andrew-healey commited on
Commit
196ecd8
·
verified ·
1 Parent(s): dbbc009

Upload folder using huggingface_hub

Browse files
12_head_baseline_lr_14e-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_14e-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": 0.0014, "decay_lr": true, "readout_zero_init": false, "query_zero_init": false, "l1_loss": false, "debugpy": false, "key": "12_head_baseline_lr_14e-4_fixed"}
12_head_baseline_lr_14e-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_14e-4_head_dim_22_fixed/log2.txt ADDED
@@ -0,0 +1,529 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ max_steps: 4375
2
+ 0 val loss 11.4043
3
+ 0 val perplexity 89706.9531
4
+ 0 train 11.388936 (lr=1.9580e-06) (hash(x)=93356070)
5
+ 10 train 10.269173 (lr=2.1538e-05) (hash(x)=91253010)
6
+ 20 train 9.815946 (lr=4.1119e-05) (hash(x)=74934453)
7
+ 30 train 9.442367 (lr=6.0699e-05) (hash(x)=79278034)
8
+ 40 train 8.884886 (lr=8.0280e-05) (hash(x)=80107892)
9
+ 50 train 8.354688 (lr=9.9860e-05) (hash(x)=70745428)
10
+ 60 train 8.016597 (lr=1.1944e-04) (hash(x)=80077589)
11
+ 70 train 7.594265 (lr=1.3902e-04) (hash(x)=76213766)
12
+ 80 train 7.605256 (lr=1.5860e-04) (hash(x)=83218328)
13
+ 90 train 7.450216 (lr=1.7818e-04) (hash(x)=74855845)
14
+ 100 val loss 7.3799
15
+ 100 val perplexity 1603.4470
16
+ 100 train 7.224006 (lr=1.9776e-04) (hash(x)=82814902)
17
+ 110 train 7.158934 (lr=2.1734e-04) (hash(x)=84907741)
18
+ 120 train 7.035166 (lr=2.3692e-04) (hash(x)=82613223)
19
+ 130 train 6.863042 (lr=2.5650e-04) (hash(x)=83540876)
20
+ 140 train 6.817091 (lr=2.7608e-04) (hash(x)=75095216)
21
+ 150 train 6.869602 (lr=2.9566e-04) (hash(x)=97190944)
22
+ 160 train 6.768077 (lr=3.1524e-04) (hash(x)=82117809)
23
+ 170 train 6.660790 (lr=3.3483e-04) (hash(x)=70514724)
24
+ 180 train 6.581505 (lr=3.5441e-04) (hash(x)=81029624)
25
+ 190 train 6.480536 (lr=3.7399e-04) (hash(x)=78996841)
26
+ 200 val loss 6.4323
27
+ 200 val perplexity 621.6266
28
+ 200 train 6.437474 (lr=3.9357e-04) (hash(x)=79845097)
29
+ 210 train 6.322279 (lr=4.1315e-04) (hash(x)=78997683)
30
+ 220 train 6.230166 (lr=4.3273e-04) (hash(x)=74895865)
31
+ 230 train 6.372021 (lr=4.5231e-04) (hash(x)=80933276)
32
+ 240 train 6.078808 (lr=4.7189e-04) (hash(x)=77664606)
33
+ 250 train 6.117267 (lr=4.9147e-04) (hash(x)=75903930)
34
+ 260 train 6.021396 (lr=5.1105e-04) (hash(x)=82996853)
35
+ 270 train 5.936547 (lr=5.3063e-04) (hash(x)=73269568)
36
+ 280 train 5.883516 (lr=5.5021e-04) (hash(x)=82906427)
37
+ 290 train 5.804130 (lr=5.6979e-04) (hash(x)=71797895)
38
+ 300 val loss 5.9383
39
+ 300 val perplexity 379.3034
40
+ 300 train 5.659815 (lr=5.8937e-04) (hash(x)=85232249)
41
+ 310 train 5.610974 (lr=6.0895e-04) (hash(x)=75252489)
42
+ 320 train 5.693245 (lr=6.2853e-04) (hash(x)=74394644)
43
+ 330 train 5.770323 (lr=6.4811e-04) (hash(x)=82772910)
44
+ 340 train 5.797123 (lr=6.6769e-04) (hash(x)=81627464)
45
+ 350 train 5.795680 (lr=6.8727e-04) (hash(x)=84678053)
46
+ 360 train 5.736790 (lr=7.0685e-04) (hash(x)=81884128)
47
+ 370 train 5.725534 (lr=7.2643e-04) (hash(x)=76379242)
48
+ 380 train 5.534879 (lr=7.4601e-04) (hash(x)=98182875)
49
+ 390 train 5.510308 (lr=7.6559e-04) (hash(x)=79710436)
50
+ 400 val loss 5.5639
51
+ 400 val perplexity 260.8391
52
+ 400 train 5.509996 (lr=7.8517e-04) (hash(x)=79841071)
53
+ 410 train 5.444727 (lr=8.0476e-04) (hash(x)=75844151)
54
+ 420 train 5.462790 (lr=8.2434e-04) (hash(x)=73125036)
55
+ 430 train 5.490418 (lr=8.4392e-04) (hash(x)=84214858)
56
+ 440 train 5.391551 (lr=8.6350e-04) (hash(x)=80456994)
57
+ 450 train 5.362626 (lr=8.8308e-04) (hash(x)=76735962)
58
+ 460 train 5.277441 (lr=9.0266e-04) (hash(x)=81845446)
59
+ 470 train 5.089274 (lr=9.2224e-04) (hash(x)=76094689)
60
+ 480 train 5.220060 (lr=9.4182e-04) (hash(x)=83806686)
