andrew-healey commited on
Commit
dbbc009
·
verified ·
1 Parent(s): 96a9928

Upload folder using huggingface_hub

Browse files
12_head_baseline_lr_12e-4_head_dim_22_fixed/args.json ADDED
@@ -0,0 +1 @@
 
 
1
+ {"hellaswag": true, "attention_kind": "selective", "log_dir": "logs/repro_selective_pattern_rankings/12_head_baseline_lr_12e-4_head_dim_22_fixed", "resume_checkpoint": null, "resume_optimizer": false, "add_a_head": false, "add_head_to_start": true, "new_head_init": "normal", "n_heads": 12, "n_embd": 264, "head_dim": 22, "protect_bos_token": true, "prevent_from_masking_myself": true, "max_steps": 4375, "warmup_steps": 250, "group": "repro_selective_pattern_rankings", "use_wandb": true, "kill_self_after_run": false, "random_seed": 1339, "memory_penalty_epsilon": 0.1, "selection_head_linear_combo": "none", "selection_head_linear_combo_scale": 1.0, "protection_kind": "none", "leaky_relu_alpha": null, "leaky_relu_bias": null, "use_compile": true, "use_mini_model": false, "upload_to_hf": true, "seq_len": 256, "batch_size": 64, "total_batch_size": 131072, "protection_head_scaling_factor": 1.0, "protection_head_bias": 0.0, "n_sliced_masks": null, "n_latent_masks": null, "mask_layernorm": false, "residual_attention_masks": false, "compute_base_shapes": false, "base_shapes_savefile": null, "mup": true, "disable_selection": false, "mup_enable_coord_check_logging": false, "max_lr": 0.0012, "decay_lr": true, "readout_zero_init": false, "query_zero_init": false, "l1_loss": false, "debugpy": false, "key": "12_head_baseline_lr_12e-4_fixed"}
12_head_baseline_lr_12e-4_head_dim_22_fixed/dataloader_04374.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:8ec6682852ed60a9d42cb4047300f65a04f87328c0ad7a4516be84d11b28f216
3
+ size 964
12_head_baseline_lr_12e-4_head_dim_22_fixed/log2.txt ADDED
@@ -0,0 +1,529 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ max_steps: 4375
2
+ 0 val loss 11.4043
3
+ 0 val perplexity 89707.0391
4
+ 0 train 11.388923 (lr=1.6783e-06) (hash(x)=93356070)
5
+ 10 train 10.320988 (lr=1.8462e-05) (hash(x)=91253010)
6
+ 20 train 9.859467 (lr=3.5245e-05) (hash(x)=74934453)
7
+ 30 train 9.540303 (lr=5.2028e-05) (hash(x)=79278034)
8
+ 40 train 9.044521 (lr=6.8811e-05) (hash(x)=80107892)
9
+ 50 train 8.542803 (lr=8.5594e-05) (hash(x)=70745428)
10
+ 60 train 8.178747 (lr=1.0238e-04) (hash(x)=80077589)
11
+ 70 train 7.689917 (lr=1.1916e-04) (hash(x)=76213766)
12
+ 80 train 7.646315 (lr=1.3594e-04) (hash(x)=83218328)
13
+ 90 train 7.497573 (lr=1.5273e-04) (hash(x)=74855845)
14
+ 100 val loss 7.4277
15
+ 100 val perplexity 1681.8641
16
+ 100 train 7.280364 (lr=1.6951e-04) (hash(x)=82814902)
17
+ 110 train 7.220215 (lr=1.8629e-04) (hash(x)=84907741)
18
+ 120 train 7.099597 (lr=2.0308e-04) (hash(x)=82613223)
19
+ 130 train 6.941185 (lr=2.1986e-04) (hash(x)=83540876)
20
+ 140 train 6.887495 (lr=2.3664e-04) (hash(x)=75095216)
21
+ 150 train 6.937623 (lr=2.5343e-04) (hash(x)=97190944)
22
+ 160 train 6.826200 (lr=2.7021e-04) (hash(x)=82117809)
23
+ 170 train 6.760761 (lr=2.8699e-04) (hash(x)=70514724)
24
+ 180 train 6.650165 (lr=3.0378e-04) (hash(x)=81029624)
25
+ 190 train 6.568184 (lr=3.2056e-04) (hash(x)=78996841)
26
+ 200 val loss 6.4945
27
+ 200 val perplexity 661.4639
28
+ 200 train 6.500025 (lr=3.3734e-04) (hash(x)=79845097)
29
+ 210 train 6.370492 (lr=3.5413e-04) (hash(x)=78997683)
30
+ 220 train 6.296419 (lr=3.7091e-04) (hash(x)=74895865)
31
+ 230 train 6.440056 (lr=3.8769e-04) (hash(x)=80933276)
32
+ 240 train 6.162028 (lr=4.0448e-04) (hash(x)=77664606)
33
+ 250 train 6.180106 (lr=4.2126e-04) (hash(x)=75903930)
34
+ 260 train 6.073304 (lr=4.3804e-04) (hash(x)=82996853)
35
+ 270 train 6.003132 (lr=4.5483e-04) (hash(x)=73269568)
36
+ 280 train 5.939777 (lr=4.7161e-04) (hash(x)=82906427)
37
+ 290 train 5.858984 (lr=4.8839e-04) (hash(x)=71797895)
38
+ 300 val loss 5.9987
39
+ 300 val perplexity 402.8957
40
+ 300 train 5.724561 (lr=5.0517e-04) (hash(x)=85232249)
41
