andrew-healey commited on
Commit
c844a52
·
verified ·
1 Parent(s): b132d66

Upload folder using huggingface_hub

Browse files
12_head_baseline_lr_25e-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_25e-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": 0.0025, "decay_lr": true, "readout_zero_init": false, "query_zero_init": false, "l1_loss": false, "debugpy": false, "key": "12_head_baseline_lr_25e-4_fixed"}
12_head_baseline_lr_25e-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_25e-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 11.388922 (lr=3.4965e-06) (hash(x)=93356070)
5
+ 10 train 10.091938 (lr=3.8462e-05) (hash(x)=91253010)
6
+ 20 train 9.566366 (lr=7.3427e-05) (hash(x)=74934453)
7
+ 30 train 8.940718 (lr=1.0839e-04) (hash(x)=79278034)
8
+ 40 train 8.181480 (lr=1.4336e-04) (hash(x)=80107892)
9
+ 50 train 7.767656 (lr=1.7832e-04) (hash(x)=70745428)
10
+ 60 train 7.758302 (lr=2.1329e-04) (hash(x)=80077589)
11
+ 70 train 7.446225 (lr=2.4825e-04) (hash(x)=76213766)
12
+ 80 train 7.441905 (lr=2.8322e-04) (hash(x)=83218328)
13
+ 90 train 7.244534 (lr=3.1818e-04) (hash(x)=74855845)
14
+ 100 val loss 7.1491
15
+ 100 val perplexity 1273.0051
16
+ 100 train 6.988572 (lr=3.5315e-04) (hash(x)=82814902)
17
+ 110 train 6.952920 (lr=3.8811e-04) (hash(x)=84907741)
18
+ 120 train 6.783826 (lr=4.2308e-04) (hash(x)=82613223)
19
+ 130 train 6.623692 (lr=4.5804e-04) (hash(x)=83540876)
20
+ 140 train 6.585614 (lr=4.9301e-04) (hash(x)=75095216)
21
+ 150 train 6.660504 (lr=5.2797e-04) (hash(x)=97190944)
22
+ 160 train 6.570115 (lr=5.6294e-04) (hash(x)=82117809)
23
+ 170 train 6.473680 (lr=5.9790e-04) (hash(x)=70514724)
24
+ 180 train 6.396311 (lr=6.3287e-04) (hash(x)=81029624)
25
+ 190 train 6.321274 (lr=6.6783e-04) (hash(x)=78996841)
26
+ 200 val loss 6.2460
27
+ 200 val perplexity 515.9409
28
+ 200 train 6.251741 (lr=7.0280e-04) (hash(x)=79845097)
29
+ 210 train 6.133491 (lr=7.3776e-04) (hash(x)=78997683)
30
+ 220 train 6.057427 (lr=7.7273e-04) (hash(x)=74895865)
31
+ 230 train 6.178236 (lr=8.0769e-04) (hash(x)=80933276)
32
+ 240 train 5.892378 (lr=8.4266e-04) (hash(x)=77664606)
33
+ 250 train 5.934065 (lr=8.7762e-04) (hash(x)=75903930)
34
+ 260 train 5.826718 (lr=9.1259e-04) (hash(x)=82996853)
35
+ 270 train 5.767688 (lr=9.4755e-04) (hash(x)=73269568)
36
+ 280 train 5.692712 (lr=9.8252e-04) (hash(x)=82906427)
37
+ 290 train 5.612326 (lr=1.0175e-03) (hash(x)=71797895)
38
+ 300 val loss 5.7717
39
+ 300 val perplexity 321.0679
40
+ 300 train 5.480511 (lr=1.0524e-03) (hash(x)=85232249)
41
+ 310 train 5.410742 (lr=1.0874e-03) (hash(x)=75252489)
42
+ 320 train 5.510778 (lr=1.1224e-03) (hash(x)=74394644)
43
+ 330 train 5.604581 (lr=1.1573e-03) (hash(x)=82772910)
44
+ 340 train 5.654929 (lr=1.1923e-03) (hash(x)=81627464)
45
+ 350 train 5.630037 (lr=1.2273e-03) (hash(x)=84678053)
46
+ 360 train 5.570426 (lr=1.2622e-03) (hash(x)=81884128)
