andrew-healey commited on
Commit
8fa1751
·
verified ·
1 Parent(s): c3d8aba

Upload folder using huggingface_hub

Browse files
12_head_baseline_lr_16e-4/args.json ADDED
@@ -0,0 +1 @@
 
 
1
+ {"hellaswag": true, "attention_kind": "selective", "log_dir": "logs/repro_selective_pattern_rankings/12_head_baseline_lr_16e-4", "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, "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": 0.0016, "decay_lr": true, "readout_zero_init": false, "query_zero_init": false, "l1_loss": false, "debugpy": false, "key": "12_head_baseline_lr_16e-4"}
12_head_baseline_lr_16e-4/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_16e-4/log2.txt ADDED
@@ -0,0 +1,529 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ max_steps: 4375
2
+ 0 val loss 10.9275
3
+ 0 val perplexity 55688.2578
4
+ 0 train 10.927715 (lr=2.2378e-06) (hash(x)=93356070)
5
+ 10 train 9.798237 (lr=2.4615e-05) (hash(x)=91253010)
6
+ 20 train 9.396315 (lr=4.6993e-05) (hash(x)=74934453)
7
+ 30 train 9.005478 (lr=6.9371e-05) (hash(x)=79278034)
8
+ 40 train 8.433649 (lr=9.1748e-05) (hash(x)=80107892)
9
+ 50 train 7.974592 (lr=1.1413e-04) (hash(x)=70745428)
10
+ 60 train 7.783370 (lr=1.3650e-04) (hash(x)=80077589)
11
+ 70 train 7.477903 (lr=1.5888e-04) (hash(x)=76213766)
12
+ 80 train 7.492089 (lr=1.8126e-04) (hash(x)=83218328)
13
+ 90 train 7.285089 (lr=2.0364e-04) (hash(x)=74855845)
14
+ 100 val loss 7.1958
15
+ 100 val perplexity 1333.7958
16
+ 100 train 7.032207 (lr=2.2601e-04) (hash(x)=82814902)
17
+ 110 train 6.953409 (lr=2.4839e-04) (hash(x)=84907741)
18
+ 120 train 6.804421 (lr=2.7077e-04) (hash(x)=82613223)
19
+ 130 train 6.645586 (lr=2.9315e-04) (hash(x)=83540876)
20
+ 140 train 6.586178 (lr=3.1552e-04) (hash(x)=75095216)
21
+ 150 train 6.651823 (lr=3.3790e-04) (hash(x)=97190944)
22
+ 160 train 6.560573 (lr=3.6028e-04) (hash(x)=82117809)
23
+ 170 train 6.456620 (lr=3.8266e-04) (hash(x)=70514724)
24
+ 180 train 6.387156 (lr=4.0503e-04) (hash(x)=81029624)
25
+ 190 train 6.303118 (lr=4.2741e-04) (hash(x)=78996841)
26
+ 200 val loss 6.2306
27
+ 200 val perplexity 508.0386
28
+ 200 train 6.240286 (lr=4.4979e-04) (hash(x)=79845097)
29
+ 210 train 6.120396 (lr=4.7217e-04) (hash(x)=78997683)
30
+ 220 train 6.037217 (lr=4.9455e-04) (hash(x)=74895865)
31
+ 230 train 6.181069 (lr=5.1692e-04) (hash(x)=80933276)
32
+ 240 train 5.878169 (lr=5.3930e-04) (hash(x)=77664606)
33
+ 250 train 5.922624 (lr=5.6168e-04) (hash(x)=75903930)
34
+ 260 train 5.814422 (lr=5.8406e-04) (hash(x)=82996853)
35
+ 270 train 5.751627 (lr=6.0643e-04) (hash(x)=73269568)
36
+ 280 train 5.699580 (lr=6.2881e-04) (hash(x)=82906427)
37
+ 290 train 5.609686 (lr=6.5119e-04) (hash(x)=71797895)
38
+ 300 val loss 5.7720
39
+ 300 val perplexity 321.1844
40
+ 300 train 5.478294 (lr=6.7357e-04) (hash(x)=85232249)
41
+ 310 train 5.390877 (lr=6.9594e-04) (hash(x)=75252489)
42
+ 320 train 5.487942 (lr=7.1832e-04) (hash(x)=74394644)
43
+ 330 train 5.577361 (lr=7.4070e-04) (hash(x)=82772910)
44
+ 340 train 5.647011 (lr=7.6308e-04) (hash(x)=81627464)
45
+ 350 train 5.597541 (lr=7.8545e-04) (hash(x)=84678053)
46
+ 360 train 5.519506 (lr=8.0783e-04) (hash(x)=81884128)
47
+ 370 train 5.539505 (lr=8.3021e-04) (hash(x)=76379242)
48
+ 380 train 5.352305 (lr=8.5259e-04) (hash(x)=98182875)
49
+ 390 train 5.305471 (lr=8.7497e-04) (hash(x)=79710436)
50
+ 400 val loss 5.3553
51
+ 400 val perplexity 211.7274
52
+ 400 train 5.303134 (lr=8.9734e-04) (hash(x)=79841071)
53
+ 410 train 5.253006 (lr=9.1972e-04) (hash(x)=75844151)
54
+ 420 train 5.238636 (lr=9.4210e-04) (hash(x)=73125036)
55
+ 430 train 5.275555 (lr=9.6448e-04) (hash(x)=84214858)
56
+ 440 train 5.152685 (lr=9.8685e-04) (hash(x)=80456994)
57
+ 450 train 5.137425 (lr=1.0092e-03) (hash(x)=76735962)
58
+ 460 train 5.050288 (lr=1.0316e-03) (hash(x)=81845446)