+ 310 train 5.669612 (lr=5.2196e-04) (hash(x)=75252489)
42
+ 320 train 5.742350 (lr=5.3874e-04) (hash(x)=74394644)
43
+ 330 train 5.832599 (lr=5.5552e-04) (hash(x)=82772910)
44
+ 340 train 5.861310 (lr=5.7231e-04) (hash(x)=81627464)
45
+ 350 train 5.852623 (lr=5.8909e-04) (hash(x)=84678053)
46
+ 360 train 5.786205 (lr=6.0587e-04) (hash(x)=81884128)
47
+ 370 train 5.791491 (lr=6.2266e-04) (hash(x)=76379242)
48
+ 380 train 5.575869 (lr=6.3944e-04) (hash(x)=98182875)
49
+ 390 train 5.573097 (lr=6.5622e-04) (hash(x)=79710436)
50
+ 400 val loss 5.6129
51
+ 400 val perplexity 273.9472
52
+ 400 train 5.560454 (lr=6.7301e-04) (hash(x)=79841071)
53
+ 410 train 5.498624 (lr=6.8979e-04) (hash(x)=75844151)
54
+ 420 train 5.502224 (lr=7.0657e-04) (hash(x)=73125036)
55
+ 430 train 5.537268 (lr=7.2336e-04) (hash(x)=84214858)
56
+ 440 train 5.426658 (lr=7.4014e-04) (hash(x)=80456994)
57
+ 450 train 5.398307 (lr=7.5692e-04) (hash(x)=76735962)
58
+ 460 train 5.320539 (lr=7.7371e-04) (hash(x)=81845446)
59
+ 470 train 5.131890 (lr=7.9049e-04) (hash(x)=76094689)
60
+ 480 train 5.265275 (lr=8.0727e-04) (hash(x)=83806686)
61
+ 490 train 5.078186 (lr=8.2406e-04) (hash(x)=84690227)
62
+ 500 val loss 5.4126
63
+ 500 val perplexity 224.2126
64
+ 500 train 5.166617 (lr=8.4084e-04) (hash(x)=71851938)
65
+ 510 train 5.247060 (lr=8.5762e-04) (hash(x)=77159346)
66
+ 520 train 5.314418 (lr=8.7441e-04) (hash(x)=80755753)
67
+ 530 train 5.409760 (lr=8.9119e-04) (hash(x)=75379679)
68
+ 540 train 5.321963 (lr=9.0797e-04) (hash(x)=82458619)
69
+ 550 train 5.298475 (lr=9.2476e-04) (hash(x)=86366396)
70
+ 560 train 5.299356 (lr=9.4154e-04) (hash(x)=85095044)
71
+ 570 train 5.267879 (lr=9.5832e-04) (hash(x)=78385159)
72
+ 580 train 5.219580 (lr=9.7510e-04) (hash(x)=79342394)
73
+ 590 train 5.229352 (lr=9.9189e-04) (hash(x)=70782192)
74
+ 600 val loss 5.2261
75
+ 600 val perplexity 186.0681
76
+ 600 train 5.205407 (lr=1.0087e-03) (hash(x)=76130353)
77
+ 610 train 5.107973 (lr=1.0255e-03) (hash(x)=74778440)
78
+ 620 train 5.091714 (lr=1.0422e-03) (hash(x)=79129709)
79
+ 630 train 4.987134 (lr=1.0590e-03) (hash(x)=76469962)
80
+ 640 train 4.985141 (lr=1.0758e-03) (hash(x)=78288049)
81
+ 650 train 4.945765 (lr=1.0926e-03) (hash(x)=76641388)
82
+ 660 train 4.876579 (lr=1.1094e-03) (hash(x)=75937906)
83
+ 670 train 4.866247 (lr=1.1262e-03) (hash(x)=74807157)
84
+ 680 train 4.776150 (lr=1.1429e-03) (hash(x)=77490144)
85
+ 690 train 4.803900 (lr=1.1597e-03) (hash(x)=72900124)
86
+ 700 val loss 5.0819
87
+ 700 val perplexity 161.0775
88
+ 700 train 4.768158 (lr=1.1765e-03) (hash(x)=73424218)
89
+ 710 train 5.090592 (lr=1.1933e-03) (hash(x)=86052345)
90
+ 720 train 5.024864 (lr=1.2000e-03) (hash(x)=77462613)
91
+ 730 train 5.027158 (lr=1.2000e-03) (hash(x)=79027471)
92
+ 740 train 4.985702 (lr=1.1999e-03) (hash(x)=78149992)
93
+ 750 train 4.951528 (lr=1.1998e-03) (hash(x)=89147499)
94
+ 760 train 4.957899 (lr=1.1996e-03) (hash(x)=74931538)
95
+ 770 train 4.900444 (lr=1.1994e-03) (hash(x)=89648204)
96
+ 780 train 4.956250 (lr=1.1992e-03) (hash(x)=83708293)
97
+ 790 train 4.969097 (lr=1.1989e-03) (hash(x)=84336512)
98
+ 800 val loss 4.9148
99
+ 800 val perplexity 136.2895
100
+ 800 train 4.834459 (lr=1.1986e-03) (hash(x)=75025285)
101
+ 810 train 4.760211 (lr=1.1982e-03) (hash(x)=79436536)
102
+ 820 train 4.753765 (lr=1.1978e-03) (hash(x)=90467390)
103
+ 830 train 4.779811 (lr=1.1974e-03) (hash(x)=77292020)
104
+ 840 train 4.700965 (lr=1.1969e-03) (hash(x)=75568927)
105
+ 850 train 4.718232 (lr=1.1964e-03) (hash(x)=79671781)
106
+ 860 train 4.803602 (lr=1.1958e-03) (hash(x)=84280943)
107
+ 870 train 4.779204 (lr=1.1952e-03) (hash(x)=80901294)
108
+ 880 train 4.901457 (lr=1.1946e-03) (hash(x)=81437584)
109
+ 890 train 4.754169 (lr=1.1939e-03) (hash(x)=73893778)
110
+ 900 val loss 4.7783
111
+ 900 val perplexity 118.8975
112
+ 900 train 4.701823 (lr=1.1932e-03) (hash(x)=74987794)
113
+ 910 train 4.714493 (lr=1.1925e-03) (hash(x)=93721374)
