4.162747 (lr=6.7908e-04) (hash(x)=161698394)
288
+ 2380 train 4.088000 (lr=6.7487e-04) (hash(x)=165588078)
289
+ 2390 train 4.095147 (lr=6.7066e-04) (hash(x)=171673672)
290
+ 2400 val loss 4.2313
291
+ 2400 val perplexity 68.8081
292
+ 2400 train 4.156996 (lr=6.6645e-04) (hash(x)=157381516)
293
+ 2410 train 4.167273 (lr=6.6223e-04) (hash(x)=163238532)
294
+ 2420 train 4.117593 (lr=6.5801e-04) (hash(x)=156178237)
295
+ 2430 train 4.001331 (lr=6.5378e-04) (hash(x)=155080617)
296
+ 2440 train 4.115452 (lr=6.4955e-04) (hash(x)=167490538)
297
+ 2450 train 4.087460 (lr=6.4532e-04) (hash(x)=163757352)
298
+ 2460 train 4.044192 (lr=6.4108e-04) (hash(x)=156734811)
299
+ 2470 train 4.036159 (lr=6.3684e-04) (hash(x)=161522341)
300
+ 2480 train 4.295784 (lr=6.3260e-04) (hash(x)=174032060)
301
+ 2490 train 4.256676 (lr=6.2836e-04) (hash(x)=163065818)
302
+ 2500 val loss 4.2011
303
+ 2500 val perplexity 66.7628
304
+ 2500 train 4.312733 (lr=6.2412e-04) (hash(x)=126959780)
305
+ 2510 train 4.239351 (lr=6.1987e-04) (hash(x)=153844939)
306
+ 2520 train 4.220837 (lr=6.1562e-04) (hash(x)=174018174)
307
+ 2530 train 4.196819 (lr=6.1137e-04) (hash(x)=150568427)
308
+ 2540 train 4.194892 (lr=6.0712e-04) (hash(x)=161725206)
309
+ 2550 train 4.165215 (lr=6.0288e-04) (hash(x)=181346417)
310
+ 2560 train 4.149391 (lr=5.9863e-04) (hash(x)=168730011)
311
+ 2570 train 4.040175 (lr=5.9438e-04) (hash(x)=155900214)
312
+ 2580 train 4.099322 (lr=5.9013e-04) (hash(x)=160014621)
313
+ 2590 train 4.066610 (lr=5.8588e-04) (hash(x)=148996689)
314
+ 2600 val loss 4.1880
315
+ 2600 val perplexity 65.8878
316
+ 2600 train 4.146353 (lr=5.8164e-04) (hash(x)=145183968)
317
+ 2610 train 3.892092 (lr=5.7740e-04) (hash(x)=159420240)
318
+ 2620 train 3.871750 (lr=5.7316e-04) (hash(x)=151169903)
319
+ 2630 train 3.883908 (lr=5.6892e-04) (hash(x)=149454811)
320
+ 2640 train 3.952943 (lr=5.6468e-04) (hash(x)=155603914)
321
+ 2650 train 3.949025 (lr=5.6045e-04) (hash(x)=169545195)
322
+ 2660 train 4.289502 (lr=5.5622e-04) (hash(x)=168454079)
323
+ 2670 train 4.188440 (lr=5.5199e-04) (hash(x)=163110417)
324
+ 2680 train 4.136687 (lr=5.4777e-04) (hash(x)=166075603)
325
+ 2690 train 4.171264 (lr=5.4355e-04) (hash(x)=174132258)
326
+ 2700 val loss 4.1766
327
+ 2700 val perplexity 65.1440
328
+ 2700 train 4.174453 (lr=5.3934e-04) (hash(x)=169846940)
329
+ 2710 train 4.076893 (lr=5.3513e-04) (hash(x)=151553731)
330
+ 2720 train 4.176074 (lr=5.3092e-04) (hash(x)=164024255)
331
+ 2730 train 4.193578 (lr=5.2673e-04) (hash(x)=156458097)
332
+ 2740 train 4.122664 (lr=5.2253e-04) (hash(x)=154663748)
333
+ 2750 train 3.989403 (lr=5.1835e-04) (hash(x)=157548181)
334
+ 2760 train 4.100259 (lr=5.1417e-04) (hash(x)=151584783)
335
+ 2770 train 4.123983 (lr=5.0999e-04) (hash(x)=152692594)
336
+ 2780 train 4.099561 (lr=5.0583e-04) (hash(x)=165952919)