47
+ 370 train 5.575796 (lr=1.2972e-03) (hash(x)=76379242)
48
+ 380 train 5.393479 (lr=1.3322e-03) (hash(x)=98182875)
49
+ 390 train 5.378513 (lr=1.3671e-03) (hash(x)=79710436)
50
+ 400 val loss 5.4224
51
+ 400 val perplexity 226.4287
52
+ 400 train 5.367092 (lr=1.4021e-03) (hash(x)=79841071)
53
+ 410 train 5.318703 (lr=1.4371e-03) (hash(x)=75844151)
54
+ 420 train 5.333406 (lr=1.4720e-03) (hash(x)=73125036)
55
+ 430 train 5.358259 (lr=1.5070e-03) (hash(x)=84214858)
56
+ 440 train 5.251061 (lr=1.5420e-03) (hash(x)=80456994)
57
+ 450 train 5.237456 (lr=1.5769e-03) (hash(x)=76735962)
58
+ 460 train 5.153664 (lr=1.6119e-03) (hash(x)=81845446)
59
+ 470 train 4.966374 (lr=1.6469e-03) (hash(x)=76094689)
60
+ 480 train 5.097123 (lr=1.6818e-03) (hash(x)=83806686)
61
+ 490 train 4.907626 (lr=1.7168e-03) (hash(x)=84690227)
62
+ 500 val loss 5.2571
63
+ 500 val perplexity 191.9149
64
+ 500 train 4.990450 (lr=1.7517e-03) (hash(x)=71851938)
65
+ 510 train 5.107912 (lr=1.7867e-03) (hash(x)=77159346)
66
+ 520 train 5.173186 (lr=1.8217e-03) (hash(x)=80755753)
67
+ 530 train 5.270996 (lr=1.8566e-03) (hash(x)=75379679)
68
+ 540 train 5.178167 (lr=1.8916e-03) (hash(x)=82458619)
69
+ 550 train 5.149359 (lr=1.9266e-03) (hash(x)=86366396)
70
+ 560 train 5.190310 (lr=1.9615e-03) (hash(x)=85095044)
71
+ 570 train 5.150440 (lr=1.9965e-03) (hash(x)=78385159)
72
+ 580 train 5.077712 (lr=2.0315e-03) (hash(x)=79342394)
73
+ 590 train 5.084577 (lr=2.0664e-03) (hash(x)=70782192)
74
+ 600 val loss 5.1211
75
+ 600 val perplexity 167.5230
76
+ 600 train 5.095017 (lr=2.1014e-03) (hash(x)=76130353)
77
+ 610 train 5.005000 (lr=2.1364e-03) (hash(x)=74778440)
78
+ 620 train 4.968255 (lr=2.1713e-03) (hash(x)=79129709)
79
+ 630 train 4.865604 (lr=2.2063e-03) (hash(x)=76469962)
80
+ 640 train 4.856482 (lr=2.2413e-03) (hash(x)=78288049)
81
+ 650 train 4.821362 (lr=2.2762e-03) (hash(x)=76641388)
82
+ 660 train 4.749958 (lr=2.3112e-03) (hash(x)=75937906)
83
+ 670 train 4.730556 (lr=2.3462e-03) (hash(x)=74807157)
84
+ 680 train 4.625620 (lr=2.3811e-03) (hash(x)=77490144)
85
+ 690 train 4.669048 (lr=2.4161e-03) (hash(x)=72900124)
86
+ 700 val loss 4.9397
87
+ 700 val perplexity 139.7256
88
+ 700 train 4.643094 (lr=2.4510e-03) (hash(x)=73424218)
89
+ 710 train 4.916427 (lr=2.4860e-03) (hash(x)=86052345)
90
+ 720 train 4.901440 (lr=2.5000e-03) (hash(x)=77462613)
91
+ 730 train 4.892340 (lr=2.4999e-03) (hash(x)=79027471)
92
+ 740 train 4.843719 (lr=2.4997e-03) (hash(x)=78149992)
93
+ 750 train 4.795235 (lr=2.4995e-03) (hash(x)=89147499)
94
+ 760 train 4.827475 (lr=2.4992e-03) (hash(x)=74931538)
95
+ 770 train 4.770164 (lr=2.4987e-03) (hash(x)=89648204)
96
+ 780 train 4.820521 (lr=2.4982e-03) (hash(x)=83708293)
97
+ 790 train 4.808646 (lr=2.4977e-03) (hash(x)=84336512)
98
+ 800 val loss 4.7862
99
+ 800 val perplexity 119.8492
100
