andrew-healey commited on
Commit
5c0f1df
·
verified ·
1 Parent(s): 3c0c3ef

Upload folder using huggingface_hub

Browse files
one_mask_per_head_1_latent_mask/args.json ADDED
@@ -0,0 +1 @@
 
 
1
+ {"hellaswag": true, "attention_kind": "selective", "log_dir": "logs/repro_selective_pattern_rankings/one_mask_per_head_1_latent_mask", "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": 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": 0.0011, "decay_lr": true, "readout_zero_init": false, "query_zero_init": false, "l1_loss": false, "debugpy": false, "key": "one_mask_per_head_1_latent_mask", "n_embd": 256}
one_mask_per_head_1_latent_mask/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_1_latent_mask/log2.txt ADDED
@@ -0,0 +1,529 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ max_steps: 4375
2
+ 0 val loss 11.2456
3
+ 0 val perplexity 76545.7266
4
+ 0 train 11.257509 (lr=1.5385e-06) (hash(x)=161979327)
5
+ 10 train 10.309372 (lr=1.6923e-05) (hash(x)=165941070)
6
+ 20 train 9.818686 (lr=3.2308e-05) (hash(x)=152918033)
7
+ 30 train 9.565734 (lr=4.7692e-05) (hash(x)=169931607)
8
+ 40 train 9.097921 (lr=6.3077e-05) (hash(x)=174342675)
9
+ 50 train 8.656292 (lr=7.8462e-05) (hash(x)=163872554)
10
+ 60 train 8.285915 (lr=9.3846e-05) (hash(x)=161534908)
11
+ 70 train 7.777361 (lr=1.0923e-04) (hash(x)=156776712)
12
+ 80 train 7.676078 (lr=1.2462e-04) (hash(x)=164996929)
13
+ 90 train 7.545615 (lr=1.4000e-04) (hash(x)=152962169)
14
+ 100 val loss 7.4813
15
+ 100 val perplexity 1774.4757
16
+ 100 train 7.335667 (lr=1.5538e-04) (hash(x)=147111701)
17
+ 110 train 7.293728 (lr=1.7077e-04) (hash(x)=167595270)
18
+ 120 train 7.198477 (lr=1.8615e-04) (hash(x)=152205426)
19
+ 130 train 7.028139 (lr=2.0154e-04) (hash(x)=158427335)
20
+ 140 train 6.988869 (lr=2.1692e-04) (hash(x)=174588384)
21
+ 150 train 7.038250 (lr=2.3231e-04) (hash(x)=157067270)
22
+ 160 train 6.943315 (lr=2.4769e-04) (hash(x)=161859595)
23
+ 170 train 6.820362 (lr=2.6308e-04) (hash(x)=162739537)
24
+ 180 train 6.739737 (lr=2.7846e-04) (hash(x)=170785304)
25
+ 190 train 6.635871 (lr=2.9385e-04) (hash(x)=168658088)
26
+ 200 val loss 6.6001
27
+ 200 val perplexity 735.1863
28
+ 200 train 6.598641 (lr=3.0923e-04) (hash(x)=150749086)
29
+ 210 train 6.463853 (lr=3.2462e-04) (hash(x)=148789576)
30
+ 220 train 6.390158 (lr=3.4000e-04) (hash(x)=153164408)
31
+ 230 train 6.557399 (lr=3.5538e-04) (hash(x)=184594366)
32
+ 240 train 6.261622 (lr=3.7077e-04) (hash(x)=157817319)
33
+ 250 train 6.285032 (lr=3.8615e-04) (hash(x)=149451562)
34
+ 260 train 6.196124 (lr=4.0154e-04) (hash(x)=165738161)
35
+ 270 train 6.130772 (lr=4.1692e-04) (hash(x)=165724764)
36
+ 280 train 6.071974 (lr=4.3231e-04) (hash(x)=155414136)
37
+ 290 train 5.996763 (lr=4.4769e-04) (hash(x)=165090903)
38
+ 300 val loss 6.1397
39
+ 300 val perplexity 463.9232
40
+ 300 train 5.858901 (lr=4.6308e-04) (hash(x)=154445545)
41
+ 310 train 5.814085 (lr=4.7846e-04) (hash(x)=158949603)
42
+ 320 train 5.896953 (lr=4.9385e-04) (hash(x)=161412973)
43
+ 330 train 5.962414 (lr=5.0923e-04) (hash(x)=164572249)
44
+ 340 train 6.004774 (lr=5.2462e-04) (hash(x)=162202191)
45
+ 350 train 5.979802 (lr=5.4000e-04) (hash(x)=158135392)
46
+ 360 train 5.935951 (lr=5.5538e-04) (hash(x)=163752391)
47
+ 370 train 5.904899 (lr=5.7077e-04) (hash(x)=172758799)
48
+ 380 train 5.707806 (lr=5.8615e-04) (hash(x)=144569488)
49
+ 390 train 5.712571 (lr=6.0154e-04) (hash(x)=150429229)
50
+ 400 val loss 5.7483
51
+ 400 val perplexity 313.6446
52
+ 400 train 5.690661 (lr=6.1692e-04) (hash(x)=154261530)
53
+ 410 train 5.624164 (lr=6.3231e-04) (hash(x)=159172506)
54
+ 420 train 5.632329 (lr=6.4769e-04) (hash(x)=154419867)
55
+ 430 train 5.656596 (lr=6.6308e-04) (hash(x)=167337099)
56
+ 440 train 5.549316 (lr=6.7846e-04) (hash(x)=154535255)
57
+ 450 train 5.511420 (lr=6.9385e-04) (hash(x)=153077071)
58