61
+ 490 train 5.031739 (lr=9.6140e-04) (hash(x)=84690227)
62
+ 500 val loss 5.3615
63
+ 500 val perplexity 213.0372
64
+ 500 train 5.113383 (lr=9.8098e-04) (hash(x)=71851938)
65
+ 510 train 5.216011 (lr=1.0006e-03) (hash(x)=77159346)
66
+ 520 train 5.263540 (lr=1.0201e-03) (hash(x)=80755753)
67
+ 530 train 5.382292 (lr=1.0397e-03) (hash(x)=75379679)
68
+ 540 train 5.278602 (lr=1.0593e-03) (hash(x)=82458619)
69
+ 550 train 5.262976 (lr=1.0789e-03) (hash(x)=86366396)
70
+ 560 train 5.259624 (lr=1.0985e-03) (hash(x)=85095044)
71
+ 570 train 5.237655 (lr=1.1180e-03) (hash(x)=78385159)
72
+ 580 train 5.183151 (lr=1.1376e-03) (hash(x)=79342394)
73
+ 590 train 5.176814 (lr=1.1572e-03) (hash(x)=70782192)
74
+ 600 val loss 5.1784
75
+ 600 val perplexity 177.3993
76
+ 600 train 5.161437 (lr=1.1768e-03) (hash(x)=76130353)
77
+ 610 train 5.088160 (lr=1.1964e-03) (hash(x)=74778440)
78
+ 620 train 5.048032 (lr=1.2159e-03) (hash(x)=79129709)
79
+ 630 train 4.952912 (lr=1.2355e-03) (hash(x)=76469962)
80
+ 640 train 4.951177 (lr=1.2551e-03) (hash(x)=78288049)
81
+ 650 train 4.911243 (lr=1.2747e-03) (hash(x)=76641388)
82
+ 660 train 4.841191 (lr=1.2943e-03) (hash(x)=75937906)
83
+ 670 train 4.830285 (lr=1.3138e-03) (hash(x)=74807157)
84
+ 680 train 4.733498 (lr=1.3334e-03) (hash(x)=77490144)
85
+ 690 train 4.765818 (lr=1.3530e-03) (hash(x)=72900124)
86
+ 700 val loss 5.0853
87
+ 700 val perplexity 161.6347
88
+ 700 train 4.777063 (lr=1.3726e-03) (hash(x)=73424218)
89
+ 710 train 5.033429 (lr=1.3922e-03) (hash(x)=86052345)
90
+ 720 train 4.980216 (lr=1.4000e-03) (hash(x)=77462613)
91
+ 730 train 4.992654 (lr=1.3999e-03) (hash(x)=79027471)
92
+ 740 train 4.945261 (lr=1.3999e-03) (hash(x)=78149992)
93
+ 750 train 4.912086 (lr=1.3997e-03) (hash(x)=89147499)
94
+ 760 train 4.916355 (lr=1.3995e-03) (hash(x)=74931538)
95
+ 770 train 4.861145 (lr=1.3993e-03) (hash(x)=89648204)
96
+ 780 train 4.906028 (lr=1.3990e-03) (hash(x)=83708293)
97
+ 790 train 4.936550 (lr=1.3987e-03) (hash(x)=84336512)
98
+ 800 val loss 4.8699
99
+ 800 val perplexity 130.3101
100
+ 800 train 4.790729 (lr=1.3983e-03) (hash(x)=75025285)
101
+ 810 train 4.709366 (lr=1.3979e-03) (hash(x)=79436536)
102
+ 820 train 4.701161 (lr=1.3974e-03) (hash(x)=90467390)
103
+ 830 train 4.730155 (lr=1.3969e-03) (hash(x)=77292020)
104
+ 840 train 4.658195 (lr=1.3964e-03) (hash(x)=75568927)
105
+ 850 train 4.683330 (lr=1.3958e-03) (hash(x)=79671781)
106
+ 860 train 4.764938 (lr=1.3951e-03) (hash(x)=84280943)
107
+ 870 train 4.751052 (lr=1.3944e-03) (hash(x)=80901294)
108
+ 880 train 4.870871 (lr=1.3937e-03) (hash(x)=81437584)
109
+ 890 train 4.714358 (lr=1.3929e-03) (hash(x)=73893778)
110
+ 900 val loss 4.7470
111
+ 900 val perplexity 115.2379
112
+ 900 train 4.678545 (lr=1.3921e-03) (hash(x)=74987794)
113
+ 910 train 4.683047 (lr=1.3912e-03) (hash(x)=93721374)
114
+ 920 train 4.659788 (lr=1.3903e-03) (hash(x)=79149678)
115
+ 930 train 4.618722 (lr=1.3893e-03) (hash(x)=83179387)
116
+ 940 train 4.617470 (lr=1.3883e-03) (hash(x)=89009978)
117
+ 950 train 4.603734 (lr=1.3872e-03) (hash(x)=76483735)
118
+ 960 train 4.535385 (lr=1.3861e-03) (hash(x)=78638850)
119
+ 970 train 4.564167 (lr=1.3850e-03) (hash(x)=90415206)
120
+ 980 train 4.506578 (lr=1.3838e-03) (hash(x)=90819195)
121
+ 990 train 4.630904 (lr=1.3825e-03) (hash(x)=85938852)
122
+ 1000 val loss 4.6188
123
+ 1000 val perplexity 101.3773
124
+ 1000 train 4.665431 (lr=1.3812e-03) (hash(x)=82436789)
125
+ 1010 train 4.704231 (lr=1.3799e-03) (hash(x)=63022149)
126
+ 1020 train 4.630691 (lr=1.3785e-03) (hash(x)=83501199)
127
+ 1030 train 4.601912 (lr=1.3771e-03) (hash(x)=77093326)
128
+ 1040 train 4.485398 (lr=1.3756e-03) (hash(x)=73437559)
129
+ 1050 train 4.576779 (lr=1.3741e-03) (hash(x)=84550388)
130