337
+ 2790 train 3.987947 (lr=5.0167e-04) (hash(x)=157484183)
338
+ 2800 val loss 4.1617
339
+ 2800 val perplexity 64.1826
340
+ 2800 train 4.069056 (lr=4.9752e-04) (hash(x)=154790034)
341
+ 2810 train 3.925876 (lr=4.9337e-04) (hash(x)=157856097)
342
+ 2820 train 3.912570 (lr=4.8924e-04) (hash(x)=168903939)
343
+ 2830 train 4.031922 (lr=4.8511e-04) (hash(x)=167001475)
344
+ 2840 train 3.863457 (lr=4.8099e-04) (hash(x)=158987150)
345
+ 2850 train 4.092248 (lr=4.7688e-04) (hash(x)=168695407)
346
+ 2860 train 4.224924 (lr=4.7279e-04) (hash(x)=164999034)
347
+ 2870 train 4.075220 (lr=4.6870e-04) (hash(x)=191197316)
348
+ 2880 train 4.034261 (lr=4.6462e-04) (hash(x)=159625200)
349
+ 2890 train 4.192124 (lr=4.6055e-04) (hash(x)=157756126)
350
+ 2900 val loss 4.1350
351
+ 2900 val perplexity 62.4876
352
+ 2900 train 4.046464 (lr=4.5649e-04) (hash(x)=164721377)
353
+ 2910 train 4.097908 (lr=4.5244e-04) (hash(x)=155873181)
354
+ 2920 train 4.040895 (lr=4.4840e-04) (hash(x)=156477082)
355
+ 2930 train 4.051638 (lr=4.4438e-04) (hash(x)=151261633)
356
+ 2940 train 3.992356 (lr=4.4037e-04) (hash(x)=160090385)
357
+ 2950 train 4.133692 (lr=4.3637e-04) (hash(x)=157300618)
358
+ 2960 train 4.044963 (lr=4.3238e-04) (hash(x)=163095555)
359
+ 2970 train 4.082992 (lr=4.2840e-04) (hash(x)=149649212)
360
+ 2980 train 3.893595 (lr=4.2444e-04) (hash(x)=174517675)
361
+ 2990 train 3.855943 (lr=4.2049e-04) (hash(x)=161344491)
362
+ 3000 val loss 4.1360
363
+ 3000 val perplexity 62.5519
364
+ 3000 train 3.887669 (lr=4.1655e-04) (hash(x)=163816905)
365
+ 3010 train 3.796937 (lr=4.1263e-04) (hash(x)=157420680)
366
+ 3020 train 3.903817 (lr=4.0872e-04) (hash(x)=153843035)
367
+ 3030 train 4.092408 (lr=4.0483e-04) (hash(x)=149770373)
368
+ 3040 train 4.162940 (lr=4.0095e-04) (hash(x)=165320741)
369
+ 3050 train 4.135208 (lr=3.9709e-04) (hash(x)=170181151)
370
+ 3060 train 4.058032 (lr=3.9324e-04) (hash(x)=163837230)
371
+ 3070 train 4.049291 (lr=3.8941e-04) (hash(x)=156812797)
372
+ 3080 train 4.023409 (lr=3.8559e-04) (hash(x)=143663627)
373
+ 3090 train 4.050633 (lr=3.8179e-04) (hash(x)=165052478)
374
+ 3100 val loss 4.1138
375
+ 3100 val perplexity 61.1799
376
+ 3100 train 3.988517 (lr=3.7801e-04) (hash(x)=151458705)
377
+ 3110 train 4.033869 (lr=3.7424e-04) (hash(x)=158228317)
378
+ 3120 train 4.208058 (lr=3.7049e-04) (hash(x)=149856801)
379
+ 3130 train 3.877982 (lr=3.6676e-04) (hash(x)=146297847)
380
+ 3140 train 3.977850 (lr=3.6304e-04) (hash(x)=172795167)
381
+ 3150 train 4.032634 (lr=3.5934e-04) (hash(x)=145936074)
382
+ 3160 train 4.117423 (lr=3.5566e-04) (hash(x)=157564303)
383
+ 3170 train 4.213943 (lr=3.5200e-04) (hash(x)=157451362)
384
+ 3180 train 4.133344 (lr=3.4836e-04) (hash(x)=156366043)
385
+ 3190 train 4.094037 (lr=3.4473e-04) (hash(x)=162344697)
386
+ 3200 val loss 4.0967