+ 460 train 5.445433 (lr=7.0923e-04) (hash(x)=150853070)
59
+ 470 train 5.264164 (lr=7.2462e-04) (hash(x)=160701113)
60
+ 480 train 5.380118 (lr=7.4000e-04) (hash(x)=154690870)
61
+ 490 train 5.209565 (lr=7.5538e-04) (hash(x)=163834286)
62
+ 500 val loss 5.5233
63
+ 500 val perplexity 250.4550
64
+ 500 train 5.277184 (lr=7.7077e-04) (hash(x)=162651717)
65
+ 510 train 5.371642 (lr=7.8615e-04) (hash(x)=160448311)
66
+ 520 train 5.401184 (lr=8.0154e-04) (hash(x)=165741643)
67
+ 530 train 5.524035 (lr=8.1692e-04) (hash(x)=153201264)
68
+ 540 train 5.421144 (lr=8.3231e-04) (hash(x)=159027160)
69
+ 550 train 5.407862 (lr=8.4769e-04) (hash(x)=170769095)
70
+ 560 train 5.411930 (lr=8.6308e-04) (hash(x)=153606190)
71
+ 570 train 5.372353 (lr=8.7846e-04) (hash(x)=159708689)
72
+ 580 train 5.312476 (lr=8.9385e-04) (hash(x)=179108359)
73
+ 590 train 5.312405 (lr=9.0923e-04) (hash(x)=157134241)
74
+ 600 val loss 5.3129
75
+ 600 val perplexity 202.9419
76
+ 600 train 5.287734 (lr=9.2462e-04) (hash(x)=162178356)
77
+ 610 train 5.213418 (lr=9.4000e-04) (hash(x)=167646928)
78
+ 620 train 5.182909 (lr=9.5538e-04) (hash(x)=160979601)
79
+ 630 train 5.092290 (lr=9.7077e-04) (hash(x)=158349468)
80
+ 640 train 5.092687 (lr=9.8615e-04) (hash(x)=150370348)
81
+ 650 train 5.051288 (lr=1.0015e-03) (hash(x)=152810075)
82
+ 660 train 4.966309 (lr=1.0169e-03) (hash(x)=145786262)
83
+ 670 train 4.974841 (lr=1.0323e-03) (hash(x)=160779604)
84
+ 680 train 4.868185 (lr=1.0477e-03) (hash(x)=152343688)
85
+ 690 train 4.901938 (lr=1.0631e-03) (hash(x)=166507600)
86
+ 700 val loss 5.1835
87
+ 700 val perplexity 178.2982
88
+ 700 train 4.874757 (lr=1.0785e-03) (hash(x)=162623700)
89
+ 710 train 5.138314 (lr=1.0938e-03) (hash(x)=162293403)
90
+ 720 train 5.096844 (lr=1.1000e-03) (hash(x)=159564657)
91
+ 730 train 5.118099 (lr=1.1000e-03) (hash(x)=154209107)
92
+ 740 train 5.086290 (lr=1.0999e-03) (hash(x)=156485709)
93
+ 750 train 5.051794 (lr=1.0998e-03) (hash(x)=160834702)
94
+ 760 train 5.055502 (lr=1.0996e-03) (hash(x)=152240275)
95
+ 770 train 5.040873 (lr=1.0994e-03) (hash(x)=168903075)
96
+ 780 train 5.101682 (lr=1.0992e-03) (hash(x)=166843738)
97
+ 790 train 5.131664 (lr=1.0990e-03) (hash(x)=166358104)
98
+ 800 val loss 5.0116
99
+ 800 val perplexity 150.1413
100
+ 800 train 4.948340 (lr=1.0987e-03) (hash(x)=153477747)
101
+ 810 train 4.875656 (lr=1.0984e-03) (hash(x)=157954385)
102
+ 820 train 4.866747 (lr=1.0980e-03) (hash(x)=174383505)
103
+ 830 train 4.883289 (lr=1.0976e-03) (hash(x)=145226181)
104
+ 840 train 4.854624 (lr=1.0972e-03) (hash(x)=157665743)
105
+ 850 train 4.852358 (lr=1.0967e-03) (hash(x)=158565836)
106
+ 860 train 4.853584 (lr=1.0962e-03) (hash(x)=171168105)
107
+ 870 train 5.249199 (lr=1.0956e-03) (hash(x)=159017777)
108
+ 880 train 5.025313 (lr=1.0950e-03) (hash(x)=168156768)
109
+ 890 train 4.923164 (lr=1.0944e-03) (hash(x)=145645245)
110
+ 900 val loss 4.8593
111
+ 900 val perplexity 128.9318
112
+ 900 train 4.744902 (lr=1.0938e-03) (hash(x)=151245521)
113
+ 910 train 4.772571 (lr=1.0931e-03) (hash(x)=182614672)
114
+ 920 train 4.760611 (lr=1.0924e-03) (hash(x)=167377432)
115
+ 930 train 4.678514 (lr=1.0916e-03) (hash(x)=171347228)
116
+ 940 train 4.700559 (lr=1.0908e-03) (hash(x)=171588755)
117
+ 950 train 4.783489 (lr=1.0900e-03) (hash(x)=159560168)
118
+ 960 train 4.671362 (lr=1.0891e-03) (hash(x)=171801357)
119
+ 970 train 4.675605 (lr=1.0882e-03) (hash(x)=172513884)
120
+ 980 train 4.694594 (lr=1.0872e-03) (hash(x)=170659685)
121
+ 990 train 4.720859 (lr=1.0863e-03) (hash(x)=165436358)
122
+ 1000 val loss 4.7293
123
+ 1000 val perplexity 113.2201
124
+ 1000 train 4.781206 (lr=1.0853e-03) (hash(x)=167684690)
125
+ 1010 train 4.735608 (lr=1.0842e-03) (hash(x)=127225492)
126
+ 1020 train 4.748623 (lr=1.0831e-03) (hash(x)=167575323)
127
+ 1030 train 4.672286 (lr=1.0820e-03) (hash(x)=161266302)
128