59
+ 470 train 4.852437 (lr=1.0540e-03) (hash(x)=76094689)
60
+ 480 train 4.947817 (lr=1.0764e-03) (hash(x)=83806686)
61
+ 490 train 4.764601 (lr=1.0987e-03) (hash(x)=84690227)
62
+ 500 val loss 5.1016
63
+ 500 val perplexity 164.2830
64
+ 500 train 4.842937 (lr=1.1211e-03) (hash(x)=71851938)
65
+ 510 train 4.935117 (lr=1.1435e-03) (hash(x)=77159346)
66
+ 520 train 5.004507 (lr=1.1659e-03) (hash(x)=80755753)
67
+ 530 train 5.050716 (lr=1.1883e-03) (hash(x)=75379679)
68
+ 540 train 4.985985 (lr=1.2106e-03) (hash(x)=82458619)
69
+ 550 train 4.935608 (lr=1.2330e-03) (hash(x)=86366396)
70
+ 560 train 4.980645 (lr=1.2554e-03) (hash(x)=85095044)
71
+ 570 train 4.892474 (lr=1.2778e-03) (hash(x)=78385159)
72
+ 580 train 4.830591 (lr=1.3001e-03) (hash(x)=79342394)
73
+ 590 train 4.841479 (lr=1.3225e-03) (hash(x)=70782192)
74
+ 600 val loss 4.8460
75
+ 600 val perplexity 127.2367
76
+ 600 train 4.819918 (lr=1.3449e-03) (hash(x)=76130353)
77
+ 610 train 4.748949 (lr=1.3673e-03) (hash(x)=74778440)
78
+ 620 train 4.723595 (lr=1.3897e-03) (hash(x)=79129709)
79
+ 630 train 4.627557 (lr=1.4120e-03) (hash(x)=76469962)
80
+ 640 train 4.624539 (lr=1.4344e-03) (hash(x)=78288049)
81
+ 650 train 4.556910 (lr=1.4568e-03) (hash(x)=76641388)
82
+ 660 train 4.469276 (lr=1.4792e-03) (hash(x)=75937906)
83
+ 670 train 4.465406 (lr=1.5015e-03) (hash(x)=74807157)
84
+ 680 train 4.367145 (lr=1.5239e-03) (hash(x)=77490144)
85
+ 690 train 4.409160 (lr=1.5463e-03) (hash(x)=72900124)
86
+ 700 val loss 4.6771
87
+ 700 val perplexity 107.4606
88
+ 700 train 4.367651 (lr=1.5687e-03) (hash(x)=73424218)
89
+ 710 train 4.661384 (lr=1.5910e-03) (hash(x)=86052345)
90
+ 720 train 4.639750 (lr=1.6000e-03) (hash(x)=77462613)
91
+ 730 train 4.650421 (lr=1.5999e-03) (hash(x)=79027471)
92
+ 740 train 4.596417 (lr=1.5998e-03) (hash(x)=78149992)
93
+ 750 train 4.533423 (lr=1.5997e-03) (hash(x)=89147499)
94
+ 760 train 4.602837 (lr=1.5995e-03) (hash(x)=74931538)
95
+ 770 train 4.533976 (lr=1.5992e-03) (hash(x)=89648204)
96
+ 780 train 4.588074 (lr=1.5989e-03) (hash(x)=83708293)
97
+ 790 train 4.583754 (lr=1.5985e-03) (hash(x)=84336512)
98
+ 800 val loss 4.5473
99
+ 800 val perplexity 94.3780
100
+ 800 train 4.473948 (lr=1.5981e-03) (hash(x)=75025285)
101
+ 810 train 4.394310 (lr=1.5976e-03) (hash(x)=79436536)
102
+ 820 train 4.396888 (lr=1.5971e-03) (hash(x)=90467390)
103
+ 830 train 4.431841 (lr=1.5965e-03) (hash(x)=77292020)
104
+ 840 train 4.356587 (lr=1.5959e-03) (hash(x)=75568927)
105
+ 850 train 4.378909 (lr=1.5952e-03) (hash(x)=79671781)
106
+ 860 train 4.464846 (lr=1.5944e-03) (hash(x)=84280943)
107
+ 870 train 4.462914 (lr=1.5936e-03) (hash(x)=80901294)
108
+ 880 train 4.570021 (lr=1.5928e-03) (hash(x)=81437584)
109
+ 890 train 4.426832 (lr=1.5919e-03) (hash(x)=73893778)
110
+ 900 val loss 4.4433
111
+ 900 val perplexity 85.0512
112
+ 900 train 4.379714 (lr=1.5909e-03) (hash(x)=74987794)
113
+ 910 train 4.395721 (lr=1.5899e-03) (hash(x)=93721374)
114
+ 920 train 4.380036 (lr=1.5889e-03) (hash(x)=79149678)
115
+ 930 train 4.338592 (lr=1.5878e-03) (hash(x)=83179387)
116
+ 940 train 4.333689 (lr=1.5866e-03) (hash(x)=89009978)
117
+ 950 train 4.338483 (lr=1.5854e-03) (hash(x)=76483735)
118
+ 960 train 4.254809 (lr=1.5841e-03) (hash(x)=78638850)
119
+ 970 train 4.295341 (lr=1.5828e-03) (hash(x)=90415206)
120
+ 980 train 4.239851 (lr=1.5815e-03) (hash(x)=90819195)
121
+ 990 train 4.368494 (lr=1.5800e-03) (hash(x)=85938852)
122
+ 1000 val loss 4.3452
123
+ 1000 val perplexity 77.1087
124
+ 1000 train 4.391242 (lr=1.5786e-03) (hash(x)=82436789)
125
+ 1010 train 4.479742 (lr=1.5770e-03) (hash(x)=63022149)
126
+ 1020 train 4.357800 (lr=1.5755e-03) (hash(x)=83501199)
127
+ 1030 train 4.340155 (lr=1.5738e-03) (hash(x)=77093326)
128
+ 1040 train 4.214557 (lr=1.5722e-03) (hash(x)=73437559)