387
+ 3200 val perplexity 60.1415
388
+ 3200 train 4.119820 (lr=3.4113e-04) (hash(x)=164206298)
389
+ 3210 train 4.183823 (lr=3.3755e-04) (hash(x)=149468251)
390
+ 3220 train 4.119967 (lr=3.3398e-04) (hash(x)=157808606)
391
+ 3230 train 4.256069 (lr=3.3043e-04) (hash(x)=157078690)
392
+ 3240 train 4.060828 (lr=3.2691e-04) (hash(x)=165699696)
393
+ 3250 train 4.009609 (lr=3.2340e-04) (hash(x)=181352842)
394
+ 3260 train 4.109667 (lr=3.1992e-04) (hash(x)=142534383)
395
+ 3270 train 4.065913 (lr=3.1646e-04) (hash(x)=165297359)
396
+ 3280 train 4.107195 (lr=3.1302e-04) (hash(x)=160326815)
397
+ 3290 train 4.027143 (lr=3.0960e-04) (hash(x)=150043063)
398
+ 3300 val loss 4.0922
399
+ 3300 val perplexity 59.8700
400
+ 3300 train 3.987126 (lr=3.0620e-04) (hash(x)=164133091)
401
+ 3310 train 3.931129 (lr=3.0282e-04) (hash(x)=145803863)
402
+ 3320 train 3.951809 (lr=2.9947e-04) (hash(x)=157282009)
403
+ 3330 train 3.897341 (lr=2.9614e-04) (hash(x)=159126415)
404
+ 3340 train 4.096988 (lr=2.9283e-04) (hash(x)=164426675)
405
+ 3350 train 4.126705 (lr=2.8954e-04) (hash(x)=170424924)
406
+ 3360 train 4.071496 (lr=2.8628e-04) (hash(x)=163161621)
407
+ 3370 train 4.130874 (lr=2.8304e-04) (hash(x)=160786731)
408
+ 3380 train 4.090637 (lr=2.7982e-04) (hash(x)=165954688)
409
+ 3390 train 4.140387 (lr=2.7663e-04) (hash(x)=147371107)
410
+ 3400 val loss 4.0727
411
+ 3400 val perplexity 58.7128
412
+ 3400 train 4.083417 (lr=2.7346e-04) (hash(x)=153067604)
413
+ 3410 train 4.077000 (lr=2.7032e-04) (hash(x)=161748196)
414
+ 3420 train 4.088540 (lr=2.6720e-04) (hash(x)=168297963)
415
+ 3430 train 4.080776 (lr=2.6411e-04) (hash(x)=145086892)
416
+ 3440 train 3.988733 (lr=2.6104e-04) (hash(x)=173125711)
417
+ 3450 train 4.091846 (lr=2.5800e-04) (hash(x)=158080216)
418
+ 3460 train 3.909902 (lr=2.5498e-04) (hash(x)=158055750)
419
+ 3470 train 4.008200 (lr=2.5199e-04) (hash(x)=151381400)
420
+ 3480 train 3.948475 (lr=2.4903e-04) (hash(x)=147262346)
421
+ 3490 train 3.927995 (lr=2.4609e-04) (hash(x)=154740086)
422
+ 3500 val loss 4.0700
423
+ 3500 val perplexity 58.5564
424
+ 3500 train 3.847829 (lr=2.4317e-04) (hash(x)=145730968)
425
+ 3510 train 3.994682 (lr=2.4029e-04) (hash(x)=151355929)
426
+ 3520 train 3.983793 (lr=2.3743e-04) (hash(x)=165136835)
427
+ 3530 train 4.062000 (lr=2.3460e-04) (hash(x)=163869056)
428
+ 3540 train 4.003016 (lr=2.3179e-04) (hash(x)=161580134)
429
+ 3550 train 4.014832 (lr=2.2901e-04) (hash(x)=168595690)
430
+ 3560 train 4.052648 (lr=2.2626e-04) (hash(x)=158576466)
431
+ 3570 train 4.015902 (lr=2.2354e-04) (hash(x)=157298619)
432
+ 3580 train 4.098791 (lr=2.2085e-04) (hash(x)=160892091)
433
+ 3590 train 3.972530 (lr=2.1818e-04) (hash(x)=165168008)
434
+ 3600 val loss 4.0521
435
+ 3600 val perplexity 57.5159
436
+ 3600 train 4.070959 (lr=2.1555e-04) (hash(x)=144911638)