+ 1040 train 4.637230 (lr=1.0809e-03) (hash(x)=155607661)
129
+ 1050 train 4.682666 (lr=1.0797e-03) (hash(x)=165685359)
130
+ 1060 train 4.536556 (lr=1.0785e-03) (hash(x)=160417824)
131
+ 1070 train 4.670729 (lr=1.0772e-03) (hash(x)=173313882)
132
+ 1080 train 4.637849 (lr=1.0759e-03) (hash(x)=165485047)
133
+ 1090 train 4.499608 (lr=1.0746e-03) (hash(x)=159848305)
134
+ 1100 val loss 4.6481
135
+ 1100 val perplexity 104.3899
136
+ 1100 train 4.663920 (lr=1.0732e-03) (hash(x)=170890601)
137
+ 1110 train 4.598657 (lr=1.0718e-03) (hash(x)=172689983)
138
+ 1120 train 4.618568 (lr=1.0704e-03) (hash(x)=154762859)
139
+ 1130 train 4.539643 (lr=1.0689e-03) (hash(x)=168196871)
140
+ 1140 train 4.572109 (lr=1.0674e-03) (hash(x)=156715577)
141
+ 1150 train 4.695937 (lr=1.0659e-03) (hash(x)=166381788)
142
+ 1160 train 4.621961 (lr=1.0643e-03) (hash(x)=162095130)
143
+ 1170 train 4.591633 (lr=1.0627e-03) (hash(x)=166094750)
144
+ 1180 train 4.530290 (lr=1.0611e-03) (hash(x)=154965753)
145
+ 1190 train 4.520170 (lr=1.0594e-03) (hash(x)=172766447)
146
+ 1200 val loss 4.5610
147
+ 1200 val perplexity 95.6822
148
+ 1200 train 4.431065 (lr=1.0577e-03) (hash(x)=161185334)
149
+ 1210 train 4.602028 (lr=1.0560e-03) (hash(x)=167224144)
150
+ 1220 train 4.441942 (lr=1.0542e-03) (hash(x)=167252866)
151
+ 1230 train 4.451216 (lr=1.0524e-03) (hash(x)=155856924)
152
+ 1240 train 4.445290 (lr=1.0506e-03) (hash(x)=155956446)
153
+ 1250 train 4.354044 (lr=1.0487e-03) (hash(x)=140794796)
154
+ 1260 train 4.483375 (lr=1.0468e-03) (hash(x)=166863869)
155
+ 1270 train 4.538429 (lr=1.0449e-03) (hash(x)=166021892)
156
+ 1280 train 4.540035 (lr=1.0429e-03) (hash(x)=170503648)
157
+ 1290 train 4.456444 (lr=1.0409e-03) (hash(x)=164012719)
158
+ 1300 val loss 4.5138
159
+ 1300 val perplexity 91.2690
160
+ 1300 train 4.521811 (lr=1.0389e-03) (hash(x)=156519834)
161
+ 1310 train 4.471607 (lr=1.0368e-03) (hash(x)=165146652)
162
+ 1320 train 4.495647 (lr=1.0347e-03) (hash(x)=174391990)
163
+ 1330 train 4.496171 (lr=1.0326e-03) (hash(x)=161131201)
164
+ 1340 train 4.479854 (lr=1.0305e-03) (hash(x)=162240912)
165
+ 1350 train 4.398129 (lr=1.0283e-03) (hash(x)=152462931)
166
+ 1360 train 4.398789 (lr=1.0261e-03) (hash(x)=150443278)
167
+ 1370 train 4.337878 (lr=1.0238e-03) (hash(x)=153721711)
168
+ 1380 train 4.479029 (lr=1.0215e-03) (hash(x)=158070628)
169
+ 1390 train 4.494162 (lr=1.0192e-03) (hash(x)=163318827)
170
+ 1400 val loss 4.4852
171
+ 1400 val perplexity 88.6971
172
+ 1400 train 4.404810 (lr=1.0169e-03) (hash(x)=154210336)
173
+ 1410 train 4.457428 (lr=1.0145e-03) (hash(x)=162393933)
174
+ 1420 train 4.450608 (lr=1.0121e-03) (hash(x)=156659308)
175
+ 1430 train 4.430614 (lr=1.0097e-03) (hash(x)=171243439)
176
+ 1440 train 4.568925 (lr=1.0072e-03) (hash(x)=178459521)
177
+ 1450 train 4.417697 (lr=1.0047e-03) (hash(x)=159166350)
178
+ 1460 train 4.345870 (lr=1.0022e-03) (hash(x)=173554569)
179
+ 1470 train 4.433871 (lr=9.9964e-04) (hash(x)=158362521)
180
+ 1480 train 4.395388 (lr=9.9706e-04) (hash(x)=165899207)
181
+ 1490 train 4.422701 (lr=9.9445e-04) (hash(x)=156470939)
182
+ 1500 val loss 4.4344
183
+ 1500 val perplexity 84.2976
184
+ 1500 train 4.394033 (lr=9.9182e-04) (hash(x)=164197780)
185
+ 1510 train 4.292045 (lr=9.8915e-04) (hash(x)=146709896)
186
+ 1520 train 4.308975 (lr=9.8646e-04) (hash(x)=151304838)
187
+ 1530 train 4.349345 (lr=9.8374e-04) (hash(x)=163187333)
188
+ 1540 train 4.299678 (lr=9.8099e-04) (hash(x)=153630537)
189
+ 1550 train 4.306401 (lr=9.7821e-04) (hash(x)=157299103)
190
+ 1560 train 4.216343 (lr=9.7540e-04) (hash(x)=153786477)
191
+ 1570 train 4.545338 (lr=9.7257e-04) (hash(x)=155037209)
192
+ 1580 train 4.423993 (lr=9.6971e-04) (hash(x)=146157308)
193
+ 1590 train 4.463340 (lr=9.6683e-04) (hash(x)=147823317)
194
+ 1600 val loss 4.4049
195
+ 1600 val perplexity 81.8482
196