129
+ 1050 train 4.304566 (lr=1.5704e-03) (hash(x)=84550388)
130
+ 1060 train 4.308388 (lr=1.5687e-03) (hash(x)=86184566)
131
+ 1070 train 4.290480 (lr=1.5668e-03) (hash(x)=85644922)
132
+ 1080 train 4.228671 (lr=1.5650e-03) (hash(x)=86179801)
133
+ 1090 train 4.205384 (lr=1.5630e-03) (hash(x)=82868303)
134
+ 1100 val loss 4.2725
135
+ 1100 val perplexity 71.7021
136
+ 1100 train 4.297600 (lr=1.5610e-03) (hash(x)=96780388)
137
+ 1110 train 4.243422 (lr=1.5590e-03) (hash(x)=87223122)
138
+ 1120 train 4.218210 (lr=1.5569e-03) (hash(x)=77292786)
139
+ 1130 train 4.236127 (lr=1.5548e-03) (hash(x)=88761375)
140
+ 1140 train 4.137029 (lr=1.5526e-03) (hash(x)=80956468)
141
+ 1150 train 4.311957 (lr=1.5504e-03) (hash(x)=84725462)
142
+ 1160 train 4.298216 (lr=1.5481e-03) (hash(x)=76098113)
143
+ 1170 train 4.277210 (lr=1.5458e-03) (hash(x)=84387685)
144
+ 1180 train 4.219014 (lr=1.5434e-03) (hash(x)=76120321)
145
+ 1190 train 4.184316 (lr=1.5410e-03) (hash(x)=85974065)
146
+ 1200 val loss 4.2007
147
+ 1200 val perplexity 66.7335
148
+ 1200 train 4.087433 (lr=1.5385e-03) (hash(x)=79340644)
149
+ 1210 train 4.184216 (lr=1.5360e-03) (hash(x)=88345025)
150
+ 1220 train 4.083379 (lr=1.5334e-03) (hash(x)=86406230)
151
+ 1230 train 4.076100 (lr=1.5308e-03) (hash(x)=80608975)
152
+ 1240 train 4.142020 (lr=1.5281e-03) (hash(x)=81635225)
153
+ 1250 train 3.971007 (lr=1.5254e-03) (hash(x)=82126376)
154
+ 1260 train 4.150791 (lr=1.5226e-03) (hash(x)=83737972)
155
+ 1270 train 4.057129 (lr=1.5198e-03) (hash(x)=81132345)
156
+ 1280 train 4.139013 (lr=1.5170e-03) (hash(x)=85850404)
157
+ 1290 train 4.141372 (lr=1.5141e-03) (hash(x)=83490640)
158
+ 1300 val loss 4.1526
159
+ 1300 val perplexity 63.5984
160
+ 1300 train 4.123569 (lr=1.5111e-03) (hash(x)=77545187)
161
+ 1310 train 4.170946 (lr=1.5081e-03) (hash(x)=86412685)
162
+ 1320 train 4.069434 (lr=1.5051e-03) (hash(x)=86429640)
163
+ 1330 train 4.196823 (lr=1.5020e-03) (hash(x)=79645281)
164
+ 1340 train 4.155946 (lr=1.4989e-03) (hash(x)=82668541)
165
+ 1350 train 4.026730 (lr=1.4957e-03) (hash(x)=73627248)
166
+ 1360 train 4.132848 (lr=1.4924e-03) (hash(x)=74845133)
167
+ 1370 train 4.035576 (lr=1.4892e-03) (hash(x)=77850497)
168
+ 1380 train 4.057953 (lr=1.4859e-03) (hash(x)=76143954)
169
+ 1390 train 4.067673 (lr=1.4825e-03) (hash(x)=80675544)
170
+ 1400 val loss 4.1263
171
+ 1400 val perplexity 61.9487
172
+ 1400 train 4.039493 (lr=1.4791e-03) (hash(x)=76010938)
173
+ 1410 train 4.072568 (lr=1.4756e-03) (hash(x)=81260300)
174
+ 1420 train 4.190105 (lr=1.4721e-03) (hash(x)=83803629)
175
+ 1430 train 4.024634 (lr=1.4686e-03) (hash(x)=83840110)
176
+ 1440 train 4.205920 (lr=1.4650e-03) (hash(x)=97711831)
177
+ 1450 train 4.081547 (lr=1.4614e-03) (hash(x)=81110139)
178
+ 1460 train 4.019911 (lr=1.4577e-03) (hash(x)=90942839)
179
+ 1470 train 4.056212 (lr=1.4540e-03) (hash(x)=80721564)
180
+ 1480 train 4.151441 (lr=1.4503e-03) (hash(x)=77852059)
181
+ 1490 train 4.045312 (lr=1.4465e-03) (hash(x)=80011365)
182
+ 1500 val loss 4.0766
183
+ 1500 val perplexity 58.9421
184
+ 1500 train 4.066617 (lr=1.4426e-03) (hash(x)=84496142)
185
+ 1510 train 4.018937 (lr=1.4388e-03) (hash(x)=70186729)
186
+ 1520 train 4.036461 (lr=1.4348e-03) (hash(x)=74854227)
187
+ 1530 train 3.943403 (lr=1.4309e-03) (hash(x)=77468161)
188
+ 1540 train 3.954084 (lr=1.4269e-03) (hash(x)=87853059)
189
+ 1550 train 3.896334 (lr=1.4228e-03) (hash(x)=71225436)
190
+ 1560 train 3.991151 (lr=1.4188e-03) (hash(x)=84072783)
191
+ 1570 train 4.035669 (lr=1.4147e-03) (hash(x)=72723098)
192
+ 1580 train 4.057684 (lr=1.4105e-03) (hash(x)=90409866)
193
+ 1590 train 4.134654 (lr=1.4063e-03) (hash(x)=87481378)
194
+ 1600 val loss 4.0487
195
+ 1600 val perplexity 57.3250
196
+ 1600 train 4.122609 (lr=1.4021e-03) (hash(x)=77643862)