437
+ 3610 train 4.033093 (lr=2.1294e-04) (hash(x)=150614513)
438
+ 3620 train 4.019637 (lr=2.1036e-04) (hash(x)=167695016)
439
+ 3630 train 3.935343 (lr=2.0781e-04) (hash(x)=182708233)
440
+ 3640 train 4.041108 (lr=2.0529e-04) (hash(x)=154146274)
441
+ 3650 train 4.013737 (lr=2.0280e-04) (hash(x)=159129968)
442
+ 3660 train 3.923046 (lr=2.0033e-04) (hash(x)=146209497)
443
+ 3670 train 3.943368 (lr=1.9790e-04) (hash(x)=145578104)
444
+ 3680 train 3.938535 (lr=1.9550e-04) (hash(x)=150243521)
445
+ 3690 train 3.891583 (lr=1.9313e-04) (hash(x)=154328313)
446
+ 3700 val loss 4.0521
447
+ 3700 val perplexity 57.5155
448
+ 3700 train 3.858954 (lr=1.9079e-04) (hash(x)=159904582)
449
+ 3710 train 4.207923 (lr=1.8848e-04) (hash(x)=175986417)
450
+ 3720 train 4.119118 (lr=1.8619e-04) (hash(x)=158509479)
451
+ 3730 train 4.135247 (lr=1.8395e-04) (hash(x)=156087760)
452
+ 3740 train 4.134570 (lr=1.8173e-04) (hash(x)=177667170)
453
+ 3750 train 4.104941 (lr=1.7954e-04) (hash(x)=145256842)
454
+ 3760 train 4.066597 (lr=1.7738e-04) (hash(x)=149855635)
455
+ 3770 train 4.105114 (lr=1.7526e-04) (hash(x)=168798886)
456
+ 3780 train 4.020910 (lr=1.7317e-04) (hash(x)=156277179)
457
+ 3790 train 3.990409 (lr=1.7111e-04) (hash(x)=149698945)
458
+ 3800 val loss 4.0388
459
+ 3800 val perplexity 56.7602
460
+ 3800 train 3.947018 (lr=1.6908e-04) (hash(x)=168039654)
461
+ 3810 train 4.060639 (lr=1.6708e-04) (hash(x)=154595574)
462
+ 3820 train 4.107314 (lr=1.6511e-04) (hash(x)=158303284)
463
+ 3830 train 4.123205 (lr=1.6318e-04) (hash(x)=164128644)
464
+ 3840 train 4.078045 (lr=1.6128e-04) (hash(x)=166842997)
465
+ 3850 train 3.989197 (lr=1.5942e-04) (hash(x)=177681163)
466
+ 3860 train 3.969201 (lr=1.5758e-04) (hash(x)=177062575)
467
+ 3870 train 4.006938 (lr=1.5578e-04) (hash(x)=158957159)
468
+ 3880 train 3.942889 (lr=1.5401e-04) (hash(x)=151567740)
469
+ 3890 train 4.020599 (lr=1.5228e-04) (hash(x)=153261675)
470
+ 3900 val loss 4.0308
471
+ 3900 val perplexity 56.3078
472
+ 3900 train 4.285314 (lr=1.5058e-04) (hash(x)=160268683)
473
+ 3910 train 4.015713 (lr=1.4891e-04) (hash(x)=153745240)
474
+ 3920 train 4.069378 (lr=1.4727e-04) (hash(x)=164537824)
475
+ 3930 train 4.015658 (lr=1.4567e-04) (hash(x)=162062116)
476
+ 3940 train 3.971916 (lr=1.4411e-04) (hash(x)=168626742)
477
+ 3950 train 4.042293 (lr=1.4257e-04) (hash(x)=148753561)
478
+ 3960 train 3.962564 (lr=1.4108e-04) (hash(x)=163526242)
479
+ 3970 train 4.179369 (lr=1.3961e-04) (hash(x)=159329323)
480
+ 3980 train 4.102938 (lr=1.3818e-04) (hash(x)=146502339)
481
+ 3990 train 3.800254 (lr=1.3678e-04) (hash(x)=158409708)
482
+ 4000 val loss 4.0259
483
+ 4000 val perplexity 56.0289
484
+ 4000 train 3.975248 (lr=1.3542e-04) (hash(x)=154645137)
485
+ 4010 train 3.980945 (lr=1.3410e-04) (hash(x)=160044026)
486