+ 1060 train 4.576516 (lr=1.3726e-03) (hash(x)=86184566)
131
+ 1070 train 4.568489 (lr=1.3710e-03) (hash(x)=85644922)
132
+ 1080 train 4.497795 (lr=1.3693e-03) (hash(x)=86179801)
133
+ 1090 train 4.471739 (lr=1.3676e-03) (hash(x)=82868303)
134
+ 1100 val loss 4.5462
135
+ 1100 val perplexity 94.2772
136
+ 1100 train 4.573340 (lr=1.3659e-03) (hash(x)=96780388)
137
+ 1110 train 4.516178 (lr=1.3641e-03) (hash(x)=87223122)
138
+ 1120 train 4.471232 (lr=1.3623e-03) (hash(x)=77292786)
139
+ 1130 train 4.503595 (lr=1.3604e-03) (hash(x)=88761375)
140
+ 1140 train 4.418395 (lr=1.3585e-03) (hash(x)=80956468)
141
+ 1150 train 4.588790 (lr=1.3566e-03) (hash(x)=84725462)
142
+ 1160 train 4.569863 (lr=1.3546e-03) (hash(x)=76098113)
143
+ 1170 train 4.541380 (lr=1.3526e-03) (hash(x)=84387685)
144
+ 1180 train 4.485799 (lr=1.3505e-03) (hash(x)=76120321)
145
+ 1190 train 4.449147 (lr=1.3484e-03) (hash(x)=85974065)
146
+ 1200 val loss 4.4633
147
+ 1200 val perplexity 86.7724
148
+ 1200 train 4.350276 (lr=1.3462e-03) (hash(x)=79340644)
149
+ 1210 train 4.442726 (lr=1.3440e-03) (hash(x)=88345025)
150
+ 1220 train 4.355567 (lr=1.3417e-03) (hash(x)=86406230)
151
+ 1230 train 4.348337 (lr=1.3394e-03) (hash(x)=80608975)
152
+ 1240 train 4.399980 (lr=1.3371e-03) (hash(x)=81635225)
153
+ 1250 train 4.234939 (lr=1.3347e-03) (hash(x)=82126376)
154
+ 1260 train 4.408626 (lr=1.3323e-03) (hash(x)=83737972)
155
+ 1270 train 4.315550 (lr=1.3299e-03) (hash(x)=81132345)
156
+ 1280 train 4.410818 (lr=1.3274e-03) (hash(x)=85850404)
157
+ 1290 train 4.424012 (lr=1.3248e-03) (hash(x)=83490640)
158
+ 1300 val loss 4.4221
159
+ 1300 val perplexity 83.2715
160
+ 1300 train 4.387352 (lr=1.3222e-03) (hash(x)=77545187)
161
+ 1310 train 4.437013 (lr=1.3196e-03) (hash(x)=86412685)
162
+ 1320 train 4.335728 (lr=1.3169e-03) (hash(x)=86429640)
163
+ 1330 train 4.469718 (lr=1.3142e-03) (hash(x)=79645281)
164
+ 1340 train 4.427654 (lr=1.3115e-03) (hash(x)=82668541)
165
+ 1350 train 4.284605 (lr=1.3087e-03) (hash(x)=73627248)
166
+ 1360 train 4.379375 (lr=1.3059e-03) (hash(x)=74845133)
167
+ 1370 train 4.300525 (lr=1.3030e-03) (hash(x)=77850497)
168
+ 1380 train 4.316377 (lr=1.3001e-03) (hash(x)=76143954)
169
+ 1390 train 4.335210 (lr=1.2972e-03) (hash(x)=80675544)
170
+ 1400 val loss 4.3955
171
+ 1400 val perplexity 81.0856
172
+ 1400 train 4.299941 (lr=1.2942e-03) (hash(x)=76010938)
173
+ 1410 train 4.361124 (lr=1.2912e-03) (hash(x)=81260300)
174
+ 1420 train 4.469017 (lr=1.2881e-03) (hash(x)=83803629)
175
+ 1430 train 4.297252 (lr=1.2850e-03) (hash(x)=83840110)
176
+ 1440 train 4.492138 (lr=1.2819e-03) (hash(x)=97711831)
177
+ 1450 train 4.361652 (lr=1.2787e-03) (hash(x)=81110139)
178
+ 1460 train 4.278702 (lr=1.2755e-03) (hash(x)=90942839)
179
+ 1470 train 4.321516 (lr=1.2723e-03) (hash(x)=80721564)
180
+ 1480 train 4.423439 (lr=1.2690e-03) (hash(x)=77852059)
181
+ 1490 train 4.314808 (lr=1.2657e-03) (hash(x)=80011365)
182
+ 1500 val loss 4.3513
183
+ 1500 val perplexity 77.5786
184
+ 1500 train 4.336928 (lr=1.2623e-03) (hash(x)=84496142)
185
+ 1510 train 4.280108 (lr=1.2589e-03) (hash(x)=70186729)
186
+ 1520 train 4.297036 (lr=1.2555e-03) (hash(x)=74854227)
187
+ 1530 train 4.204149 (lr=1.2520e-03) (hash(x)=77468161)
188
+ 1540 train 4.225402 (lr=1.2485e-03) (hash(x)=87853059)
189
+ 1550 train 4.153730 (lr=1.2450e-03) (hash(x)=71225436)
190
+ 1560 train 4.257272 (lr=1.2414e-03) (hash(x)=84072783)
191
+ 1570 train 4.306070 (lr=1.2378e-03) (hash(x)=72723098)
192
+ 1580 train 4.323853 (lr=1.2342e-03) (hash(x)=90409866)
193
+ 1590 train 4.395305 (lr=1.2305e-03) (hash(x)=87481378)
194
+ 1600 val loss 4.3239
195
+ 1600 val perplexity 75.4836
196
+ 1600 train 4.385012 (lr=1.2268e-03) (hash(x)=77643862)
197
+ 1610 train 4.270534 (lr=1.2231e-03) (hash(x)=88862575)
198