197
+ 1610 train 3.985092 (lr=1.3978e-03) (hash(x)=88862575)
198
+ 1620 train 4.015446 (lr=1.3935e-03) (hash(x)=84612581)
199
+ 1630 train 4.048377 (lr=1.3891e-03) (hash(x)=87075989)
200
+ 1640 train 3.935255 (lr=1.3847e-03) (hash(x)=88277361)
201
+ 1650 train 4.055773 (lr=1.3803e-03) (hash(x)=78750236)
202
+ 1660 train 3.963664 (lr=1.3758e-03) (hash(x)=82604581)
203
+ 1670 train 3.810824 (lr=1.3713e-03) (hash(x)=68482265)
204
+ 1680 train 3.942805 (lr=1.3668e-03) (hash(x)=75088835)
205
+ 1690 train 3.890950 (lr=1.3622e-03) (hash(x)=66060989)
206
+ 1700 val loss 4.0408
207
+ 1700 val perplexity 56.8725
208
+ 1700 train 4.039040 (lr=1.3576e-03) (hash(x)=79986754)
209
+ 1710 train 4.011769 (lr=1.3530e-03) (hash(x)=83657930)
210
+ 1720 train 3.953026 (lr=1.3483e-03) (hash(x)=81754135)
211
+ 1730 train 4.062155 (lr=1.3436e-03) (hash(x)=78904427)
212
+ 1740 train 3.945317 (lr=1.3389e-03) (hash(x)=85920177)
213
+ 1750 train 4.003551 (lr=1.3341e-03) (hash(x)=86573211)
214
+ 1760 train 3.965342 (lr=1.3293e-03) (hash(x)=81737128)
215
+ 1770 train 3.954886 (lr=1.3244e-03) (hash(x)=77400968)
216
+ 1780 train 3.967692 (lr=1.3195e-03) (hash(x)=73545497)
217
+ 1790 train 4.235353 (lr=1.3146e-03) (hash(x)=71641943)
218
+ 1800 val loss 3.9927
219
+ 1800 val perplexity 54.2015
220
+ 1800 train 4.073539 (lr=1.3097e-03) (hash(x)=87819781)
221
+ 1810 train 3.925332 (lr=1.3047e-03) (hash(x)=86870770)
222
+ 1820 train 3.933777 (lr=1.2997e-03) (hash(x)=82522211)
223
+ 1830 train 4.032731 (lr=1.2947e-03) (hash(x)=61947437)
224
+ 1840 train 3.897365 (lr=1.2896e-03) (hash(x)=79865406)
225
+ 1850 train 3.855808 (lr=1.2845e-03) (hash(x)=79828721)
226
+ 1860 train 3.989238 (lr=1.2794e-03) (hash(x)=80869571)
227
+ 1870 train 3.983275 (lr=1.2742e-03) (hash(x)=73780971)
228
+ 1880 train 3.937798 (lr=1.2690e-03) (hash(x)=79249549)
229
+ 1890 train 3.997724 (lr=1.2638e-03) (hash(x)=81041904)
230
+ 1900 val loss 3.9662
231
+ 1900 val perplexity 52.7818
232
+ 1900 train 4.077607 (lr=1.2586e-03) (hash(x)=82456430)
233
+ 1910 train 3.830054 (lr=1.2533e-03) (hash(x)=82222135)
234
+ 1920 train 3.878909 (lr=1.2480e-03) (hash(x)=70033249)
235
+ 1930 train 3.865349 (lr=1.2427e-03) (hash(x)=72887360)
236
+ 1940 train 3.905010 (lr=1.2373e-03) (hash(x)=83251100)
237
+ 1950 train 3.947239 (lr=1.2319e-03) (hash(x)=79660266)
238
+ 1960 train 3.834379 (lr=1.2265e-03) (hash(x)=78919068)
239
+ 1970 train 3.944201 (lr=1.2211e-03) (hash(x)=94690431)
240
+ 1980 train 3.759509 (lr=1.2157e-03) (hash(x)=88153756)
241
+ 1990 train 3.819460 (lr=1.2102e-03) (hash(x)=77195688)
242
+ 2000 val loss 3.9420
243
+ 2000 val perplexity 51.5205
244
+ 2000 train 3.900645 (lr=1.2047e-03) (hash(x)=81308591)
245
+ 2010 train 3.956593 (lr=1.1991e-03) (hash(x)=80205479)
246
+ 2020 train 3.866267 (lr=1.1936e-03) (hash(x)=81008704)
247
+ 2030 train 3.939819 (lr=1.1880e-03) (hash(x)=82914358)
248
+ 2040 train 3.945701 (lr=1.1824e-03) (hash(x)=88064399)
249
+ 2050 train 3.958043 (lr=1.1768e-03) (hash(x)=82889390)
250
+ 2060 train 4.031107 (lr=1.1712e-03) (hash(x)=80913554)
251
+ 2070 train 3.861921 (lr=1.1655e-03) (hash(x)=78121791)
252
+ 2080 train 4.000007 (lr=1.1598e-03) (hash(x)=75356657)
253
+ 2090 train 3.839210 (lr=1.1541e-03) (hash(x)=74048412)
254
+ 2100 val loss 3.9144
255
+ 2100 val perplexity 50.1215
256
+ 2100 train 3.750068 (lr=1.1484e-03) (hash(x)=68928225)
257
+ 2110 train 3.805278 (lr=1.1426e-03) (hash(x)=93734745)
258
+ 2120 train 3.806856 (lr=1.1369e-03) (hash(x)=72878164)
259
+ 2130 train 3.934175 (lr=1.1311e-03) (hash(x)=71508226)
260
+ 2140 train 3.885393 (lr=1.1253e-03) (hash(x)=84168671)
261
+ 2150 train 3.812306 (lr=1.1195e-03) (hash(x)=74673639)
262
+ 2160 train 3.959175 (lr=1.1136e-03) (hash(x)=75470031)
263
+ 2170 train 3.916228 (lr=1.1078e-03) (hash(x)=74307890)