114
+ 920 train 4.691214 (lr=1.1917e-03) (hash(x)=79149678)
115
+ 930 train 4.646092 (lr=1.1908e-03) (hash(x)=83179387)
116
+ 940 train 4.641201 (lr=1.1900e-03) (hash(x)=89009978)
117
+ 950 train 4.628469 (lr=1.1891e-03) (hash(x)=76483735)
118
+ 960 train 4.558436 (lr=1.1881e-03) (hash(x)=78638850)
119
+ 970 train 4.585842 (lr=1.1871e-03) (hash(x)=90415206)
120
+ 980 train 4.524825 (lr=1.1861e-03) (hash(x)=90819195)
121
+ 990 train 4.650132 (lr=1.1850e-03) (hash(x)=85938852)
122
+ 1000 val loss 4.6362
123
+ 1000 val perplexity 103.1468
124
+ 1000 train 4.684929 (lr=1.1839e-03) (hash(x)=82436789)
125
+ 1010 train 4.718949 (lr=1.1828e-03) (hash(x)=63022149)
126
+ 1020 train 4.652229 (lr=1.1816e-03) (hash(x)=83501199)
127
+ 1030 train 4.626018 (lr=1.1804e-03) (hash(x)=77093326)
128
+ 1040 train 4.503421 (lr=1.1791e-03) (hash(x)=73437559)
129
+ 1050 train 4.598055 (lr=1.1778e-03) (hash(x)=84550388)
130
+ 1060 train 4.592630 (lr=1.1765e-03) (hash(x)=86184566)
131
+ 1070 train 4.582988 (lr=1.1751e-03) (hash(x)=85644922)
132
+ 1080 train 4.516943 (lr=1.1737e-03) (hash(x)=86179801)
133
+ 1090 train 4.487288 (lr=1.1723e-03) (hash(x)=82868303)
134
+ 1100 val loss 4.5552
135
+ 1100 val perplexity 95.1249
136
+ 1100 train 4.586456 (lr=1.1708e-03) (hash(x)=96780388)
137
+ 1110 train 4.523304 (lr=1.1693e-03) (hash(x)=87223122)
138
+ 1120 train 4.484558 (lr=1.1677e-03) (hash(x)=77292786)
139
+ 1130 train 4.512111 (lr=1.1661e-03) (hash(x)=88761375)
140
+ 1140 train 4.425292 (lr=1.1645e-03) (hash(x)=80956468)
141
+ 1150 train 4.600706 (lr=1.1628e-03) (hash(x)=84725462)
142
+ 1160 train 4.584165 (lr=1.1611e-03) (hash(x)=76098113)
143
+ 1170 train 4.553464 (lr=1.1593e-03) (hash(x)=84387685)
144
+ 1180 train 4.499890 (lr=1.1576e-03) (hash(x)=76120321)
145
+ 1190 train 4.461966 (lr=1.1557e-03) (hash(x)=85974065)
146
+ 1200 val loss 4.4771
147
+ 1200 val perplexity 87.9824
148
+ 1200 train 4.357356 (lr=1.1539e-03) (hash(x)=79340644)
149
+ 1210 train 4.454120 (lr=1.1520e-03) (hash(x)=88345025)
150
+ 1220 train 4.363471 (lr=1.1501e-03) (hash(x)=86406230)
151
+ 1230 train 4.355181 (lr=1.1481e-03) (hash(x)=80608975)
152
+ 1240 train 4.410853 (lr=1.1461e-03) (hash(x)=81635225)
153
+ 1250 train 4.249072 (lr=1.1441e-03) (hash(x)=82126376)
154
+ 1260 train 4.414835 (lr=1.1420e-03) (hash(x)=83737972)
155
+ 1270 train 4.322860 (lr=1.1399e-03) (hash(x)=81132345)
156
+ 1280 train 4.416454 (lr=1.1377e-03) (hash(x)=85850404)
157
+ 1290 train 4.426085 (lr=1.1356e-03) (hash(x)=83490640)
158
+ 1300 val loss 4.4300
159
+ 1300 val perplexity 83.9286
160
+ 1300 train 4.398197 (lr=1.1333e-03) (hash(x)=77545187)
161
+ 1310 train 4.448874 (lr=1.1311e-03) (hash(x)=86412685)
162
+ 1320 train 4.343222 (lr=1.1288e-03) (hash(x)=86429640)
163
+ 1330 train 4.482787 (lr=1.1265e-03) (hash(x)=79645281)
164
+ 1340 train 4.437263 (lr=1.1241e-03) (hash(x)=82668541)
165
+ 1350 train 4.295218 (lr=1.1218e-03) (hash(x)=73627248)
166
+ 1360 train 4.391268 (lr=1.1193e-03) (hash(x)=74845133)
167
+ 1370 train 4.304973 (lr=1.1169e-03) (hash(x)=77850497)
168
+ 1380 train 4.330341 (lr=1.1144e-03) (hash(x)=76143954)
169
+ 1390 train 4.339757 (lr=1.1119e-03) (hash(x)=80675544)
170
+ 1400 val loss 4.4037
171
+ 1400 val perplexity 81.7533
172
+ 1400 train 4.310254 (lr=1.1093e-03) (hash(x)=76010938)
173
+ 1410 train 4.363564 (lr=1.1067e-03) (hash(x)=81260300)
174
+ 1420 train 4.473086 (lr=1.1041e-03) (hash(x)=83803629)
175
+ 1430 train 4.303625 (lr=1.1015e-03) (hash(x)=83840110)
176
+ 1440 train 4.499125 (lr=1.0988e-03) (hash(x)=97711831)
177
+ 1450 train 4.370738 (lr=1.0961e-03) (hash(x)=81110139)
178
+ 1460 train 4.287063 (lr=1.0933e-03) (hash(x)=90942839)
179
+ 1470 train 4.329041 (lr=1.0905e-03) (hash(x)=80721564)
180
+ 1480 train 4.421760 (lr=1.0877e-03) (hash(x)=77852059)
181
+ 1490 train 4.318762 (lr=1.0849e-03) (hash(x)=80011365)
182
+ 1500 val loss 4.3593
183
+ 1500 val perplexity 78.2000
184