+ 1600 train 4.450758 (lr=9.6391e-04) (hash(x)=169140159)
197
+ 1610 train 4.228226 (lr=9.6097e-04) (hash(x)=157811493)
198
+ 1620 train 4.320873 (lr=9.5801e-04) (hash(x)=168940851)
199
+ 1630 train 4.309066 (lr=9.5502e-04) (hash(x)=151149213)
200
+ 1640 train 4.407758 (lr=9.5200e-04) (hash(x)=161822009)
201
+ 1650 train 4.822773 (lr=9.4896e-04) (hash(x)=151056313)
202
+ 1660 train 4.334801 (lr=9.4589e-04) (hash(x)=156920306)
203
+ 1670 train 4.246860 (lr=9.4280e-04) (hash(x)=159695815)
204
+ 1680 train 4.256786 (lr=9.3968e-04) (hash(x)=160232907)
205
+ 1690 train 4.249704 (lr=9.3654e-04) (hash(x)=158099357)
206
+ 1700 val loss 4.3996
207
+ 1700 val perplexity 81.4190
208
+ 1700 train 4.300536 (lr=9.3337e-04) (hash(x)=164127381)
209
+ 1710 train 4.480193 (lr=9.3018e-04) (hash(x)=156592040)
210
+ 1720 train 4.395718 (lr=9.2696e-04) (hash(x)=160220265)
211
+ 1730 train 4.340187 (lr=9.2372e-04) (hash(x)=156162204)
212
+ 1740 train 4.335075 (lr=9.2046e-04) (hash(x)=152767893)
213
+ 1750 train 4.317155 (lr=9.1717e-04) (hash(x)=162707306)
214
+ 1760 train 4.336840 (lr=9.1386e-04) (hash(x)=154472004)
215
+ 1770 train 4.199519 (lr=9.1053e-04) (hash(x)=160374234)
216
+ 1780 train 4.353246 (lr=9.0718e-04) (hash(x)=159709225)
217
+ 1790 train 4.387311 (lr=9.0380e-04) (hash(x)=152095944)
218
+ 1800 val loss 4.3510
219
+ 1800 val perplexity 77.5581
220
+ 1800 train 4.361043 (lr=9.0040e-04) (hash(x)=157920255)
221
+ 1810 train 4.106506 (lr=8.9698e-04) (hash(x)=167237680)
222
+ 1820 train 4.224635 (lr=8.9354e-04) (hash(x)=161255774)
223
+ 1830 train 4.243127 (lr=8.9008e-04) (hash(x)=154275481)
224
+ 1840 train 4.194548 (lr=8.8660e-04) (hash(x)=152398178)
225
+ 1850 train 4.306996 (lr=8.8309e-04) (hash(x)=147646022)
226
+ 1860 train 4.320871 (lr=8.7957e-04) (hash(x)=158895559)
227
+ 1870 train 4.264083 (lr=8.7602e-04) (hash(x)=162742598)
228
+ 1880 train 4.255224 (lr=8.7245e-04) (hash(x)=169911029)
229
+ 1890 train 4.294629 (lr=8.6887e-04) (hash(x)=172068270)
230
+ 1900 val loss 4.3496
231
+ 1900 val perplexity 77.4450
232
+ 1900 train 4.375401 (lr=8.6527e-04) (hash(x)=159975457)
233
+ 1910 train 4.366086 (lr=8.6164e-04) (hash(x)=178775667)
234
+ 1920 train 4.240202 (lr=8.5800e-04) (hash(x)=151180831)
235
+ 1930 train 4.261526 (lr=8.5434e-04) (hash(x)=154239412)
236
+ 1940 train 4.273993 (lr=8.5066e-04) (hash(x)=139144465)
237
+ 1950 train 4.226842 (lr=8.4696e-04) (hash(x)=194482413)
238
+ 1960 train 4.242887 (lr=8.4324e-04) (hash(x)=155812344)
239
+ 1970 train 4.242938 (lr=8.3951e-04) (hash(x)=157577031)
240
+ 1980 train 4.280186 (lr=8.3576e-04) (hash(x)=152045027)
241
+ 1990 train 4.183309 (lr=8.3199e-04) (hash(x)=141742221)
242
+ 2000 val loss 4.3045
243
+ 2000 val perplexity 74.0322
244
+ 2000 train 4.309828 (lr=8.2821e-04) (hash(x)=160555812)
245
+ 2010 train 4.243379 (lr=8.2441e-04) (hash(x)=160371335)
246
+ 2020 train 4.242740 (lr=8.2059e-04) (hash(x)=167890767)
247
+ 2030 train 4.316898 (lr=8.1676e-04) (hash(x)=165127715)
248
+ 2040 train 4.250819 (lr=8.1291e-04) (hash(x)=151587302)
249
+ 2050 train 4.273315 (lr=8.0905e-04) (hash(x)=172347989)
250
+ 2060 train 4.351806 (lr=8.0517e-04) (hash(x)=161301919)
251
+ 2070 train 4.282417 (lr=8.0128e-04) (hash(x)=161472934)
252
+ 2080 train 4.363932 (lr=7.9737e-04) (hash(x)=159777322)
253
+ 2090 train 4.283125 (lr=7.9345e-04) (hash(x)=151149770)
254
+ 2100 val loss 4.2823
255
+ 2100 val perplexity 72.4069
256
+ 2100 train 4.109067 (lr=7.8951e-04) (hash(x)=160283520)
257
+ 2110 train 4.297503 (lr=7.8556e-04) (hash(x)=154222025)
258
+ 2120 train 4.228489 (lr=7.8160e-04) (hash(x)=155638178)
259
+ 2130 train 4.138748 (lr=7.7762e-04) (hash(x)=149125384)
260
+ 2140 train 4.191515 (lr=7.7363e-04) (hash(x)=159496074)
261
+ 2150 train 4.274934 (lr=7.6963e-04) (hash(x)=162019154)
262
+ 2160 train 4.435386 (lr=7.6562e-04) (hash(x)=155466928)
263
+ 2170 train 4.312972 (lr=7.6160e-04) (hash(x)=173469994)