264
+ 2180 train 4.036257 (lr=1.1019e-03) (hash(x)=77214245)
265
+ 2190 train 3.817413 (lr=1.0960e-03) (hash(x)=88628359)
266
+ 2200 val loss 3.8925
267
+ 2200 val perplexity 49.0355
268
+ 2200 train 3.845771 (lr=1.0901e-03) (hash(x)=74779126)
269
+ 2210 train 3.885524 (lr=1.0842e-03) (hash(x)=79817976)
270
+ 2220 train 3.830046 (lr=1.0783e-03) (hash(x)=75258996)
271
+ 2230 train 3.814875 (lr=1.0723e-03) (hash(x)=88164047)
272
+ 2240 train 3.783941 (lr=1.0663e-03) (hash(x)=78580686)
273
+ 2250 train 3.712483 (lr=1.0604e-03) (hash(x)=76510617)
274
+ 2260 train 3.798095 (lr=1.0544e-03) (hash(x)=76043966)
275
+ 2270 train 3.773125 (lr=1.0484e-03) (hash(x)=73062098)
276
+ 2280 train 3.993357 (lr=1.0424e-03) (hash(x)=81885909)
277
+ 2290 train 3.936231 (lr=1.0363e-03) (hash(x)=81481741)
278
+ 2300 val loss 3.8664
279
+ 2300 val perplexity 47.7709
280
+ 2300 train 3.787530 (lr=1.0303e-03) (hash(x)=82104275)
281
+ 2310 train 3.833493 (lr=1.0243e-03) (hash(x)=82137309)
282
+ 2320 train 3.807029 (lr=1.0182e-03) (hash(x)=86741167)
283
+ 2330 train 3.828388 (lr=1.0121e-03) (hash(x)=84796263)
284
+ 2340 train 3.813998 (lr=1.0060e-03) (hash(x)=169966529)
285
+ 2350 train 3.694920 (lr=9.9995e-04) (hash(x)=80006061)
286
+ 2360 train 3.722418 (lr=9.9385e-04) (hash(x)=71090999)
287
+ 2370 train 3.860324 (lr=9.8775e-04) (hash(x)=86641850)
288
+ 2380 train 3.756783 (lr=9.8163e-04) (hash(x)=84573253)
289
+ 2390 train 3.801387 (lr=9.7551e-04) (hash(x)=74483764)
290
+ 2400 val loss 3.8533
291
+ 2400 val perplexity 47.1489
292
+ 2400 train 3.771990 (lr=9.6938e-04) (hash(x)=78327659)
293
+ 2410 train 3.843506 (lr=9.6325e-04) (hash(x)=77273627)
294
+ 2420 train 3.702762 (lr=9.5710e-04) (hash(x)=76938049)
295
+ 2430 train 3.704423 (lr=9.5096e-04) (hash(x)=98449442)
296
+ 2440 train 3.565814 (lr=9.4480e-04) (hash(x)=78157797)
297
+ 2450 train 3.601907 (lr=9.3865e-04) (hash(x)=80637582)
298
+ 2460 train 3.584134 (lr=9.3248e-04) (hash(x)=80023854)
299
+ 2470 train 3.657673 (lr=9.2632e-04) (hash(x)=86678884)
300
+ 2480 train 3.872123 (lr=9.2015e-04) (hash(x)=84086469)
301
+ 2490 train 3.875577 (lr=9.1398e-04) (hash(x)=83131141)
302
+ 2500 val loss 3.8284
303
+ 2500 val perplexity 45.9879
304
+ 2500 train 3.797068 (lr=9.0780e-04) (hash(x)=82583497)
305
+ 2510 train 3.868979 (lr=9.0163e-04) (hash(x)=79164326)
306
+ 2520 train 3.779405 (lr=8.9545e-04) (hash(x)=76456503)
307
+ 2530 train 3.836169 (lr=8.8927e-04) (hash(x)=82001228)
308
+ 2540 train 3.863338 (lr=8.8309e-04) (hash(x)=75615595)
309
+ 2550 train 3.820961 (lr=8.7691e-04) (hash(x)=78397869)
310
+ 2560 train 3.824438 (lr=8.7073e-04) (hash(x)=75547032)
311
+ 2570 train 3.729818 (lr=8.6455e-04) (hash(x)=89201025)
312
+ 2580 train 3.710370 (lr=8.5837e-04) (hash(x)=89856704)
313
+ 2590 train 3.738738 (lr=8.5220e-04) (hash(x)=82175682)
314
+ 2600 val loss 3.8086
315
+ 2600 val perplexity 45.0855
316
+ 2600 train 3.756269 (lr=8.4602e-04) (hash(x)=78312826)
317
+ 2610 train 3.725816 (lr=8.3985e-04) (hash(x)=77066588)
318
+ 2620 train 3.563008 (lr=8.3368e-04) (hash(x)=78666061)
319
+ 2630 train 3.563302 (lr=8.2752e-04) (hash(x)=93762143)
320
+ 2640 train 3.595202 (lr=8.2135e-04) (hash(x)=83191587)
321
+ 2650 train 3.499893 (lr=8.1520e-04) (hash(x)=87169585)
322
+ 2660 train 3.705549 (lr=8.0904e-04) (hash(x)=86426388)
323
+ 2670 train 3.836143 (lr=8.0290e-04) (hash(x)=76692638)
324
+ 2680 train 3.756015 (lr=7.9675e-04) (hash(x)=77446063)
325
+ 2690 train 3.762798 (lr=7.9062e-04) (hash(x)=79809050)
326
+ 2700 val loss 3.8000
327
+ 2700 val perplexity 44.7008
328
+ 2700 train 3.690426 (lr=7.8449e-04) (hash(x)=83116823)
329
+ 2710 train 3.803932 (lr=7.7837e-04) (hash(x)=75622148)
330
+ 2720 train 3.742031 (lr=7.7225e-04) (hash(x)=80690512)
331
+ 2730 train 3.815257 (lr=7.6615e-04) (hash(x)=78009984)
332