+ 1620 train 4.280600 (lr=1.2193e-03) (hash(x)=84612581)
199
+ 1630 train 4.320957 (lr=1.2155e-03) (hash(x)=87075989)
200
+ 1640 train 4.204237 (lr=1.2116e-03) (hash(x)=88277361)
201
+ 1650 train 4.321087 (lr=1.2078e-03) (hash(x)=78750236)
202
+ 1660 train 4.241452 (lr=1.2039e-03) (hash(x)=82604581)
203
+ 1670 train 4.077402 (lr=1.1999e-03) (hash(x)=68482265)
204
+ 1680 train 4.203552 (lr=1.1960e-03) (hash(x)=75088835)
205
+ 1690 train 4.152542 (lr=1.1920e-03) (hash(x)=66060989)
206
+ 1700 val loss 4.3217
207
+ 1700 val perplexity 75.3173
208
+ 1700 train 4.314887 (lr=1.1879e-03) (hash(x)=79986754)
209
+ 1710 train 4.302011 (lr=1.1839e-03) (hash(x)=83657930)
210
+ 1720 train 4.249424 (lr=1.1798e-03) (hash(x)=81754135)
211
+ 1730 train 4.323024 (lr=1.1756e-03) (hash(x)=78904427)
212
+ 1740 train 4.229667 (lr=1.1715e-03) (hash(x)=85920177)
213
+ 1750 train 4.279835 (lr=1.1673e-03) (hash(x)=86573211)
214
+ 1760 train 4.236691 (lr=1.1631e-03) (hash(x)=81737128)
215
+ 1770 train 4.226747 (lr=1.1589e-03) (hash(x)=77400968)
216
+ 1780 train 4.240953 (lr=1.1546e-03) (hash(x)=73545497)
217
+ 1790 train 4.491923 (lr=1.1503e-03) (hash(x)=71641943)
218
+ 1800 val loss 4.2727
219
+ 1800 val perplexity 71.7125
220
+ 1800 train 4.338652 (lr=1.1460e-03) (hash(x)=87819781)
221
+ 1810 train 4.198794 (lr=1.1416e-03) (hash(x)=86870770)
222
+ 1820 train 4.205305 (lr=1.1372e-03) (hash(x)=82522211)
223
+ 1830 train 4.291976 (lr=1.1328e-03) (hash(x)=61947437)
224
+ 1840 train 4.169750 (lr=1.1284e-03) (hash(x)=79865406)
225
+ 1850 train 4.134159 (lr=1.1239e-03) (hash(x)=79828721)
226
+ 1860 train 4.278442 (lr=1.1194e-03) (hash(x)=80869571)
227
+ 1870 train 4.254370 (lr=1.1149e-03) (hash(x)=73780971)
228
+ 1880 train 4.224275 (lr=1.1104e-03) (hash(x)=79249549)
229
+ 1890 train 4.273096 (lr=1.1058e-03) (hash(x)=81041904)
230
+ 1900 val loss 4.2495
231
+ 1900 val perplexity 70.0704
232
+ 1900 train 4.342704 (lr=1.1012e-03) (hash(x)=82456430)
233
+ 1910 train 4.114508 (lr=1.0966e-03) (hash(x)=82222135)
234
+ 1920 train 4.154860 (lr=1.0920e-03) (hash(x)=70033249)
235
+ 1930 train 4.138369 (lr=1.0873e-03) (hash(x)=72887360)
236
+ 1940 train 4.185388 (lr=1.0827e-03) (hash(x)=83251100)
237
+ 1950 train 4.224561 (lr=1.0779e-03) (hash(x)=79660266)
238
+ 1960 train 4.104279 (lr=1.0732e-03) (hash(x)=78919068)
239
+ 1970 train 4.204835 (lr=1.0685e-03) (hash(x)=94690431)
240
+ 1980 train 4.041948 (lr=1.0637e-03) (hash(x)=88153756)
241
+ 1990 train 4.095307 (lr=1.0589e-03) (hash(x)=77195688)
242
+ 2000 val loss 4.2264
243
+ 2000 val perplexity 68.4718
244
+ 2000 train 4.188838 (lr=1.0541e-03) (hash(x)=81308591)
245
+ 2010 train 4.242614 (lr=1.0492e-03) (hash(x)=80205479)
246
+ 2020 train 4.156966 (lr=1.0444e-03) (hash(x)=81008704)
247
+ 2030 train 4.226622 (lr=1.0395e-03) (hash(x)=82914358)
248
+ 2040 train 4.236078 (lr=1.0346e-03) (hash(x)=88064399)
249
+ 2050 train 4.235309 (lr=1.0297e-03) (hash(x)=82889390)
250
+ 2060 train 4.306657 (lr=1.0248e-03) (hash(x)=80913554)
251
+ 2070 train 4.147379 (lr=1.0198e-03) (hash(x)=78121791)
252
+ 2080 train 4.283545 (lr=1.0148e-03) (hash(x)=75356657)
253
+ 2090 train 4.110053 (lr=1.0098e-03) (hash(x)=74048412)
254
+ 2100 val loss 4.2012
255
+ 2100 val perplexity 66.7695
256
+ 2100 train 4.031313 (lr=1.0048e-03) (hash(x)=68928225)
257
+ 2110 train 4.083453 (lr=9.9981e-04) (hash(x)=93734745)
258
+ 2120 train 4.077950 (lr=9.9476e-04) (hash(x)=72878164)
259
+ 2130 train 4.205284 (lr=9.8970e-04) (hash(x)=71508226)
260
+ 2140 train 4.162617 (lr=9.8463e-04) (hash(x)=84168671)
261
+ 2150 train 4.100291 (lr=9.7953e-04) (hash(x)=74673639)
262
+ 2160 train 4.249869 (lr=9.7443e-04) (hash(x)=75470031)
263
+ 2170 train 4.203418 (lr=9.6930e-04) (hash(x)=74307890)
264
+ 2180 train 4.318921 (lr=9.6417e-04) (hash(x)=77214245)
265