264
+ 2180 train 4.420057 (lr=7.5756e-04) (hash(x)=163448753)
265
+ 2190 train 4.250426 (lr=7.5351e-04) (hash(x)=166595123)
266
+ 2200 val loss 4.2630
267
+ 2200 val perplexity 71.0219
268
+ 2200 train 4.129596 (lr=7.4945e-04) (hash(x)=165958875)
269
+ 2210 train 4.267982 (lr=7.4538e-04) (hash(x)=165019039)
270
+ 2220 train 4.251116 (lr=7.4130e-04) (hash(x)=167325225)
271
+ 2230 train 4.086644 (lr=7.3721e-04) (hash(x)=154287439)
272
+ 2240 train 4.151548 (lr=7.3312e-04) (hash(x)=158565817)
273
+ 2250 train 4.116206 (lr=7.2901e-04) (hash(x)=160592835)
274
+ 2260 train 4.158035 (lr=7.2489e-04) (hash(x)=144591648)
275
+ 2270 train 4.037133 (lr=7.2076e-04) (hash(x)=150486876)
276
+ 2280 train 4.230866 (lr=7.1663e-04) (hash(x)=159194788)
277
+ 2290 train 4.307661 (lr=7.1248e-04) (hash(x)=164315345)
278
+ 2300 val loss 4.2406
279
+ 2300 val perplexity 69.4495
280
+ 2300 train 4.229617 (lr=7.0833e-04) (hash(x)=170470665)
281
+ 2310 train 4.193537 (lr=7.0417e-04) (hash(x)=176539068)
282
+ 2320 train 4.178369 (lr=7.0001e-04) (hash(x)=161008474)
283
+ 2330 train 4.245328 (lr=6.9583e-04) (hash(x)=168474380)
284
+ 2340 train 4.246149 (lr=6.9165e-04) (hash(x)=156072131)
285
+ 2350 train 4.341769 (lr=6.8747e-04) (hash(x)=165127087)
286
+ 2360 train 4.322897 (lr=6.8327e-04) (hash(x)=159462801)
287
+ 2370 train 4.164763 (lr=6.7908e-04) (hash(x)=161698394)
288
+ 2380 train 4.088496 (lr=6.7487e-04) (hash(x)=165588078)
289
+ 2390 train 4.097767 (lr=6.7066e-04) (hash(x)=171673672)
290
+ 2400 val loss 4.2320
291
+ 2400 val perplexity 68.8528
292
+ 2400 train 4.161695 (lr=6.6645e-04) (hash(x)=157381516)
293
+ 2410 train 4.165929 (lr=6.6223e-04) (hash(x)=163238532)
294
+ 2420 train 4.126830 (lr=6.5801e-04) (hash(x)=156178237)
295
+ 2430 train 4.006164 (lr=6.5378e-04) (hash(x)=155080617)
296
+ 2440 train 4.118982 (lr=6.4955e-04) (hash(x)=167490538)
297
+ 2450 train 4.090334 (lr=6.4532e-04) (hash(x)=163757352)
298
+ 2460 train 4.045750 (lr=6.4108e-04) (hash(x)=156734811)
299
+ 2470 train 4.037716 (lr=6.3684e-04) (hash(x)=161522341)
300
+ 2480 train 4.298586 (lr=6.3260e-04) (hash(x)=174032060)
301
+ 2490 train 4.259520 (lr=6.2836e-04) (hash(x)=163065818)
302
+ 2500 val loss 4.2052
303
+ 2500 val perplexity 67.0363
304
+ 2500 train 4.316655 (lr=6.2412e-04) (hash(x)=126959780)
305
+ 2510 train 4.242546 (lr=6.1987e-04) (hash(x)=153844939)
306
+ 2520 train 4.220299 (lr=6.1562e-04) (hash(x)=174018174)
307
+ 2530 train 4.196525 (lr=6.1137e-04) (hash(x)=150568427)
308
+ 2540 train 4.195255 (lr=6.0712e-04) (hash(x)=161725206)
309
+ 2550 train 4.163591 (lr=6.0288e-04) (hash(x)=181346417)
310
+ 2560 train 4.154365 (lr=5.9863e-04) (hash(x)=168730011)
311
+ 2570 train 4.048563 (lr=5.9438e-04) (hash(x)=155900214)
312
+ 2580 train 4.096811 (lr=5.9013e-04) (hash(x)=160014621)
313
+ 2590 train 4.067186 (lr=5.8588e-04) (hash(x)=148996689)
314
+ 2600 val loss 4.1917
315
+ 2600 val perplexity 66.1372
316
+ 2600 train 4.147632 (lr=5.8164e-04) (hash(x)=145183968)
317
+ 2610 train 3.898639 (lr=5.7740e-04) (hash(x)=159420240)
318
+ 2620 train 3.874911 (lr=5.7316e-04) (hash(x)=151169903)
319
+ 2630 train 3.888545 (lr=5.6892e-04) (hash(x)=149454811)
320
+ 2640 train 3.955657 (lr=5.6468e-04) (hash(x)=155603914)
321
+ 2650 train 3.949263 (lr=5.6045e-04) (hash(x)=169545195)
322
+ 2660 train 4.298136 (lr=5.5622e-04) (hash(x)=168454079)
323
+ 2670 train 4.196206 (lr=5.5199e-04) (hash(x)=163110417)
324
+ 2680 train 4.138322 (lr=5.4777e-04) (hash(x)=166075603)
325
+ 2690 train 4.173648 (lr=5.4355e-04) (hash(x)=174132258)
326
+ 2700 val loss 4.1782
327
+ 2700 val perplexity 65.2473
328
+ 2700 train 4.177142 (lr=5.3934e-04) (hash(x)=169846940)
329
+ 2710 train 4.076103 (lr=5.3513e-04) (hash(x)=151553731)
330
+ 2720 train 4.179401 (lr=5.3092e-04) (hash(x)=164024255)
331
+ 2730 train 4.197304 (lr=5.2673e-04) (hash(x)=156458097)