+ 800 train 4.704918 (lr=2.4970e-03) (hash(x)=75025285)
101
+ 810 train 4.644829 (lr=2.4963e-03) (hash(x)=79436536)
102
+ 820 train 4.625872 (lr=2.4954e-03) (hash(x)=90467390)
103
+ 830 train 4.660602 (lr=2.4945e-03) (hash(x)=77292020)
104
+ 840 train 4.592151 (lr=2.4935e-03) (hash(x)=75568927)
105
+ 850 train 4.614832 (lr=2.4925e-03) (hash(x)=79671781)
106
+ 860 train 4.700879 (lr=2.4913e-03) (hash(x)=84280943)
107
+ 870 train 4.687429 (lr=2.4901e-03) (hash(x)=80901294)
108
+ 880 train 4.797977 (lr=2.4887e-03) (hash(x)=81437584)
109
+ 890 train 4.652145 (lr=2.4873e-03) (hash(x)=73893778)
110
+ 900 val loss 4.6817
111
+ 900 val perplexity 107.9544
112
+ 900 train 4.616809 (lr=2.4858e-03) (hash(x)=74987794)
113
+ 910 train 4.627005 (lr=2.4843e-03) (hash(x)=93721374)
114
+ 920 train 4.612716 (lr=2.4826e-03) (hash(x)=79149678)
115
+ 930 train 4.568168 (lr=2.4809e-03) (hash(x)=83179387)
116
+ 940 train 4.568726 (lr=2.4791e-03) (hash(x)=89009978)
117
+ 950 train 4.560432 (lr=2.4772e-03) (hash(x)=76483735)
118
+ 960 train 4.480444 (lr=2.4752e-03) (hash(x)=78638850)
119
+ 970 train 4.517279 (lr=2.4732e-03) (hash(x)=90415206)
120
+ 980 train 4.464985 (lr=2.4710e-03) (hash(x)=90819195)
121
+ 990 train 4.593828 (lr=2.4688e-03) (hash(x)=85938852)
122
+ 1000 val loss 4.5743
123
+ 1000 val perplexity 96.9607
124
+ 1000 train 4.614430 (lr=2.4665e-03) (hash(x)=82436789)
125
+ 1010 train 4.685489 (lr=2.4641e-03) (hash(x)=63022149)
126
+ 1020 train 4.596513 (lr=2.4617e-03) (hash(x)=83501199)
127
+ 1030 train 4.570286 (lr=2.4591e-03) (hash(x)=77093326)
128
+ 1040 train 4.455762 (lr=2.4565e-03) (hash(x)=73437559)
129
+ 1050 train 4.538451 (lr=2.4538e-03) (hash(x)=84550388)
130
+ 1060 train 4.541142 (lr=2.4510e-03) (hash(x)=86184566)
131
+ 1070 train 4.533621 (lr=2.4482e-03) (hash(x)=85644922)
132
+ 1080 train 4.468680 (lr=2.4452e-03) (hash(x)=86179801)
133
+ 1090 train 4.441399 (lr=2.4422e-03) (hash(x)=82868303)
134
+ 1100 val loss 4.5148
135
+ 1100 val perplexity 91.3556
136
+ 1100 train 4.531556 (lr=2.4391e-03) (hash(x)=96780388)
137
+ 1110 train 4.480387 (lr=2.4360e-03) (hash(x)=87223122)
138
+ 1120 train 4.452402 (lr=2.4327e-03) (hash(x)=77292786)
139
+ 1130 train 4.472954 (lr=2.4294e-03) (hash(x)=88761375)
140
+ 1140 train 4.381264 (lr=2.4260e-03) (hash(x)=80956468)
141
+ 1150 train 4.554913 (lr=2.4225e-03) (hash(x)=84725462)
142
+ 1160 train 4.539227 (lr=2.4189e-03) (hash(x)=76098113)
143
+ 1170 train 4.514234 (lr=2.4153e-03) (hash(x)=84387685)
144
+ 1180 train 4.461565 (lr=2.4116e-03) (hash(x)=76120321)
145
+ 1190 train 4.421763 (lr=2.4078e-03) (hash(x)=85974065)
146
+ 1200 val loss 4.4434
147
+ 1200 val perplexity 85.0654
148
+ 1200 train 4.331007 (lr=2.4039e-03) (hash(x)=79340644)
149
+ 1210 train 4.422689 (lr=2.4000e-03) (hash(x)=88345025)
150
+ 1220 train 4.325808 (lr=2.3960e-03) (hash(x)=86406230)
151