+ 2190 train 4.106559 (lr=9.5901e-04) (hash(x)=88628359)
266
+ 2200 val loss 4.1826
267
+ 2200 val perplexity 65.5387
268
+ 2200 train 4.121032 (lr=9.5385e-04) (hash(x)=74779126)
269
+ 2210 train 4.155656 (lr=9.4867e-04) (hash(x)=79817976)
270
+ 2220 train 4.110496 (lr=9.4348e-04) (hash(x)=75258996)
271
+ 2230 train 4.086431 (lr=9.3827e-04) (hash(x)=88164047)
272
+ 2240 train 4.063107 (lr=9.3306e-04) (hash(x)=78580686)
273
+ 2250 train 4.004002 (lr=9.2783e-04) (hash(x)=76510617)
274
+ 2260 train 4.082332 (lr=9.2259e-04) (hash(x)=76043966)
275
+ 2270 train 4.064415 (lr=9.1733e-04) (hash(x)=73062098)
276
+ 2280 train 4.283959 (lr=9.1207e-04) (hash(x)=81885909)
277
+ 2290 train 4.226594 (lr=9.0680e-04) (hash(x)=81481741)
278
+ 2300 val loss 4.1619
279
+ 2300 val perplexity 64.1950
280
+ 2300 train 4.087507 (lr=9.0151e-04) (hash(x)=82104275)
281
+ 2310 train 4.126513 (lr=8.9622e-04) (hash(x)=82137309)
282
+ 2320 train 4.107329 (lr=8.9092e-04) (hash(x)=86741167)
283
+ 2330 train 4.122222 (lr=8.8561e-04) (hash(x)=84796263)
284
+ 2340 train 4.099080 (lr=8.8029e-04) (hash(x)=169966529)
285
+ 2350 train 3.972561 (lr=8.7496e-04) (hash(x)=80006061)
286
+ 2360 train 4.026200 (lr=8.6962e-04) (hash(x)=71090999)
287
+ 2370 train 4.159523 (lr=8.6428e-04) (hash(x)=86641850)
288
+ 2380 train 4.057641 (lr=8.5893e-04) (hash(x)=84573253)
289
+ 2390 train 4.094704 (lr=8.5357e-04) (hash(x)=74483764)
290
+ 2400 val loss 4.1565
291
+ 2400 val perplexity 63.8453
292
+ 2400 train 4.066572 (lr=8.4821e-04) (hash(x)=78327659)
293
+ 2410 train 4.131469 (lr=8.4284e-04) (hash(x)=77273627)
294
+ 2420 train 3.995170 (lr=8.3747e-04) (hash(x)=76938049)
295
+ 2430 train 3.999827 (lr=8.3209e-04) (hash(x)=98449442)
296
+ 2440 train 3.854039 (lr=8.2670e-04) (hash(x)=78157797)
297
+ 2450 train 3.899999 (lr=8.2132e-04) (hash(x)=80637582)
298
+ 2460 train 3.887086 (lr=8.1592e-04) (hash(x)=80023854)
299
+ 2470 train 3.947611 (lr=8.1053e-04) (hash(x)=86678884)
300
+ 2480 train 4.172515 (lr=8.0513e-04) (hash(x)=84086469)
301
+ 2490 train 4.188626 (lr=7.9973e-04) (hash(x)=83131141)
302
+ 2500 val loss 4.1305
303
+ 2500 val perplexity 62.2095
304
+ 2500 train 4.110740 (lr=7.9433e-04) (hash(x)=82583497)
305
+ 2510 train 4.170415 (lr=7.8892e-04) (hash(x)=79164326)
306
+ 2520 train 4.076923 (lr=7.8352e-04) (hash(x)=76456503)
307
+ 2530 train 4.134107 (lr=7.7811e-04) (hash(x)=82001228)
308
+ 2540 train 4.170575 (lr=7.7270e-04) (hash(x)=75615595)
309
+ 2550 train 4.116284 (lr=7.6730e-04) (hash(x)=78397869)
310
+ 2560 train 4.120837 (lr=7.6189e-04) (hash(x)=75547032)
311
+ 2570 train 4.031884 (lr=7.5648e-04) (hash(x)=89201025)
312
+ 2580 train 4.017317 (lr=7.5108e-04) (hash(x)=89856704)
313
+ 2590 train 4.026735 (lr=7.4567e-04) (hash(x)=82175682)
314
+ 2600 val loss 4.1136
315
+ 2600 val perplexity 61.1674
316
+ 2600 train 4.035299 (lr=7.4027e-04) (hash(x)=78312826)
317
+ 2610 train 4.026513 (lr=7.3487e-04) (hash(x)=77066588)
318
+ 2620 train 3.872179 (lr=7.2947e-04) (hash(x)=78666061)
319
+ 2630 train 3.865324 (lr=7.2408e-04) (hash(x)=93762143)
320
+ 2640 train 3.882591 (lr=7.1868e-04) (hash(x)=83191587)
321
+ 2650 train 3.805134 (lr=7.1330e-04) (hash(x)=87169585)
322
+ 2660 train 4.019222 (lr=7.0791e-04) (hash(x)=86426388)
323
+ 2670 train 4.156046 (lr=7.0253e-04) (hash(x)=76692638)
324
+ 2680 train 4.072478 (lr=6.9716e-04) (hash(x)=77446063)
325
+ 2690 train 4.068733 (lr=6.9179e-04) (hash(x)=79809050)
326
+ 2700 val loss 4.1085
327
+ 2700 val perplexity 60.8536
328
+ 2700 train 4.003058 (lr=6.8643e-04) (hash(x)=83116823)
329
+ 2710 train 4.110912 (lr=6.8107e-04) (hash(x)=75622148)
330
+ 2720 train 4.043295 (lr=6.7572e-04) (hash(x)=80690512)
331
+ 2730 train 4.113472 (lr=6.7038e-04) (hash(x)=78009984)
332
+ 2740 train 4.063675 (lr=6.6504e-04) (hash(x)=74205488)
333