+ 1500 train 4.345961 (lr=1.0820e-03) (hash(x)=84496142)
185
+ 1510 train 4.282455 (lr=1.0791e-03) (hash(x)=70186729)
186
+ 1520 train 4.305882 (lr=1.0761e-03) (hash(x)=74854227)
187
+ 1530 train 4.210709 (lr=1.0732e-03) (hash(x)=77468161)
188
+ 1540 train 4.232373 (lr=1.0702e-03) (hash(x)=87853059)
189
+ 1550 train 4.163240 (lr=1.0671e-03) (hash(x)=71225436)
190
+ 1560 train 4.262134 (lr=1.0641e-03) (hash(x)=84072783)
191
+ 1570 train 4.312644 (lr=1.0610e-03) (hash(x)=72723098)
192
+ 1580 train 4.325500 (lr=1.0579e-03) (hash(x)=90409866)
193
+ 1590 train 4.402428 (lr=1.0547e-03) (hash(x)=87481378)
194
+ 1600 val loss 4.3277
195
+ 1600 val perplexity 75.7705
196
+ 1600 train 4.387433 (lr=1.0515e-03) (hash(x)=77643862)
197
+ 1610 train 4.270691 (lr=1.0483e-03) (hash(x)=88862575)
198
+ 1620 train 4.288310 (lr=1.0451e-03) (hash(x)=84612581)
199
+ 1630 train 4.320275 (lr=1.0418e-03) (hash(x)=87075989)
200
+ 1640 train 4.207576 (lr=1.0385e-03) (hash(x)=88277361)
201
+ 1650 train 4.323485 (lr=1.0352e-03) (hash(x)=78750236)
202
+ 1660 train 4.244304 (lr=1.0319e-03) (hash(x)=82604581)
203
+ 1670 train 4.083309 (lr=1.0285e-03) (hash(x)=68482265)
204
+ 1680 train 4.205079 (lr=1.0251e-03) (hash(x)=75088835)
205
+ 1690 train 4.150445 (lr=1.0217e-03) (hash(x)=66060989)
206
+ 1700 val loss 4.3253
207
+ 1700 val perplexity 75.5851
208
+ 1700 train 4.321501 (lr=1.0182e-03) (hash(x)=79986754)
209
+ 1710 train 4.307797 (lr=1.0147e-03) (hash(x)=83657930)
210
+ 1720 train 4.252460 (lr=1.0112e-03) (hash(x)=81754135)
211
+ 1730 train 4.329279 (lr=1.0077e-03) (hash(x)=78904427)
212
+ 1740 train 4.233785 (lr=1.0041e-03) (hash(x)=85920177)
213
+ 1750 train 4.282675 (lr=1.0006e-03) (hash(x)=86573211)
214
+ 1760 train 4.236879 (lr=9.9694e-04) (hash(x)=81737128)
215
+ 1770 train 4.229470 (lr=9.9331e-04) (hash(x)=77400968)
216
+ 1780 train 4.245444 (lr=9.8965e-04) (hash(x)=73545497)
217
+ 1790 train 4.493844 (lr=9.8597e-04) (hash(x)=71641943)
218
+ 1800 val loss 4.2753
219
+ 1800 val perplexity 71.9044
220
+ 1800 train 4.339240 (lr=9.8226e-04) (hash(x)=87819781)
221
+ 1810 train 4.199331 (lr=9.7853e-04) (hash(x)=86870770)
222
+ 1820 train 4.207456 (lr=9.7477e-04) (hash(x)=82522211)
223
+ 1830 train 4.296036 (lr=9.7100e-04) (hash(x)=61947437)
224
+ 1840 train 4.170052 (lr=9.6719e-04) (hash(x)=79865406)
225
+ 1850 train 4.142939 (lr=9.6337e-04) (hash(x)=79828721)
226
+ 1860 train 4.286947 (lr=9.5953e-04) (hash(x)=80869571)
227
+ 1870 train 4.262673 (lr=9.5566e-04) (hash(x)=73780971)
228
+ 1880 train 4.228723 (lr=9.5177e-04) (hash(x)=79249549)
229
+ 1890 train 4.283707 (lr=9.4786e-04) (hash(x)=81041904)
230
+ 1900 val loss 4.2543
231
+ 1900 val perplexity 70.4076
232
+ 1900 train 4.343535 (lr=9.4393e-04) (hash(x)=82456430)
233
+ 1910 train 4.117029 (lr=9.3997e-04) (hash(x)=82222135)
234
+ 1920 train 4.153120 (lr=9.3600e-04) (hash(x)=70033249)
235
+ 1930 train 4.138804 (lr=9.3200e-04) (hash(x)=72887360)
236
+ 1940 train 4.187682 (lr=9.2799e-04) (hash(x)=83251100)
237
+ 1950 train 4.229957 (lr=9.2396e-04) (hash(x)=79660266)
238
+ 1960 train 4.109115 (lr=9.1990e-04) (hash(x)=78919068)
239
+ 1970 train 4.206552 (lr=9.1583e-04) (hash(x)=94690431)
240
+ 1980 train 4.053576 (lr=9.1174e-04) (hash(x)=88153756)
241
+ 1990 train 4.096329 (lr=9.0763e-04) (hash(x)=77195688)
242
+ 2000 val loss 4.2305
243
+ 2000 val perplexity 68.7528
244
+ 2000 train 4.194915 (lr=9.0350e-04) (hash(x)=81308591)
245
+ 2010 train 4.243697 (lr=8.9936e-04) (hash(x)=80205479)
246
+ 2020 train 4.156584 (lr=8.9519e-04) (hash(x)=81008704)
247
+ 2030 train 4.222412 (lr=8.9101e-04) (hash(x)=82914358)
248
+ 2040 train 4.238915 (lr=8.8681e-04) (hash(x)=88064399)
249
+ 2050 train 4.241400 (lr=8.8260e-04) (hash(x)=82889390)
250
+ 2060 train 4.311831 (lr=8.7837e-04) (hash(x)=80913554)
251
+ 2070 train 4.150552 (lr=8.7412e-04) (hash(x)=78121791)
252
+ 2080 train 4.287324 (lr=8.6986e-04) (hash(x)=75356657)
253
+ 2090 train 4.107797 (lr=8.6558e-04) (hash(x)=74048412)
254