+ 4020 train 3.961409 (lr=1.3280e-04) (hash(x)=155544686)
487
+ 4030 train 3.949321 (lr=1.3155e-04) (hash(x)=147589421)
488
+ 4040 train 4.060771 (lr=1.3032e-04) (hash(x)=158607990)
489
+ 4050 train 3.967876 (lr=1.2914e-04) (hash(x)=154108872)
490
+ 4060 train 4.047066 (lr=1.2798e-04) (hash(x)=167101465)
491
+ 4070 train 4.012476 (lr=1.2687e-04) (hash(x)=160323363)
492
+ 4080 train 3.997480 (lr=1.2578e-04) (hash(x)=150423692)
493
+ 4090 train 3.975677 (lr=1.2474e-04) (hash(x)=167461885)
494
+ 4100 val loss 4.0167
495
+ 4100 val perplexity 55.5162
496
+ 4100 train 4.014330 (lr=1.2373e-04) (hash(x)=151783377)
497
+ 4110 train 4.009905 (lr=1.2275e-04) (hash(x)=165679649)
498
+ 4120 train 3.907338 (lr=1.2181e-04) (hash(x)=153643039)
499
+ 4130 train 3.953771 (lr=1.2091e-04) (hash(x)=149648749)
500
+ 4140 train 3.989294 (lr=1.2004e-04) (hash(x)=162751418)
501
+ 4150 train 3.987859 (lr=1.1920e-04) (hash(x)=157047478)
502
+ 4160 train 3.925452 (lr=1.1841e-04) (hash(x)=156391812)
503
+ 4170 train 3.997423 (lr=1.1764e-04) (hash(x)=180926985)
504
+ 4180 train 4.042068 (lr=1.1692e-04) (hash(x)=168699059)
505
+ 4190 train 3.961898 (lr=1.1623e-04) (hash(x)=155410372)
506
+ 4200 val loss 4.0132
507
+ 4200 val perplexity 55.3224
508
+ 4200 train 3.969065 (lr=1.1557e-04) (hash(x)=152615252)
509
+ 4210 train 3.935683 (lr=1.1496e-04) (hash(x)=155679453)
510
+ 4220 train 4.048816 (lr=1.1437e-04) (hash(x)=152954399)
511
+ 4230 train 4.056474 (lr=1.1383e-04) (hash(x)=164608084)
512
+ 4240 train 4.079751 (lr=1.1332e-04) (hash(x)=162463867)
513
+ 4250 train 3.996219 (lr=1.1285e-04) (hash(x)=168832088)
514
+ 4260 train 3.965529 (lr=1.1241e-04) (hash(x)=162698882)
515
+ 4270 train 3.995934 (lr=1.1201e-04) (hash(x)=180861492)
516
+ 4280 train 3.886028 (lr=1.1164e-04) (hash(x)=160824300)
517
+ 4290 train 3.946921 (lr=1.1132e-04) (hash(x)=157750938)
518
+ 4300 val loss 4.0129
519
+ 4300 val perplexity 55.3095
520
+ 4300 train 3.896388 (lr=1.1103e-04) (hash(x)=157216833)
521
+ 4310 train 3.897977 (lr=1.1077e-04) (hash(x)=176184661)
522
+ 4320 train 4.060944 (lr=1.1055e-04) (hash(x)=161537117)
523
+ 4330 train 4.016997 (lr=1.1037e-04) (hash(x)=160700216)
524
+ 4340 train 3.971396 (lr=1.1022e-04) (hash(x)=175519903)
525
+ 4350 train 3.970876 (lr=1.1011e-04) (hash(x)=167988701)
526
+ 4360 train 3.922414 (lr=1.1004e-04) (hash(x)=162619084)
527
+ 4370 train 4.303748 (lr=1.1000e-04) (hash(x)=166292684)
528
+ 4374 val loss 4.0068
529
+ 4374 val perplexity 54.9684
one_mask_per_head_2_latent_masks/model_04374.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:444fec1612eb3130b35126e2cf55bf3f73b605c81356bdeef8f4a33706d24c20
3
+ size 95220594
one_mask_per_head_2_latent_masks/optimizer_04374.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:ef049463eda873326d3571617584612317c4b2e62b5826bc579f9e18130c6aee
3
+ size 184163894