+ 2740 train 3.762790 (lr=7.6005e-04) (hash(x)=74205488)
333
+ 2750 train 3.771251 (lr=7.5396e-04) (hash(x)=91013332)
334
+ 2760 train 3.689183 (lr=7.4788e-04) (hash(x)=76954961)
335
+ 2770 train 3.705366 (lr=7.4181e-04) (hash(x)=79390317)
336
+ 2780 train 3.578420 (lr=7.3575e-04) (hash(x)=70168783)
337
+ 2790 train 3.755299 (lr=7.2970e-04) (hash(x)=76028417)
338
+ 2800 val loss 3.7776
339
+ 2800 val perplexity 43.7097
340
+ 2800 train 3.569932 (lr=7.2366e-04) (hash(x)=77656050)
341
+ 2810 train 3.598113 (lr=7.1763e-04) (hash(x)=94248216)
342
+ 2820 train 3.667575 (lr=7.1162e-04) (hash(x)=78305078)
343
+ 2830 train 3.576174 (lr=7.0562e-04) (hash(x)=79948848)
344
+ 2840 train 3.431547 (lr=6.9963e-04) (hash(x)=85341024)
345
+ 2850 train 3.736544 (lr=6.9365e-04) (hash(x)=78735170)
346
+ 2860 train 3.995527 (lr=6.8769e-04) (hash(x)=71616419)
347
+ 2870 train 3.746640 (lr=6.8174e-04) (hash(x)=78656517)
348
+ 2880 train 3.777730 (lr=6.7581e-04) (hash(x)=80073987)
349
+ 2890 train 3.712286 (lr=6.6989e-04) (hash(x)=76894809)
350
+ 2900 val loss 3.7498
351
+ 2900 val perplexity 42.5126
352
+ 2900 train 3.688190 (lr=6.6398e-04) (hash(x)=80499838)
353
+ 2910 train 3.744892 (lr=6.5810e-04) (hash(x)=72673354)
354
+ 2920 train 3.733337 (lr=6.5222e-04) (hash(x)=84265768)
355
+ 2930 train 3.747352 (lr=6.4637e-04) (hash(x)=79612060)
356
+ 2940 train 3.607414 (lr=6.4053e-04) (hash(x)=74970087)
357
+ 2950 train 3.764172 (lr=6.3471e-04) (hash(x)=84166818)
358
+ 2960 train 3.749575 (lr=6.2891e-04) (hash(x)=89410221)
359
+ 2970 train 3.689886 (lr=6.2313e-04) (hash(x)=75672566)
360
+ 2980 train 3.549690 (lr=6.1737e-04) (hash(x)=81760314)
361
+ 2990 train 3.509017 (lr=6.1162e-04) (hash(x)=80605200)
362
+ 3000 val loss 3.7470
363
+ 3000 val perplexity 42.3957
364
+ 3000 train 3.507289 (lr=6.0590e-04) (hash(x)=83804735)
365
+ 3010 train 3.433452 (lr=6.0019e-04) (hash(x)=77015303)
366
+ 3020 train 3.515005 (lr=5.9451e-04) (hash(x)=81464523)
367
+ 3030 train 3.684551 (lr=5.8884e-04) (hash(x)=80168230)
368
+ 3040 train 3.628102 (lr=5.8320e-04) (hash(x)=84817006)
369
+ 3050 train 3.752707 (lr=5.7758e-04) (hash(x)=71601811)
370
+ 3060 train 3.680906 (lr=5.7199e-04) (hash(x)=85499733)
371
+ 3070 train 3.742000 (lr=5.6641e-04) (hash(x)=78670408)
372
+ 3080 train 3.634930 (lr=5.6086e-04) (hash(x)=77120468)
373
+ 3090 train 3.736116 (lr=5.5533e-04) (hash(x)=77927426)
374
+ 3100 val loss 3.7228
375
+ 3100 val perplexity 41.3792
376
+ 3100 train 3.758689 (lr=5.4983e-04) (hash(x)=83998606)
377
+ 3110 train 3.662683 (lr=5.4435e-04) (hash(x)=77990218)
378
+ 3120 train 3.568766 (lr=5.3889e-04) (hash(x)=81623970)
379
+ 3130 train 3.546069 (lr=5.3346e-04) (hash(x)=75014781)
380
+ 3140 train 3.597431 (lr=5.2806e-04) (hash(x)=72591250)
381
+ 3150 train 3.618393 (lr=5.2268e-04) (hash(x)=81421847)
382
+ 3160 train 3.647749 (lr=5.1733e-04) (hash(x)=75112631)
383
+ 3170 train 3.732754 (lr=5.1200e-04) (hash(x)=87518033)
384
+ 3180 train 3.610738 (lr=5.0670e-04) (hash(x)=78648348)
385
+ 3190 train 3.703344 (lr=5.0143e-04) (hash(x)=78270029)
386
+ 3200 val loss 3.7011
387
+ 3200 val perplexity 40.4901
388
+ 3200 train 3.762161 (lr=4.9619e-04) (hash(x)=83380714)
389
+ 3210 train 3.962884 (lr=4.9097e-04) (hash(x)=67804991)
390
+ 3220 train 3.685901 (lr=4.8579e-04) (hash(x)=76968804)
391
+ 3230 train 3.626415 (lr=4.8063e-04) (hash(x)=75798670)
392
+ 3240 train 3.695642 (lr=4.7550e-04) (hash(x)=83192811)
393
+ 3250 train 3.631952 (lr=4.7041e-04) (hash(x)=81659789)
394
+ 3260 train 3.694153 (lr=4.6534e-04) (hash(x)=80622502)
395
+ 3270 train 3.644301 (lr=4.6030e-04) (hash(x)=85436511)
396
+ 3280 train 3.593134 (lr=4.5530e-04) (hash(x)=80589180)
397
+ 3290 train 3.739891 (lr=4.5032e-04) (hash(x)=85747193)
398
+ 3300 val loss 3.6965
399
+ 3300 val perplexity 40.3070
400