+ 2100 val loss 4.2056
255
+ 2100 val perplexity 67.0626
256
+ 2100 train 4.028209 (lr=8.6128e-04) (hash(x)=68928225)
257
+ 2110 train 4.084877 (lr=8.5698e-04) (hash(x)=93734745)
258
+ 2120 train 4.080456 (lr=8.5265e-04) (hash(x)=72878164)
259
+ 2130 train 4.210809 (lr=8.4832e-04) (hash(x)=71508226)
260
+ 2140 train 4.165883 (lr=8.4396e-04) (hash(x)=84168671)
261
+ 2150 train 4.098943 (lr=8.3960e-04) (hash(x)=74673639)
262
+ 2160 train 4.254272 (lr=8.3522e-04) (hash(x)=75470031)
263
+ 2170 train 4.213081 (lr=8.3083e-04) (hash(x)=74307890)
264
+ 2180 train 4.327392 (lr=8.2643e-04) (hash(x)=77214245)
265
+ 2190 train 4.107833 (lr=8.2201e-04) (hash(x)=88628359)
266
+ 2200 val loss 4.1851
267
+ 2200 val perplexity 65.7022
268
+ 2200 train 4.120584 (lr=8.1759e-04) (hash(x)=74779126)
269
+ 2210 train 4.159240 (lr=8.1315e-04) (hash(x)=79817976)
270
+ 2220 train 4.115835 (lr=8.0870e-04) (hash(x)=75258996)
271
+ 2230 train 4.085845 (lr=8.0423e-04) (hash(x)=88164047)
272
+ 2240 train 4.068602 (lr=7.9976e-04) (hash(x)=78580686)
273
+ 2250 train 4.008166 (lr=7.9528e-04) (hash(x)=76510617)
274
+ 2260 train 4.086077 (lr=7.9079e-04) (hash(x)=76043966)
275
+ 2270 train 4.061698 (lr=7.8629e-04) (hash(x)=73062098)
276
+ 2280 train 4.290317 (lr=7.8177e-04) (hash(x)=81885909)
277
+ 2290 train 4.221574 (lr=7.7725e-04) (hash(x)=81481741)
278
+ 2300 val loss 4.1632
279
+ 2300 val perplexity 64.2775
280
+ 2300 train 4.082384 (lr=7.7273e-04) (hash(x)=82104275)
281
+ 2310 train 4.132091 (lr=7.6819e-04) (hash(x)=82137309)
282
+ 2320 train 4.107729 (lr=7.6364e-04) (hash(x)=86741167)
283
+ 2330 train 4.122018 (lr=7.5909e-04) (hash(x)=84796263)
284
+ 2340 train 4.091414 (lr=7.5453e-04) (hash(x)=169966529)
285
+ 2350 train 3.972918 (lr=7.4996e-04) (hash(x)=80006061)
286
+ 2360 train 4.034052 (lr=7.4539e-04) (hash(x)=71090999)
287
+ 2370 train 4.160900 (lr=7.4081e-04) (hash(x)=86641850)
288
+ 2380 train 4.058736 (lr=7.3622e-04) (hash(x)=84573253)
289
+ 2390 train 4.092303 (lr=7.3163e-04) (hash(x)=74483764)
290
+ 2400 val loss 4.1563
291
+ 2400 val perplexity 63.8335
292
+ 2400 train 4.067576 (lr=7.2704e-04) (hash(x)=78327659)
293
+ 2410 train 4.130264 (lr=7.2243e-04) (hash(x)=77273627)
294
+ 2420 train 3.993855 (lr=7.1783e-04) (hash(x)=76938049)
295
+ 2430 train 4.001208 (lr=7.1322e-04) (hash(x)=98449442)
296
+ 2440 train 3.857067 (lr=7.0860e-04) (hash(x)=78157797)
297
+ 2450 train 3.901651 (lr=7.0399e-04) (hash(x)=80637582)
298
+ 2460 train 3.890954 (lr=6.9936e-04) (hash(x)=80023854)
299
+ 2470 train 3.947004 (lr=6.9474e-04) (hash(x)=86678884)
300
+ 2480 train 4.173583 (lr=6.9011e-04) (hash(x)=84086469)
301
+ 2490 train 4.191680 (lr=6.8548e-04) (hash(x)=83131141)
302
+ 2500 val loss 4.1303
303
+ 2500 val perplexity 62.1954
304
+ 2500 train 4.106811 (lr=6.8085e-04) (hash(x)=82583497)
305
+ 2510 train 4.170501 (lr=6.7622e-04) (hash(x)=79164326)
306
+ 2520 train 4.076901 (lr=6.7159e-04) (hash(x)=76456503)
307
+ 2530 train 4.137943 (lr=6.6695e-04) (hash(x)=82001228)
308
+ 2540 train 4.171067 (lr=6.6232e-04) (hash(x)=75615595)
309
+ 2550 train 4.117160 (lr=6.5768e-04) (hash(x)=78397869)
310
+ 2560 train 4.118159 (lr=6.5305e-04) (hash(x)=75547032)
311
+ 2570 train 4.029750 (lr=6.4841e-04) (hash(x)=89201025)
312
+ 2580 train 4.019063 (lr=6.4378e-04) (hash(x)=89856704)
313
+ 2590 train 4.026861 (lr=6.3915e-04) (hash(x)=82175682)
314
+ 2600 val loss 4.1146
315
+ 2600 val perplexity 61.2299
316
+ 2600 train 4.033970 (lr=6.3452e-04) (hash(x)=78312826)
317
+ 2610 train 4.029638 (lr=6.2989e-04) (hash(x)=77066588)
318
+ 2620 train 3.869991 (lr=6.2526e-04) (hash(x)=78666061)
319
+ 2630 train 3.864846 (lr=6.2064e-04) (hash(x)=93762143)
320
+ 2640 train 3.881523 (lr=6.1601e-04) (hash(x)=83191587)
321
+ 2650 train 3.803981 (lr=6.1140e-04) (hash(x)=87169585)
322
+ 2660 train 4.019794 (lr=6.0678e-04) (hash(x)=86426388)
323
+ 2670 train 4.160552 (lr=6.0217e-04) (hash(x)=76692638)
324
+ 2680 train 4.076728 (lr=5.9757e-04) (hash(x)=77446063)