332
+ 2740 train 4.132477 (lr=5.2253e-04) (hash(x)=154663748)
333
+ 2750 train 3.989853 (lr=5.1835e-04) (hash(x)=157548181)
334
+ 2760 train 4.103660 (lr=5.1417e-04) (hash(x)=151584783)
335
+ 2770 train 4.124932 (lr=5.0999e-04) (hash(x)=152692594)
336
+ 2780 train 4.101386 (lr=5.0583e-04) (hash(x)=165952919)
337
+ 2790 train 3.995130 (lr=5.0167e-04) (hash(x)=157484183)
338
+ 2800 val loss 4.1638
339
+ 2800 val perplexity 64.3180
340
+ 2800 train 4.067719 (lr=4.9752e-04) (hash(x)=154790034)
341
+ 2810 train 3.931058 (lr=4.9337e-04) (hash(x)=157856097)
342
+ 2820 train 3.911710 (lr=4.8924e-04) (hash(x)=168903939)
343
+ 2830 train 4.038960 (lr=4.8511e-04) (hash(x)=167001475)
344
+ 2840 train 3.864670 (lr=4.8099e-04) (hash(x)=158987150)
345
+ 2850 train 4.100732 (lr=4.7688e-04) (hash(x)=168695407)
346
+ 2860 train 4.234872 (lr=4.7279e-04) (hash(x)=164999034)
347
+ 2870 train 4.077283 (lr=4.6870e-04) (hash(x)=191197316)
348
+ 2880 train 4.039079 (lr=4.6462e-04) (hash(x)=159625200)
349
+ 2890 train 4.193256 (lr=4.6055e-04) (hash(x)=157756126)
350
+ 2900 val loss 4.1397
351
+ 2900 val perplexity 62.7864
352
+ 2900 train 4.053407 (lr=4.5649e-04) (hash(x)=164721377)
353
+ 2910 train 4.099225 (lr=4.5244e-04) (hash(x)=155873181)
354
+ 2920 train 4.047504 (lr=4.4840e-04) (hash(x)=156477082)
355
+ 2930 train 4.054124 (lr=4.4438e-04) (hash(x)=151261633)
356
+ 2940 train 3.996733 (lr=4.4037e-04) (hash(x)=160090385)
357
+ 2950 train 4.143826 (lr=4.3637e-04) (hash(x)=157300618)
358
+ 2960 train 4.052995 (lr=4.3238e-04) (hash(x)=163095555)
359
+ 2970 train 4.084399 (lr=4.2840e-04) (hash(x)=149649212)
360
+ 2980 train 3.894443 (lr=4.2444e-04) (hash(x)=174517675)
361
+ 2990 train 3.865269 (lr=4.2049e-04) (hash(x)=161344491)
362
+ 3000 val loss 4.1406
363
+ 3000 val perplexity 62.8406
364
+ 3000 train 3.892428 (lr=4.1655e-04) (hash(x)=163816905)
365
+ 3010 train 3.801877 (lr=4.1263e-04) (hash(x)=157420680)
366
+ 3020 train 3.910977 (lr=4.0872e-04) (hash(x)=153843035)
367
+ 3030 train 4.093313 (lr=4.0483e-04) (hash(x)=149770373)
368
+ 3040 train 4.168060 (lr=4.0095e-04) (hash(x)=165320741)
369
+ 3050 train 4.142211 (lr=3.9709e-04) (hash(x)=170181151)
370
+ 3060 train 4.059856 (lr=3.9324e-04) (hash(x)=163837230)
371
+ 3070 train 4.050051 (lr=3.8941e-04) (hash(x)=156812797)
372
+ 3080 train 4.029430 (lr=3.8559e-04) (hash(x)=143663627)
373
+ 3090 train 4.051957 (lr=3.8179e-04) (hash(x)=165052478)
374
+ 3100 val loss 4.1192
375
+ 3100 val perplexity 61.5127
376
+ 3100 train 3.990771 (lr=3.7801e-04) (hash(x)=151458705)
377
+ 3110 train 4.040544 (lr=3.7424e-04) (hash(x)=158228317)
378
+ 3120 train 4.205076 (lr=3.7049e-04) (hash(x)=149856801)
379
+ 3130 train 3.879729 (lr=3.6676e-04) (hash(x)=146297847)
380
+ 3140 train 3.978199 (lr=3.6304e-04) (hash(x)=172795167)
381
+ 3150 train 4.038453 (lr=3.5934e-04) (hash(x)=145936074)
382
+ 3160 train 4.121060 (lr=3.5566e-04) (hash(x)=157564303)
383
+ 3170 train 4.222927 (lr=3.5200e-04) (hash(x)=157451362)
384
+ 3180 train 4.140738 (lr=3.4836e-04) (hash(x)=156366043)
385
+ 3190 train 4.101098 (lr=3.4473e-04) (hash(x)=162344697)
386
+ 3200 val loss 4.1005
387
+ 3200 val perplexity 60.3733
388
+ 3200 train 4.122525 (lr=3.4113e-04) (hash(x)=164206298)
389
+ 3210 train 4.192482 (lr=3.3755e-04) (hash(x)=149468251)
390
+ 3220 train 4.123498 (lr=3.3398e-04) (hash(x)=157808606)
391
+ 3230 train 4.260423 (lr=3.3043e-04) (hash(x)=157078690)
392
+ 3240 train 4.069648 (lr=3.2691e-04) (hash(x)=165699696)
393
+ 3250 train 4.017827 (lr=3.2340e-04) (hash(x)=181352842)
394
+ 3260 train 4.112412 (lr=3.1992e-04) (hash(x)=142534383)
395
+ 3270 train 4.069358 (lr=3.1646e-04) (hash(x)=165297359)
396
+ 3280 train 4.114295 (lr=3.1302e-04) (hash(x)=160326815)
397
+ 3290 train 4.039746 (lr=3.0960e-04) (hash(x)=150043063)
398
+ 3300 val loss 4.0967
399
+ 3300 val perplexity 60.1412
400