+ 3300 train 3.599995 (lr=4.4538e-04) (hash(x)=74635692)
401
+ 3310 train 3.584977 (lr=4.4047e-04) (hash(x)=82045455)
402
+ 3320 train 3.455269 (lr=4.3559e-04) (hash(x)=73249173)
403
+ 3330 train 3.422512 (lr=4.3074e-04) (hash(x)=71492338)
404
+ 3340 train 3.744324 (lr=4.2593e-04) (hash(x)=76374071)
405
+ 3350 train 3.710243 (lr=4.2115e-04) (hash(x)=80474064)
406
+ 3360 train 3.618067 (lr=4.1640e-04) (hash(x)=84390892)
407
+ 3370 train 3.661542 (lr=4.1169e-04) (hash(x)=83399949)
408
+ 3380 train 3.718322 (lr=4.0702e-04) (hash(x)=77648059)
409
+ 3390 train 3.695032 (lr=4.0237e-04) (hash(x)=80152701)
410
+ 3400 val loss 3.6735
411
+ 3400 val perplexity 39.3905
412
+ 3400 train 3.749348 (lr=3.9777e-04) (hash(x)=81952545)
413
+ 3410 train 3.700473 (lr=3.9320e-04) (hash(x)=80908993)
414
+ 3420 train 3.759974 (lr=3.8866e-04) (hash(x)=79610037)
415
+ 3430 train 3.612681 (lr=3.8416e-04) (hash(x)=87624382)
416
+ 3440 train 3.679648 (lr=3.7970e-04) (hash(x)=82336381)
417
+ 3450 train 3.658126 (lr=3.7527e-04) (hash(x)=87687835)
418
+ 3460 train 3.649705 (lr=3.7088e-04) (hash(x)=76376135)
419
+ 3470 train 3.575760 (lr=3.6653e-04) (hash(x)=79331391)
420
+ 3480 train 3.468257 (lr=3.6222e-04) (hash(x)=107002681)
421
+ 3490 train 3.559164 (lr=3.5794e-04) (hash(x)=77465514)
422
+ 3500 val loss 3.6693
423
+ 3500 val perplexity 39.2243
424
+ 3500 train 3.509536 (lr=3.5371e-04) (hash(x)=88237229)
425
+ 3510 train 3.579407 (lr=3.4951e-04) (hash(x)=81011739)
426
+ 3520 train 3.646174 (lr=3.4535e-04) (hash(x)=64643427)
427
+ 3530 train 3.916560 (lr=3.4123e-04) (hash(x)=78029539)
428
+ 3540 train 3.732683 (lr=3.3715e-04) (hash(x)=83188968)
429
+ 3550 train 3.746043 (lr=3.3311e-04) (hash(x)=86104185)
430
+ 3560 train 3.668815 (lr=3.2911e-04) (hash(x)=81993629)
431
+ 3570 train 3.655949 (lr=3.2515e-04) (hash(x)=72719368)
432
+ 3580 train 3.649710 (lr=3.2123e-04) (hash(x)=90091487)
433
+ 3590 train 3.545308 (lr=3.1736e-04) (hash(x)=77393152)
434
+ 3600 val loss 3.6504
435
+ 3600 val perplexity 38.4920
436
+ 3600 train 3.668592 (lr=3.1352e-04) (hash(x)=73103504)
437
+ 3610 train 3.660382 (lr=3.0973e-04) (hash(x)=84429400)
438
+ 3620 train 3.644264 (lr=3.0598e-04) (hash(x)=77021795)
439
+ 3630 train 3.595557 (lr=3.0227e-04) (hash(x)=92830605)
440
+ 3640 train 3.801348 (lr=2.9860e-04) (hash(x)=78313175)
441
+ 3650 train 3.606744 (lr=2.9498e-04) (hash(x)=85395549)
442
+ 3660 train 3.391429 (lr=2.9139e-04) (hash(x)=78114459)
443
+ 3670 train 3.513697 (lr=2.8786e-04) (hash(x)=74968316)
444
+ 3680 train 3.441600 (lr=2.8436e-04) (hash(x)=73358737)
445
+ 3690 train 3.520941 (lr=2.8091e-04) (hash(x)=76399442)
446
+ 3700 val loss 3.6490
447
+ 3700 val perplexity 38.4343
448
+ 3700 train 3.542142 (lr=2.7751e-04) (hash(x)=74039273)
449
+ 3710 train 3.702939 (lr=2.7415e-04) (hash(x)=79299680)
450
+ 3720 train 3.610700 (lr=2.7083e-04) (hash(x)=94701498)
451
+ 3730 train 3.698984 (lr=2.6756e-04) (hash(x)=75352071)
452
+ 3740 train 3.582694 (lr=2.6433e-04) (hash(x)=77909487)
453
+ 3750 train 3.707860 (lr=2.6115e-04) (hash(x)=73986730)
454
+ 3760 train 3.650366 (lr=2.5801e-04) (hash(x)=79325763)
455
+ 3770 train 3.702325 (lr=2.5492e-04) (hash(x)=72457818)
456
+ 3780 train 3.816529 (lr=2.5188e-04) (hash(x)=71775590)
457
+ 3790 train 3.714643 (lr=2.4888e-04) (hash(x)=82638943)
458
+ 3800 val loss 3.6313
459
+ 3800 val perplexity 37.7628
460
+ 3800 train 3.704069 (lr=2.4593e-04) (hash(x)=79965893)
461
+ 3810 train 3.629180 (lr=2.4302e-04) (hash(x)=72598235)
462
+ 3820 train 3.610856 (lr=2.4017e-04) (hash(x)=83113889)
463
+ 3830 train 3.662986 (lr=2.3736e-04) (hash(x)=74434590)
464
+ 3840 train 3.588402 (lr=2.3459e-04) (hash(x)=82860348)
465
+ 3850 train 3.533422 (lr=2.3188e-04) (hash(x)=78067565)
466
+ 3860 train 3.581328 (lr=2.2921e-04) (hash(x)=82592498)
467
+ 3870 train 3.518932 (lr=2.2659e-04) (hash(x)=81820733)