325
+ 2690 train 4.073716 (lr=5.9296e-04) (hash(x)=79809050)
326
+ 2700 val loss 4.1053
327
+ 2700 val perplexity 60.6625
328
+ 2700 train 3.997258 (lr=5.8837e-04) (hash(x)=83116823)
329
+ 2710 train 4.115988 (lr=5.8378e-04) (hash(x)=75622148)
330
+ 2720 train 4.041673 (lr=5.7919e-04) (hash(x)=80690512)
331
+ 2730 train 4.112023 (lr=5.7461e-04) (hash(x)=78009984)
332
+ 2740 train 4.067030 (lr=5.7004e-04) (hash(x)=74205488)
333
+ 2750 train 4.070109 (lr=5.6547e-04) (hash(x)=91013332)
334
+ 2760 train 4.000547 (lr=5.6091e-04) (hash(x)=76954961)
335
+ 2770 train 4.002381 (lr=5.5636e-04) (hash(x)=79390317)
336
+ 2780 train 3.865647 (lr=5.5181e-04) (hash(x)=70168783)
337
+ 2790 train 4.031262 (lr=5.4727e-04) (hash(x)=76028417)
338
+ 2800 val loss 4.0888
339
+ 2800 val perplexity 59.6665
340
+ 2800 train 3.862155 (lr=5.4275e-04) (hash(x)=77656050)
341
+ 2810 train 3.906030 (lr=5.3823e-04) (hash(x)=94248216)
342
+ 2820 train 3.968856 (lr=5.3371e-04) (hash(x)=78305078)
343
+ 2830 train 3.885490 (lr=5.2921e-04) (hash(x)=79948848)
344
+ 2840 train 3.749341 (lr=5.2472e-04) (hash(x)=85341024)
345
+ 2850 train 4.055304 (lr=5.2024e-04) (hash(x)=78735170)
346
+ 2860 train 4.305359 (lr=5.1577e-04) (hash(x)=71616419)
347
+ 2870 train 4.076962 (lr=5.1130e-04) (hash(x)=78656517)
348
+ 2880 train 4.090511 (lr=5.0685e-04) (hash(x)=80073987)
349
+ 2890 train 4.023489 (lr=5.0241e-04) (hash(x)=76894809)
350
+ 2900 val loss 4.0639
351
+ 2900 val perplexity 58.2019
352
+ 2900 train 4.002770 (lr=4.9799e-04) (hash(x)=80499838)
353
+ 2910 train 4.053166 (lr=4.9357e-04) (hash(x)=72673354)
354
+ 2920 train 4.034405 (lr=4.8917e-04) (hash(x)=84265768)
355
+ 2930 train 4.064227 (lr=4.8478e-04) (hash(x)=79612060)
356
+ 2940 train 3.913123 (lr=4.8040e-04) (hash(x)=74970087)
357
+ 2950 train 4.073566 (lr=4.7604e-04) (hash(x)=84166818)
358
+ 2960 train 4.060753 (lr=4.7168e-04) (hash(x)=89410221)
359
+ 2970 train 4.001463 (lr=4.6735e-04) (hash(x)=75672566)
360
+ 2980 train 3.853348 (lr=4.6302e-04) (hash(x)=81760314)
361
+ 2990 train 3.827268 (lr=4.5872e-04) (hash(x)=80605200)
362
+ 3000 val loss 4.0631
363
+ 3000 val perplexity 58.1523
364
+ 3000 train 3.826334 (lr=4.5442e-04) (hash(x)=83804735)
365
+ 3010 train 3.750636 (lr=4.5014e-04) (hash(x)=77015303)
366
+ 3020 train 3.824512 (lr=4.4588e-04) (hash(x)=81464523)
367
+ 3030 train 4.018211 (lr=4.4163e-04) (hash(x)=80168230)
368
+ 3040 train 3.958422 (lr=4.3740e-04) (hash(x)=84817006)
369
+ 3050 train 4.066715 (lr=4.3319e-04) (hash(x)=71601811)
370
+ 3060 train 3.987371 (lr=4.2899e-04) (hash(x)=85499733)
371
+ 3070 train 4.043922 (lr=4.2481e-04) (hash(x)=78670408)
372
+ 3080 train 3.933670 (lr=4.2064e-04) (hash(x)=77120468)
373
+ 3090 train 4.045580 (lr=4.1650e-04) (hash(x)=77927426)
374
+ 3100 val loss 4.0449
375
+ 3100 val perplexity 57.1067
376
+ 3100 train 4.063379 (lr=4.1237e-04) (hash(x)=83998606)
377
+ 3110 train 3.958214 (lr=4.0826e-04) (hash(x)=77990218)
378
+ 3120 train 3.872759 (lr=4.0417e-04) (hash(x)=81623970)
379
+ 3130 train 3.842596 (lr=4.0010e-04) (hash(x)=75014781)
380
+ 3140 train 3.902738 (lr=3.9604e-04) (hash(x)=72591250)
381
+ 3150 train 3.913351 (lr=3.9201e-04) (hash(x)=81421847)
382
+ 3160 train 3.962830 (lr=3.8800e-04) (hash(x)=75112631)
383
+ 3170 train 4.058686 (lr=3.8400e-04) (hash(x)=87518033)
384
+ 3180 train 3.914291 (lr=3.8003e-04) (hash(x)=78648348)
385
+ 3190 train 4.027238 (lr=3.7607e-04) (hash(x)=78270029)
386
+ 3200 val loss 4.0276
387
+ 3200 val perplexity 56.1245
388
+ 3200 train 4.103144 (lr=3.7214e-04) (hash(x)=83380714)
389
+ 3210 train 4.280261 (lr=3.6823e-04) (hash(x)=67804991)
390
+ 3220 train 4.011722 (lr=3.6434e-04) (hash(x)=76968804)
391
+ 3230 train 3.942992 (lr=3.6047e-04) (hash(x)=75798670)
392
+ 3240 train 4.022935 (lr=3.5663e-04) (hash(x)=83192811)
393
+ 3250 train 3.957005 (lr=3.5281e-04) (hash(x)=81659789)
394
+ 3260 train 3.998947 (lr=3.4900e-04) (hash(x)=80622502)
395