+ 3300 train 3.995378 (lr=3.0620e-04) (hash(x)=164133091)
401
+ 3310 train 3.936071 (lr=3.0282e-04) (hash(x)=145803863)
402
+ 3320 train 3.958537 (lr=2.9947e-04) (hash(x)=157282009)
403
+ 3330 train 3.903255 (lr=2.9614e-04) (hash(x)=159126415)
404
+ 3340 train 4.104558 (lr=2.9283e-04) (hash(x)=164426675)
405
+ 3350 train 4.135722 (lr=2.8954e-04) (hash(x)=170424924)
406
+ 3360 train 4.073749 (lr=2.8628e-04) (hash(x)=163161621)
407
+ 3370 train 4.136198 (lr=2.8304e-04) (hash(x)=160786731)
408
+ 3380 train 4.095407 (lr=2.7982e-04) (hash(x)=165954688)
409
+ 3390 train 4.145745 (lr=2.7663e-04) (hash(x)=147371107)
410
+ 3400 val loss 4.0771
411
+ 3400 val perplexity 58.9738
412
+ 3400 train 4.087373 (lr=2.7346e-04) (hash(x)=153067604)
413
+ 3410 train 4.083257 (lr=2.7032e-04) (hash(x)=161748196)
414
+ 3420 train 4.097847 (lr=2.6720e-04) (hash(x)=168297963)
415
+ 3430 train 4.087326 (lr=2.6411e-04) (hash(x)=145086892)
416
+ 3440 train 3.991506 (lr=2.6104e-04) (hash(x)=173125711)
417
+ 3450 train 4.092023 (lr=2.5800e-04) (hash(x)=158080216)
418
+ 3460 train 3.913092 (lr=2.5498e-04) (hash(x)=158055750)
419
+ 3470 train 4.009450 (lr=2.5199e-04) (hash(x)=151381400)
420
+ 3480 train 3.954942 (lr=2.4903e-04) (hash(x)=147262346)
421
+ 3490 train 3.932483 (lr=2.4609e-04) (hash(x)=154740086)
422
+ 3500 val loss 4.0752
423
+ 3500 val perplexity 58.8637
424
+ 3500 train 3.858031 (lr=2.4317e-04) (hash(x)=145730968)
425
+ 3510 train 3.999305 (lr=2.4029e-04) (hash(x)=151355929)
426
+ 3520 train 3.987920 (lr=2.3743e-04) (hash(x)=165136835)
427
+ 3530 train 4.064452 (lr=2.3460e-04) (hash(x)=163869056)
428
+ 3540 train 4.007653 (lr=2.3179e-04) (hash(x)=161580134)
429
+ 3550 train 4.021440 (lr=2.2901e-04) (hash(x)=168595690)
430
+ 3560 train 4.060184 (lr=2.2626e-04) (hash(x)=158576466)
431
+ 3570 train 4.017753 (lr=2.2354e-04) (hash(x)=157298619)
432
+ 3580 train 4.110590 (lr=2.2085e-04) (hash(x)=160892091)
433
+ 3590 train 3.973662 (lr=2.1818e-04) (hash(x)=165168008)
434
+ 3600 val loss 4.0573
435
+ 3600 val perplexity 57.8203
436
+ 3600 train 4.077959 (lr=2.1555e-04) (hash(x)=144911638)
437
+ 3610 train 4.037361 (lr=2.1294e-04) (hash(x)=150614513)
438
+ 3620 train 4.024478 (lr=2.1036e-04) (hash(x)=167695016)
439
+ 3630 train 3.942936 (lr=2.0781e-04) (hash(x)=182708233)
440
+ 3640 train 4.047745 (lr=2.0529e-04) (hash(x)=154146274)
441
+ 3650 train 4.016556 (lr=2.0280e-04) (hash(x)=159129968)
442
+ 3660 train 3.926732 (lr=2.0033e-04) (hash(x)=146209497)
443
+ 3670 train 3.949498 (lr=1.9790e-04) (hash(x)=145578104)
444
+ 3680 train 3.940126 (lr=1.9550e-04) (hash(x)=150243521)
445
+ 3690 train 3.897836 (lr=1.9313e-04) (hash(x)=154328313)
446
+ 3700 val loss 4.0585
447
+ 3700 val perplexity 57.8864
448
+ 3700 train 3.868111 (lr=1.9079e-04) (hash(x)=159904582)
449
+ 3710 train 4.210571 (lr=1.8848e-04) (hash(x)=175986417)
450
+ 3720 train 4.122455 (lr=1.8619e-04) (hash(x)=158509479)
451
+ 3730 train 4.135634 (lr=1.8395e-04) (hash(x)=156087760)
452
+ 3740 train 4.144529 (lr=1.8173e-04) (hash(x)=177667170)
453
+ 3750 train 4.106473 (lr=1.7954e-04) (hash(x)=145256842)
454
+ 3760 train 4.074240 (lr=1.7738e-04) (hash(x)=149855635)
455
+ 3770 train 4.110331 (lr=1.7526e-04) (hash(x)=168798886)
456
+ 3780 train 4.025065 (lr=1.7317e-04) (hash(x)=156277179)
457
+ 3790 train 3.996068 (lr=1.7111e-04) (hash(x)=149698945)
458
+ 3800 val loss 4.0443
459
+ 3800 val perplexity 57.0698
460
+ 3800 train 3.952396 (lr=1.6908e-04) (hash(x)=168039654)
461
+ 3810 train 4.060846 (lr=1.6708e-04) (hash(x)=154595574)
462
+ 3820 train 4.109933 (lr=1.6511e-04) (hash(x)=158303284)
463
+ 3830 train 4.130737 (lr=1.6318e-04) (hash(x)=164128644)
464
+ 3840 train 4.081497 (lr=1.6128e-04) (hash(x)=166842997)
465
+ 3850 train 3.995714 (lr=1.5942e-04) (hash(x)=177681163)
466
+ 3860 train 3.973776 (lr=1.5758e-04) (hash(x)=177062575)
467
+ 3870 train 4.014819 (lr=1.5578e-04) (hash(x)=158957159)