468
+ 3880 train 3.536292 (lr=2.2402e-04) (hash(x)=87709040)
469
+ 3890 train 3.571911 (lr=2.2150e-04) (hash(x)=70379093)
470
+ 3900 val loss 3.6242
471
+ 3900 val perplexity 37.4962
472
+ 3900 train 3.685142 (lr=2.1902e-04) (hash(x)=76597431)
473
+ 3910 train 3.572523 (lr=2.1659e-04) (hash(x)=90490716)
474
+ 3920 train 3.691116 (lr=2.1422e-04) (hash(x)=81970659)
475
+ 3930 train 3.665605 (lr=2.1189e-04) (hash(x)=81496334)
476
+ 3940 train 3.663921 (lr=2.0961e-04) (hash(x)=75717605)
477
+ 3950 train 3.593484 (lr=2.0738e-04) (hash(x)=82749357)
478
+ 3960 train 3.625359 (lr=2.0520e-04) (hash(x)=89355157)
479
+ 3970 train 3.582020 (lr=2.0307e-04) (hash(x)=78980403)
480
+ 3980 train 3.558674 (lr=2.0099e-04) (hash(x)=76627217)
481
+ 3990 train 3.559707 (lr=1.9896e-04) (hash(x)=72412879)
482
+ 4000 val loss 3.6184
483
+ 4000 val perplexity 37.2769
484
+ 4000 train 3.570160 (lr=1.9698e-04) (hash(x)=83018142)
485
+ 4010 train 3.604418 (lr=1.9505e-04) (hash(x)=81272436)
486
+ 4020 train 3.520159 (lr=1.9317e-04) (hash(x)=85497482)
487
+ 4030 train 3.567759 (lr=1.9134e-04) (hash(x)=85241734)
488
+ 4040 train 3.542062 (lr=1.8956e-04) (hash(x)=77925307)
489
+ 4050 train 3.547979 (lr=1.8783e-04) (hash(x)=84826179)
490
+ 4060 train 3.601512 (lr=1.8616e-04) (hash(x)=83606764)
491
+ 4070 train 3.513038 (lr=1.8453e-04) (hash(x)=80567590)
492
+ 4080 train 3.585178 (lr=1.8296e-04) (hash(x)=76860998)
493
+ 4090 train 3.624927 (lr=1.8144e-04) (hash(x)=74902328)
494
+ 4100 val loss 3.6064
495
+ 4100 val perplexity 36.8345
496
+ 4100 train 3.657998 (lr=1.7997e-04) (hash(x)=82832041)
497
+ 4110 train 3.531200 (lr=1.7855e-04) (hash(x)=79143262)
498
+ 4120 train 3.581289 (lr=1.7718e-04) (hash(x)=77038149)
499
+ 4130 train 3.739970 (lr=1.7586e-04) (hash(x)=86339074)
500
+ 4140 train 3.825858 (lr=1.7460e-04) (hash(x)=76686216)
501
+ 4150 train 3.577947 (lr=1.7339e-04) (hash(x)=70522682)
502
+ 4160 train 3.551043 (lr=1.7223e-04) (hash(x)=90958555)
503
+ 4170 train 3.554443 (lr=1.7112e-04) (hash(x)=91463532)
504
+ 4180 train 3.550604 (lr=1.7006e-04) (hash(x)=81959329)
505
+ 4190 train 3.515916 (lr=1.6906e-04) (hash(x)=83146752)
506
+ 4200 val loss 3.6058
507
+ 4200 val perplexity 36.8097
508
+ 4200 train 3.495444 (lr=1.6811e-04) (hash(x)=78361715)
509
+ 4210 train 3.635114 (lr=1.6721e-04) (hash(x)=87364889)
510
+ 4220 train 3.503727 (lr=1.6636e-04) (hash(x)=70465156)
511
+ 4230 train 3.664410 (lr=1.6557e-04) (hash(x)=84524081)
512
+ 4240 train 3.597541 (lr=1.6483e-04) (hash(x)=77824868)
513
+ 4250 train 3.621438 (lr=1.6414e-04) (hash(x)=81710711)
514
+ 4260 train 3.612156 (lr=1.6350e-04) (hash(x)=76362728)
515
+ 4270 train 3.609385 (lr=1.6292e-04) (hash(x)=83115208)
516
+ 4280 train 3.519208 (lr=1.6239e-04) (hash(x)=87218314)
517
+ 4290 train 3.681349 (lr=1.6192e-04) (hash(x)=74582673)
518
+ 4300 val loss 3.6010
519
+ 4300 val perplexity 36.6345
520
+ 4300 train 3.536857 (lr=1.6149e-04) (hash(x)=77379615)
521
+ 4310 train 3.525330 (lr=1.6112e-04) (hash(x)=78669579)
522
+ 4320 train 3.625381 (lr=1.6080e-04) (hash(x)=83066608)
523
+ 4330 train 3.605047 (lr=1.6054e-04) (hash(x)=83037340)
524
+ 4340 train 3.625861 (lr=1.6032e-04) (hash(x)=82849771)
525
+ 4350 train 3.572292 (lr=1.6017e-04) (hash(x)=76693985)
526
+ 4360 train 3.671343 (lr=1.6006e-04) (hash(x)=77745394)
527
+ 4370 train 3.632418 (lr=1.6001e-04) (hash(x)=79954388)
528
+ 4374 val loss 3.5950
529
+ 4374 val perplexity 36.4166
12_head_baseline_lr_16e-4/model_04374.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:573509bd72fd6ae49daa28a60de025938f1c166fdfc7d918e51ef3f112f5597e
3
+ size 498607490
12_head_baseline_lr_16e-4/optimizer_04374.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:3ebf816ac38809b59c43a083cda242d5aae4f0a986609cb581e44b9f0ca9199f
3
+ size 990934406