+ 3270 train 3.970541 (lr=3.4523e-04) (hash(x)=85436511)
396
+ 3280 train 3.912679 (lr=3.4147e-04) (hash(x)=80589180)
397
+ 3290 train 4.050023 (lr=3.3774e-04) (hash(x)=85747193)
398
+ 3300 val loss 4.0219
399
+ 3300 val perplexity 55.8085
400
+ 3300 train 3.920900 (lr=3.3403e-04) (hash(x)=74635692)
401
+ 3310 train 3.908852 (lr=3.3035e-04) (hash(x)=82045455)
402
+ 3320 train 3.777364 (lr=3.2669e-04) (hash(x)=73249173)
403
+ 3330 train 3.737711 (lr=3.2306e-04) (hash(x)=71492338)
404
+ 3340 train 4.068841 (lr=3.1945e-04) (hash(x)=76374071)
405
+ 3350 train 4.038703 (lr=3.1586e-04) (hash(x)=80474064)
406
+ 3360 train 3.951686 (lr=3.1230e-04) (hash(x)=84390892)
407
+ 3370 train 3.997801 (lr=3.0877e-04) (hash(x)=83399949)
408
+ 3380 train 4.044109 (lr=3.0526e-04) (hash(x)=77648059)
409
+ 3390 train 4.017727 (lr=3.0178e-04) (hash(x)=80152701)
410
+ 3400 val loss 4.0030
411
+ 3400 val perplexity 54.7618
412
+ 3400 train 4.065221 (lr=2.9833e-04) (hash(x)=81952545)
413
+ 3410 train 4.024789 (lr=2.9490e-04) (hash(x)=80908993)
414
+ 3420 train 4.068767 (lr=2.9149e-04) (hash(x)=79610037)
415
+ 3430 train 3.933214 (lr=2.8812e-04) (hash(x)=87624382)
416
+ 3440 train 3.991707 (lr=2.8477e-04) (hash(x)=82336381)
417
+ 3450 train 3.968160 (lr=2.8145e-04) (hash(x)=87687835)
418
+ 3460 train 3.972480 (lr=2.7816e-04) (hash(x)=76376135)
419
+ 3470 train 3.887361 (lr=2.7490e-04) (hash(x)=79331391)
420
+ 3480 train 3.792932 (lr=2.7166e-04) (hash(x)=107002681)
421
+ 3490 train 3.874091 (lr=2.6846e-04) (hash(x)=77465514)
422
+ 3500 val loss 4.0006
423
+ 3500 val perplexity 54.6323
424
+ 3500 train 3.834973 (lr=2.6528e-04) (hash(x)=88237229)
425
+ 3510 train 3.903488 (lr=2.6213e-04) (hash(x)=81011739)
426
+ 3520 train 3.950758 (lr=2.5901e-04) (hash(x)=64643427)
427
+ 3530 train 4.231858 (lr=2.5592e-04) (hash(x)=78029539)
428
+ 3540 train 4.062687 (lr=2.5286e-04) (hash(x)=83188968)
429
+ 3550 train 4.081599 (lr=2.4983e-04) (hash(x)=86104185)
430
+ 3560 train 4.008024 (lr=2.4683e-04) (hash(x)=81993629)
431
+ 3570 train 3.979062 (lr=2.4386e-04) (hash(x)=72719368)
432
+ 3580 train 3.986550 (lr=2.4093e-04) (hash(x)=90091487)
433
+ 3590 train 3.874116 (lr=2.3802e-04) (hash(x)=77393152)
434
+ 3600 val loss 3.9823
435
+ 3600 val perplexity 53.6401
436
+ 3600 train 3.985742 (lr=2.3514e-04) (hash(x)=73103504)
437
+ 3610 train 3.991526 (lr=2.3230e-04) (hash(x)=84429400)
438
+ 3620 train 3.964933 (lr=2.2948e-04) (hash(x)=77021795)
439
+ 3630 train 3.920199 (lr=2.2670e-04) (hash(x)=92830605)
440
+ 3640 train 4.110807 (lr=2.2395e-04) (hash(x)=78313175)
441
+ 3650 train 3.913970 (lr=2.2123e-04) (hash(x)=85395549)
442
+ 3660 train 3.710909 (lr=2.1855e-04) (hash(x)=78114459)
443
+ 3670 train 3.835298 (lr=2.1589e-04) (hash(x)=74968316)
444
+ 3680 train 3.762897 (lr=2.1327e-04) (hash(x)=73358737)
445
+ 3690 train 3.871409 (lr=2.1068e-04) (hash(x)=76399442)
446
+ 3700 val loss 3.9841
447
+ 3700 val perplexity 53.7360
448
+ 3700 train 3.873192 (lr=2.0813e-04) (hash(x)=74039273)
449
+ 3710 train 4.042569 (lr=2.0561e-04) (hash(x)=79299680)
450
+ 3720 train 3.961908 (lr=2.0312e-04) (hash(x)=94701498)
451
+ 3730 train 4.034923 (lr=2.0067e-04) (hash(x)=75352071)
452
+ 3740 train 3.915213 (lr=1.9825e-04) (hash(x)=77909487)
453
+ 3750 train 4.017241 (lr=1.9586e-04) (hash(x)=73986730)
454
+ 3760 train 3.979407 (lr=1.9351e-04) (hash(x)=79325763)
455
+ 3770 train 4.028449 (lr=1.9119e-04) (hash(x)=72457818)
456
+ 3780 train 4.165260 (lr=1.8891e-04) (hash(x)=71775590)
457
+ 3790 train 4.047787 (lr=1.8666e-04) (hash(x)=82638943)
458
+ 3800 val loss 3.9683
459
+ 3800 val perplexity 52.8968
460
+ 3800 train 4.022591 (lr=1.8445e-04) (hash(x)=79965893)
461
+ 3810 train 3.955323 (lr=1.8227e-04) (hash(x)=72598235)
462
+ 3820 train 3.942094 (lr=1.8013e-04) (hash(x)=83113889)
463
+ 3830 train 3.996552 (lr=1.7802e-04) (hash(x)=74434590)
464
+ 3840 train 3.917689 (lr=1.7595e-04) (hash(x)=82860348)
465
+ 3850 train 3.851673 (lr=1.7391e-04) (hash(x)=78067565)