+ 2750 train 4.069544 (lr=6.5971e-04) (hash(x)=91013332)
334
+ 2760 train 3.991330 (lr=6.5439e-04) (hash(x)=76954961)
335
+ 2770 train 4.005804 (lr=6.4908e-04) (hash(x)=79390317)
336
+ 2780 train 3.870386 (lr=6.4378e-04) (hash(x)=70168783)
337
+ 2790 train 4.029764 (lr=6.3849e-04) (hash(x)=76028417)
338
+ 2800 val loss 4.0861
339
+ 2800 val perplexity 59.5073
340
+ 2800 train 3.857347 (lr=6.3320e-04) (hash(x)=77656050)
341
+ 2810 train 3.907868 (lr=6.2793e-04) (hash(x)=94248216)
342
+ 2820 train 3.968151 (lr=6.2267e-04) (hash(x)=78305078)
343
+ 2830 train 3.883700 (lr=6.1741e-04) (hash(x)=79948848)
344
+ 2840 train 3.751257 (lr=6.1217e-04) (hash(x)=85341024)
345
+ 2850 train 4.059037 (lr=6.0694e-04) (hash(x)=78735170)
346
+ 2860 train 4.307292 (lr=6.0173e-04) (hash(x)=71616419)
347
+ 2870 train 4.079407 (lr=5.9652e-04) (hash(x)=78656517)
348
+ 2880 train 4.089142 (lr=5.9133e-04) (hash(x)=80073987)
349
+ 2890 train 4.028285 (lr=5.8615e-04) (hash(x)=76894809)
350
+ 2900 val loss 4.0618
351
+ 2900 val perplexity 58.0794
352
+ 2900 train 3.998420 (lr=5.8099e-04) (hash(x)=80499838)
353
+ 2910 train 4.052412 (lr=5.7583e-04) (hash(x)=72673354)
354
+ 2920 train 4.029347 (lr=5.7070e-04) (hash(x)=84265768)
355
+ 2930 train 4.059742 (lr=5.6557e-04) (hash(x)=79612060)
356
+ 2940 train 3.913079 (lr=5.6047e-04) (hash(x)=74970087)
357
+ 2950 train 4.067758 (lr=5.5537e-04) (hash(x)=84166818)
358
+ 2960 train 4.059626 (lr=5.5030e-04) (hash(x)=89410221)
359
+ 2970 train 3.996549 (lr=5.4524e-04) (hash(x)=75672566)
360
+ 2980 train 3.851335 (lr=5.4019e-04) (hash(x)=81760314)
361
+ 2990 train 3.827325 (lr=5.3517e-04) (hash(x)=80605200)
362
+ 3000 val loss 4.0606
363
+ 3000 val perplexity 58.0099
364
+ 3000 train 3.822360 (lr=5.3016e-04) (hash(x)=83804735)
365
+ 3010 train 3.753647 (lr=5.2517e-04) (hash(x)=77015303)
366
+ 3020 train 3.822413 (lr=5.2019e-04) (hash(x)=81464523)
367
+ 3030 train 4.015129 (lr=5.1524e-04) (hash(x)=80168230)
368
+ 3040 train 3.958127 (lr=5.1030e-04) (hash(x)=84817006)
369
+ 3050 train 4.069607 (lr=5.0539e-04) (hash(x)=71601811)
370
+ 3060 train 3.985832 (lr=5.0049e-04) (hash(x)=85499733)
371
+ 3070 train 4.041065 (lr=4.9561e-04) (hash(x)=78670408)
372
+ 3080 train 3.936787 (lr=4.9075e-04) (hash(x)=77120468)
373
+ 3090 train 4.040203 (lr=4.8591e-04) (hash(x)=77927426)
374
+ 3100 val loss 4.0419
375
+ 3100 val perplexity 56.9366
376
+ 3100 train 4.057204 (lr=4.8110e-04) (hash(x)=83998606)
377
+ 3110 train 3.955440 (lr=4.7630e-04) (hash(x)=77990218)
378
+ 3120 train 3.872147 (lr=4.7153e-04) (hash(x)=81623970)
379
+ 3130 train 3.845905 (lr=4.6678e-04) (hash(x)=75014781)
380
+ 3140 train 3.899098 (lr=4.6205e-04) (hash(x)=72591250)
381
+ 3150 train 3.911500 (lr=4.5734e-04) (hash(x)=81421847)
382
+ 3160 train 3.961646 (lr=4.5266e-04) (hash(x)=75112631)
383
+ 3170 train 4.056727 (lr=4.4800e-04) (hash(x)=87518033)
384
+ 3180 train 3.902801 (lr=4.4337e-04) (hash(x)=78648348)
385
+ 3190 train 4.026871 (lr=4.3875e-04) (hash(x)=78270029)
386
+ 3200 val loss 4.0230
387
+ 3200 val perplexity 55.8685
388
+ 3200 train 4.099795 (lr=4.3417e-04) (hash(x)=83380714)
389
+ 3210 train 4.280150 (lr=4.2960e-04) (hash(x)=67804991)
390
+ 3220 train 4.010918 (lr=4.2507e-04) (hash(x)=76968804)
391
+ 3230 train 3.943396 (lr=4.2055e-04) (hash(x)=75798670)
392
+ 3240 train 4.016018 (lr=4.1607e-04) (hash(x)=83192811)
393
+ 3250 train 3.953175 (lr=4.1161e-04) (hash(x)=81659789)
394
+ 3260 train 3.999028 (lr=4.0717e-04) (hash(x)=80622502)
395
+ 3270 train 3.965233 (lr=4.0276e-04) (hash(x)=85436511)
396
+ 3280 train 3.914549 (lr=3.9838e-04) (hash(x)=80589180)
397
+ 3290 train 4.049939 (lr=3.9403e-04) (hash(x)=85747193)
398
+ 3300 val loss 4.0186
399
+ 3300 val perplexity 55.6217
400
+ 3300 train 3.919935 (lr=3.8971e-04) (hash(x)=74635692)