468
+ 3880 train 3.948854 (lr=1.5401e-04) (hash(x)=151567740)
469
+ 3890 train 4.028297 (lr=1.5228e-04) (hash(x)=153261675)
470
+ 3900 val loss 4.0359
471
+ 3900 val perplexity 56.5945
472
+ 3900 train 4.289242 (lr=1.5058e-04) (hash(x)=160268683)
473
+ 3910 train 4.020354 (lr=1.4891e-04) (hash(x)=153745240)
474
+ 3920 train 4.077204 (lr=1.4727e-04) (hash(x)=164537824)
475
+ 3930 train 4.021866 (lr=1.4567e-04) (hash(x)=162062116)
476
+ 3940 train 3.980704 (lr=1.4411e-04) (hash(x)=168626742)
477
+ 3950 train 4.043627 (lr=1.4257e-04) (hash(x)=148753561)
478
+ 3960 train 3.967503 (lr=1.4108e-04) (hash(x)=163526242)
479
+ 3970 train 4.181691 (lr=1.3961e-04) (hash(x)=159329323)
480
+ 3980 train 4.110963 (lr=1.3818e-04) (hash(x)=146502339)
481
+ 3990 train 3.806314 (lr=1.3678e-04) (hash(x)=158409708)
482
+ 4000 val loss 4.0309
483
+ 4000 val perplexity 56.3121
484
+ 4000 train 3.983942 (lr=1.3542e-04) (hash(x)=154645137)
485
+ 4010 train 3.984140 (lr=1.3410e-04) (hash(x)=160044026)
486
+ 4020 train 3.968870 (lr=1.3280e-04) (hash(x)=155544686)
487
+ 4030 train 3.952895 (lr=1.3155e-04) (hash(x)=147589421)
488
+ 4040 train 4.060854 (lr=1.3032e-04) (hash(x)=158607990)
489
+ 4050 train 3.972727 (lr=1.2914e-04) (hash(x)=154108872)
490
+ 4060 train 4.060097 (lr=1.2798e-04) (hash(x)=167101465)
491
+ 4070 train 4.020504 (lr=1.2687e-04) (hash(x)=160323363)
492
+ 4080 train 4.007334 (lr=1.2578e-04) (hash(x)=150423692)
493
+ 4090 train 3.980417 (lr=1.2474e-04) (hash(x)=167461885)
494
+ 4100 val loss 4.0218
495
+ 4100 val perplexity 55.8020
496
+ 4100 train 4.014617 (lr=1.2373e-04) (hash(x)=151783377)
497
+ 4110 train 4.008676 (lr=1.2275e-04) (hash(x)=165679649)
498
+ 4120 train 3.907412 (lr=1.2181e-04) (hash(x)=153643039)
499
+ 4130 train 3.958738 (lr=1.2091e-04) (hash(x)=149648749)
500
+ 4140 train 3.996169 (lr=1.2004e-04) (hash(x)=162751418)
501
+ 4150 train 3.988765 (lr=1.1920e-04) (hash(x)=157047478)
502
+ 4160 train 3.930102 (lr=1.1841e-04) (hash(x)=156391812)
503
+ 4170 train 4.002793 (lr=1.1764e-04) (hash(x)=180926985)
504
+ 4180 train 4.045938 (lr=1.1692e-04) (hash(x)=168699059)
505
+ 4190 train 3.968961 (lr=1.1623e-04) (hash(x)=155410372)
506
+ 4200 val loss 4.0188
507
+ 4200 val perplexity 55.6334
508
+ 4200 train 3.973525 (lr=1.1557e-04) (hash(x)=152615252)
509
+ 4210 train 3.947087 (lr=1.1496e-04) (hash(x)=155679453)
510
+ 4220 train 4.049628 (lr=1.1437e-04) (hash(x)=152954399)
511
+ 4230 train 4.060158 (lr=1.1383e-04) (hash(x)=164608084)
512
+ 4240 train 4.082090 (lr=1.1332e-04) (hash(x)=162463867)
513
+ 4250 train 3.997517 (lr=1.1285e-04) (hash(x)=168832088)
514
+ 4260 train 3.975362 (lr=1.1241e-04) (hash(x)=162698882)
515
+ 4270 train 4.004496 (lr=1.1201e-04) (hash(x)=180861492)
516
+ 4280 train 3.887822 (lr=1.1164e-04) (hash(x)=160824300)
517
+ 4290 train 3.950050 (lr=1.1132e-04) (hash(x)=157750938)
518
+ 4300 val loss 4.0187
519
+ 4300 val perplexity 55.6273
520
+ 4300 train 3.906092 (lr=1.1103e-04) (hash(x)=157216833)
521
+ 4310 train 3.900333 (lr=1.1077e-04) (hash(x)=176184661)
522
+ 4320 train 4.066880 (lr=1.1055e-04) (hash(x)=161537117)
523
+ 4330 train 4.020887 (lr=1.1037e-04) (hash(x)=160700216)
524
+ 4340 train 3.973645 (lr=1.1022e-04) (hash(x)=175519903)
525
+ 4350 train 3.975668 (lr=1.1011e-04) (hash(x)=167988701)
526
+ 4360 train 3.926420 (lr=1.1004e-04) (hash(x)=162619084)
527
+ 4370 train 4.311178 (lr=1.1000e-04) (hash(x)=166292684)
528
+ 4374 val loss 4.0124
529
+ 4374 val perplexity 55.2775
one_mask_per_head_1_latent_mask/model_04374.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:59e311b17d864dbdd0d5d94e897108b2dc1d56d2945d211122b0650b5c7372aa
3
+ size 95220594
one_mask_per_head_1_latent_mask/optimizer_04374.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:34c288344619d967328ce4004fbb035d226ce20966f39f3e4aa48070967cdc96
3
+ size 184163894