466
+ 3860 train 3.922087 (lr=1.7191e-04) (hash(x)=82592498)
467
+ 3870 train 3.856236 (lr=1.6994e-04) (hash(x)=81820733)
468
+ 3880 train 3.869605 (lr=1.6801e-04) (hash(x)=87709040)
469
+ 3890 train 3.914414 (lr=1.6612e-04) (hash(x)=70379093)
470
+ 3900 val loss 3.9627
471
+ 3900 val perplexity 52.5977
472
+ 3900 train 4.035837 (lr=1.6427e-04) (hash(x)=76597431)
473
+ 3910 train 3.921295 (lr=1.6245e-04) (hash(x)=90490716)
474
+ 3920 train 4.024273 (lr=1.6066e-04) (hash(x)=81970659)
475
+ 3930 train 4.013020 (lr=1.5892e-04) (hash(x)=81496334)
476
+ 3940 train 3.984635 (lr=1.5721e-04) (hash(x)=75717605)
477
+ 3950 train 3.934394 (lr=1.5554e-04) (hash(x)=82749357)
478
+ 3960 train 3.966752 (lr=1.5390e-04) (hash(x)=89355157)
479
+ 3970 train 3.921142 (lr=1.5230e-04) (hash(x)=78980403)
480
+ 3980 train 3.894027 (lr=1.5074e-04) (hash(x)=76627217)
481
+ 3990 train 3.892213 (lr=1.4922e-04) (hash(x)=72412879)
482
+ 4000 val loss 3.9576
483
+ 4000 val perplexity 52.3341
484
+ 4000 train 3.907177 (lr=1.4773e-04) (hash(x)=83018142)
485
+ 4010 train 3.923580 (lr=1.4629e-04) (hash(x)=81272436)
486
+ 4020 train 3.854667 (lr=1.4488e-04) (hash(x)=85497482)
487
+ 4030 train 3.929371 (lr=1.4351e-04) (hash(x)=85241734)
488
+ 4040 train 3.891002 (lr=1.4217e-04) (hash(x)=77925307)
489
+ 4050 train 3.889904 (lr=1.4088e-04) (hash(x)=84826179)
490
+ 4060 train 3.948967 (lr=1.3962e-04) (hash(x)=83606764)
491
+ 4070 train 3.860100 (lr=1.3840e-04) (hash(x)=80567590)
492
+ 4080 train 3.912318 (lr=1.3722e-04) (hash(x)=76860998)
493
+ 4090 train 3.949124 (lr=1.3608e-04) (hash(x)=74902328)
494
+ 4100 val loss 3.9477
495
+ 4100 val perplexity 51.8135
496
+ 4100 train 3.993071 (lr=1.3497e-04) (hash(x)=82832041)
497
+ 4110 train 3.874076 (lr=1.3391e-04) (hash(x)=79143262)
498
+ 4120 train 3.900429 (lr=1.3288e-04) (hash(x)=77038149)
499
+ 4130 train 4.080311 (lr=1.3190e-04) (hash(x)=86339074)
500
+ 4140 train 4.147448 (lr=1.3095e-04) (hash(x)=76686216)
501
+ 4150 train 3.923403 (lr=1.3004e-04) (hash(x)=70522682)
502
+ 4160 train 3.889619 (lr=1.2917e-04) (hash(x)=90958555)
503
+ 4170 train 3.911567 (lr=1.2834e-04) (hash(x)=91463532)
504
+ 4180 train 3.899094 (lr=1.2755e-04) (hash(x)=81959329)
505
+ 4190 train 3.857088 (lr=1.2679e-04) (hash(x)=83146752)
506
+ 4200 val loss 3.9496
507
+ 4200 val perplexity 51.9136
508
+ 4200 train 3.840910 (lr=1.2608e-04) (hash(x)=78361715)
509
+ 4210 train 3.988470 (lr=1.2541e-04) (hash(x)=87364889)
510
+ 4220 train 3.843263 (lr=1.2477e-04) (hash(x)=70465156)
511
+ 4230 train 3.992702 (lr=1.2418e-04) (hash(x)=84524081)
512
+ 4240 train 3.952963 (lr=1.2362e-04) (hash(x)=77824868)
513
+ 4250 train 3.965195 (lr=1.2311e-04) (hash(x)=81710711)
514
+ 4260 train 3.941341 (lr=1.2263e-04) (hash(x)=76362728)
515
+ 4270 train 3.940145 (lr=1.2219e-04) (hash(x)=83115208)
516
+ 4280 train 3.848452 (lr=1.2179e-04) (hash(x)=87218314)
517
+ 4290 train 4.010308 (lr=1.2144e-04) (hash(x)=74582673)
518
+ 4300 val loss 3.9448
519
+ 4300 val perplexity 51.6648
520
+ 4300 train 3.888010 (lr=1.2112e-04) (hash(x)=77379615)
521
+ 4310 train 3.860597 (lr=1.2084e-04) (hash(x)=78669579)
522
+ 4320 train 3.978181 (lr=1.2060e-04) (hash(x)=83066608)
523
+ 4330 train 3.941510 (lr=1.2040e-04) (hash(x)=83037340)
524
+ 4340 train 3.983102 (lr=1.2024e-04) (hash(x)=82849771)
525
+ 4350 train 3.915341 (lr=1.2012e-04) (hash(x)=76693985)
526
+ 4360 train 4.004005 (lr=1.2004e-04) (hash(x)=77745394)
527
+ 4370 train 3.967594 (lr=1.2000e-04) (hash(x)=79954388)
528
+ 4374 val loss 3.9386
529
+ 4374 val perplexity 51.3482
12_head_baseline_lr_12e-4_head_dim_22_fixed/model_04374.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:5d02f0a59678c4dc5f59b3fd6f6fcd0775a54ef72796213ecd6d86fd111363b7
3
+ size 96858242
12_head_baseline_lr_12e-4_head_dim_22_fixed/optimizer_04374.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:51b448e8a584b06668f2fa757fb6cfdb4929448bb9d2946ee62527c9dc473af7
3
+ size 187435910