401
+ 3310 train 3.903064 (lr=3.8541e-04) (hash(x)=82045455)
402
+ 3320 train 3.775055 (lr=3.8114e-04) (hash(x)=73249173)
403
+ 3330 train 3.734545 (lr=3.7690e-04) (hash(x)=71492338)
404
+ 3340 train 4.067096 (lr=3.7269e-04) (hash(x)=76374071)
405
+ 3350 train 4.036598 (lr=3.6851e-04) (hash(x)=80474064)
406
+ 3360 train 3.950661 (lr=3.6435e-04) (hash(x)=84390892)
407
+ 3370 train 3.995116 (lr=3.6023e-04) (hash(x)=83399949)
408
+ 3380 train 4.046431 (lr=3.5614e-04) (hash(x)=77648059)
409
+ 3390 train 4.021377 (lr=3.5208e-04) (hash(x)=80152701)
410
+ 3400 val loss 4.0004
411
+ 3400 val perplexity 54.6204
412
+ 3400 train 4.065968 (lr=3.4805e-04) (hash(x)=81952545)
413
+ 3410 train 4.024397 (lr=3.4405e-04) (hash(x)=80908993)
414
+ 3420 train 4.064485 (lr=3.4008e-04) (hash(x)=79610037)
415
+ 3430 train 3.930072 (lr=3.3614e-04) (hash(x)=87624382)
416
+ 3440 train 3.990571 (lr=3.3224e-04) (hash(x)=82336381)
417
+ 3450 train 3.968970 (lr=3.2836e-04) (hash(x)=87687835)
418
+ 3460 train 3.971879 (lr=3.2452e-04) (hash(x)=76376135)
419
+ 3470 train 3.881922 (lr=3.2072e-04) (hash(x)=79331391)
420
+ 3480 train 3.793657 (lr=3.1694e-04) (hash(x)=107002681)
421
+ 3490 train 3.870641 (lr=3.1320e-04) (hash(x)=77465514)
422
+ 3500 val loss 3.9964
423
+ 3500 val perplexity 54.4010
424
+ 3500 train 3.834505 (lr=3.0949e-04) (hash(x)=88237229)
425
+ 3510 train 3.902920 (lr=3.0582e-04) (hash(x)=81011739)
426
+ 3520 train 3.948442 (lr=3.0218e-04) (hash(x)=64643427)
427
+ 3530 train 4.221539 (lr=2.9858e-04) (hash(x)=78029539)
428
+ 3540 train 4.060056 (lr=2.9501e-04) (hash(x)=83188968)
429
+ 3550 train 4.077145 (lr=2.9147e-04) (hash(x)=86104185)
430
+ 3560 train 4.004606 (lr=2.8797e-04) (hash(x)=81993629)
431
+ 3570 train 3.976983 (lr=2.8451e-04) (hash(x)=72719368)
432
+ 3580 train 3.985878 (lr=2.8108e-04) (hash(x)=90091487)
433
+ 3590 train 3.873783 (lr=2.7769e-04) (hash(x)=77393152)
434
+ 3600 val loss 3.9783
435
+ 3600 val perplexity 53.4270
436
+ 3600 train 3.983576 (lr=2.7433e-04) (hash(x)=73103504)
437
+ 3610 train 3.985864 (lr=2.7101e-04) (hash(x)=84429400)
438
+ 3620 train 3.963993 (lr=2.6773e-04) (hash(x)=77021795)
439
+ 3630 train 3.915364 (lr=2.6448e-04) (hash(x)=92830605)
440
+ 3640 train 4.111061 (lr=2.6127e-04) (hash(x)=78313175)
441
+ 3650 train 3.909412 (lr=2.5810e-04) (hash(x)=85395549)
442
+ 3660 train 3.703520 (lr=2.5497e-04) (hash(x)=78114459)
443
+ 3670 train 3.830884 (lr=2.5187e-04) (hash(x)=74968316)
444
+ 3680 train 3.759836 (lr=2.4882e-04) (hash(x)=73358737)
445
+ 3690 train 3.865160 (lr=2.4580e-04) (hash(x)=76399442)
446
+ 3700 val loss 3.9813
447
+ 3700 val perplexity 53.5865
448
+ 3700 train 3.867809 (lr=2.4282e-04) (hash(x)=74039273)
449
+ 3710 train 4.040471 (lr=2.3988e-04) (hash(x)=79299680)
450
+ 3720 train 3.957799 (lr=2.3698e-04) (hash(x)=94701498)
451
+ 3730 train 4.036037 (lr=2.3411e-04) (hash(x)=75352071)
452
+ 3740 train 3.911677 (lr=2.3129e-04) (hash(x)=77909487)
453
+ 3750 train 4.015388 (lr=2.2850e-04) (hash(x)=73986730)
454
+ 3760 train 3.974713 (lr=2.2576e-04) (hash(x)=79325763)
455
+ 3770 train 4.025911 (lr=2.2306e-04) (hash(x)=72457818)
456
+ 3780 train 4.156987 (lr=2.2039e-04) (hash(x)=71775590)
457
+ 3790 train 4.044651 (lr=2.1777e-04) (hash(x)=82638943)
458
+ 3800 val loss 3.9651
459
+ 3800 val perplexity 52.7269
460
+ 3800 train 4.019753 (lr=2.1519e-04) (hash(x)=79965893)
461
+ 3810 train 3.952237 (lr=2.1265e-04) (hash(x)=72598235)
462
+ 3820 train 3.936225 (lr=2.1015e-04) (hash(x)=83113889)
463
+ 3830 train 3.995459 (lr=2.0769e-04) (hash(x)=74434590)
464
+ 3840 train 3.915186 (lr=2.0527e-04) (hash(x)=82860348)
465
+ 3850 train 3.851002 (lr=2.0289e-04) (hash(x)=78067565)
466
+ 3860 train 3.917578 (lr=2.0056e-04) (hash(x)=82592498)
467
+ 3870 train 3.852836 (lr=1.9827e-04) (hash(x)=81820733)
468
+ 3880 train 3.866150 (lr=1.9602e-04) (hash(x)=87709040)
469
+ 3890 train 3.911203 (lr=1.9381e-04) (hash(x)=70379093)
470
+ 3900 val loss 3.9588
471
+ 3900 val perplexity 52.3954
472
+ 3900 train 4.032923 (lr=1.9164e-04) (hash(x)=76597431)
473
+ 3910 train 3.916196 (lr=1.8952e-04) (hash(x)=90490716)
474
+ 3920 train 4.023167 (lr=1.8744e-04) (hash(x)=81970659)
475
+ 3930 train 4.008700 (lr=1.8540e-04) (hash(x)=81496334)
476
+ 3940 train 3.983585 (lr=1.8341e-04) (hash(x)=75717605)
477
+ 3950 train 3.929852 (lr=1.8146e-04) (hash(x)=82749357)
478
+ 3960 train 3.960742 (lr=1.7955e-04) (hash(x)=89355157)
479
+ 3970 train 3.917400 (lr=1.7769e-04) (hash(x)=78980403)
480
+ 3980 train 3.888764 (lr=1.7587e-04) (hash(x)=76627217)
481
+ 3990 train 3.889933 (lr=1.7409e-04) (hash(x)=72412879)
482
+ 4000 val loss 3.9545
483
+ 4000 val perplexity 52.1691
484
+ 4000 train 3.906086 (lr=1.7236e-04) (hash(x)=83018142)
485
+ 4010 train 3.918824 (lr=1.7067e-04) (hash(x)=81272436)
486
+ 4020 train 3.851087 (lr=1.6902e-04) (hash(x)=85497482)
487
+ 4030 train 3.922402 (lr=1.6742e-04) (hash(x)=85241734)
488
+ 4040 train 3.886203 (lr=1.6587e-04) (hash(x)=77925307)
489
+ 4050 train 3.884327 (lr=1.6436e-04) (hash(x)=84826179)
490
+ 4060 train 3.944555 (lr=1.6289e-04) (hash(x)=83606764)
491
+ 4070 train 3.854738 (lr=1.6147e-04) (hash(x)=80567590)
492
+ 4080 train 3.908295 (lr=1.6009e-04) (hash(x)=76860998)
493
+ 4090 train 3.944118 (lr=1.5876e-04) (hash(x)=74902328)
494
+ 4100 val loss 3.9438
495
+ 4100 val perplexity 51.6167
496
+ 4100 train 3.987155 (lr=1.5747e-04) (hash(x)=82832041)
497
+ 4110 train 3.870873 (lr=1.5623e-04) (hash(x)=79143262)
498
+ 4120 train 3.895625 (lr=1.5503e-04) (hash(x)=77038149)
499
+ 4130 train 4.079492 (lr=1.5388e-04) (hash(x)=86339074)
500
+ 4140 train 4.140169 (lr=1.5277e-04) (hash(x)=76686216)
501
+ 4150 train 3.919492 (lr=1.5171e-04) (hash(x)=70522682)
502
+ 4160 train 3.885240 (lr=1.5070e-04) (hash(x)=90958555)
503
+ 4170 train 3.911674 (lr=1.4973e-04) (hash(x)=91463532)
504
+ 4180 train 3.896229 (lr=1.4880e-04) (hash(x)=81959329)
505
+ 4190 train 3.850025 (lr=1.4793e-04) (hash(x)=83146752)
506
+ 4200 val loss 3.9471
507
+ 4200 val perplexity 51.7833
508
+ 4200 train 3.840179 (lr=1.4709e-04) (hash(x)=78361715)
509
+ 4210 train 3.983580 (lr=1.4631e-04) (hash(x)=87364889)
510
+ 4220 train 3.839051 (lr=1.4557e-04) (hash(x)=70465156)
511
+ 4230 train 3.990428 (lr=1.4487e-04) (hash(x)=84524081)
512
+ 4240 train 3.947452 (lr=1.4423e-04) (hash(x)=77824868)
513
+ 4250 train 3.959488 (lr=1.4362e-04) (hash(x)=81710711)
514
+ 4260 train 3.940060 (lr=1.4307e-04) (hash(x)=76362728)
515
+ 4270 train 3.934490 (lr=1.4256e-04) (hash(x)=83115208)
516
+ 4280 train 3.843919 (lr=1.4209e-04) (hash(x)=87218314)
517
+ 4290 train 4.006782 (lr=1.4168e-04) (hash(x)=74582673)
518
+ 4300 val loss 3.9404
519
+ 4300 val perplexity 51.4415
520
+ 4300 train 3.882605 (lr=1.4131e-04) (hash(x)=77379615)
521
+ 4310 train 3.856438 (lr=1.4098e-04) (hash(x)=78669579)
522
+ 4320 train 3.975764 (lr=1.4070e-04) (hash(x)=83066608)
523
+ 4330 train 3.936054 (lr=1.4047e-04) (hash(x)=83037340)
524
+ 4340 train 3.978107 (lr=1.4028e-04) (hash(x)=82849771)
525
+ 4350 train 3.914237 (lr=1.4015e-04) (hash(x)=76693985)
526
+ 4360 train 3.998804 (lr=1.4005e-04) (hash(x)=77745394)
527
+ 4370 train 3.959105 (lr=1.4001e-04) (hash(x)=79954388)
528
+ 4374 val loss 3.9348
529
+ 4374 val perplexity 51.1504
12_head_baseline_lr_14e-4_head_dim_22_fixed/model_04374.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:26f0ac2e678a5f662984c8beb64fe1ca2772298cd4a89747761891445d613a2e
3
+ size 96858242
12_head_baseline_lr_14e-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:a1fd2a9a7ba59139f2e3a82f8253e1c456cec0da58b3607d658815dfdcd94c7c
3
+ size 187435910