+ 1230 train 4.319077 (lr=2.3919e-03) (hash(x)=80608975)
152
+ 1240 train 4.381937 (lr=2.3877e-03) (hash(x)=81635225)
153
+ 1250 train 4.213146 (lr=2.3834e-03) (hash(x)=82126376)
154
+ 1260 train 4.389433 (lr=2.3791e-03) (hash(x)=83737972)
155
+ 1270 train 4.300915 (lr=2.3747e-03) (hash(x)=81132345)
156
+ 1280 train 4.392656 (lr=2.3703e-03) (hash(x)=85850404)
157
+ 1290 train 4.394496 (lr=2.3657e-03) (hash(x)=83490640)
158
+ 1300 val loss 4.4021
159
+ 1300 val perplexity 81.6201
160
+ 1300 train 4.371149 (lr=2.3611e-03) (hash(x)=77545187)
161
+ 1310 train 4.423258 (lr=2.3564e-03) (hash(x)=86412685)
162
+ 1320 train 4.319774 (lr=2.3517e-03) (hash(x)=86429640)
163
+ 1330 train 4.451466 (lr=2.3469e-03) (hash(x)=79645281)
164
+ 1340 train 4.409052 (lr=2.3420e-03) (hash(x)=82668541)
165
+ 1350 train 4.278049 (lr=2.3370e-03) (hash(x)=73627248)
166
+ 1360 train 4.375099 (lr=2.3319e-03) (hash(x)=74845133)
167
+ 1370 train 4.284807 (lr=2.3268e-03) (hash(x)=77850497)
168
+ 1380 train 4.307235 (lr=2.3216e-03) (hash(x)=76143954)
169
+ 1390 train 4.311766 (lr=2.3164e-03) (hash(x)=80675544)
170
+ 1400 val loss 4.3814
171
+ 1400 val perplexity 79.9532
172
+ 1400 train 4.293755 (lr=2.3111e-03) (hash(x)=76010938)
173
+ 1410 train 4.339595 (lr=2.3057e-03) (hash(x)=81260300)
174
+ 1420 train 4.458948 (lr=2.3002e-03) (hash(x)=83803629)
175
+ 1430 train 4.279525 (lr=2.2947e-03) (hash(x)=83840110)
176
+ 1440 train 4.478382 (lr=2.2891e-03) (hash(x)=97711831)
177
+ 1450 train 4.349561 (lr=2.2834e-03) (hash(x)=81110139)
178
+ 1460 train 4.274888 (lr=2.2777e-03) (hash(x)=90942839)
179
+ 1470 train 4.309854 (lr=2.2719e-03) (hash(x)=80721564)
180
+ 1480 train 4.408832 (lr=2.2661e-03) (hash(x)=77852059)
181
+ 1490 train 4.309026 (lr=2.2601e-03) (hash(x)=80011365)
182
+ 1500 val loss 4.3429
183
+ 1500 val perplexity 76.9276
184
+ 1500 train 4.334823 (lr=2.2541e-03) (hash(x)=84496142)
185
+ 1510 train 4.267076 (lr=2.2481e-03) (hash(x)=70186729)
186
+ 1520 train 4.294215 (lr=2.2420e-03) (hash(x)=74854227)
187
+ 1530 train 4.200268 (lr=2.2358e-03) (hash(x)=77468161)
188
+ 1540 train 4.209461 (lr=2.2295e-03) (hash(x)=87853059)
189
+ 1550 train 4.151201 (lr=2.2232e-03) (hash(x)=71225436)
190
+ 1560 train 4.249141 (lr=2.2168e-03) (hash(x)=84072783)
191
+ 1570 train 4.304510 (lr=2.2104e-03) (hash(x)=72723098)
192
+ 1580 train 4.313807 (lr=2.2039e-03) (hash(x)=90409866)
193
+ 1590 train 4.389756 (lr=2.1973e-03) (hash(x)=87481378)
194
+ 1600 val loss 4.3147
195
+ 1600 val perplexity 74.7883
196
+ 1600 train 4.374784 (lr=2.1907e-03) (hash(x)=77643862)
197
+ 1610 train 4.259751 (lr=2.1840e-03) (hash(x)=88862575)
198
+ 1620 train 4.275518 (lr=2.1773e-03) (hash(x)=84612581)
199
+ 1630 train 4.310367 (lr=2.1705e-03) (hash(x)=87075989)
200
+ 1640 train 4.195184 (lr=2.1636e-03) (hash(x)=88277361)
201
+ 1650 train 4.313665 (lr=2.1567e-03) (hash(x)=78750236)
202
+ 1660 train 4.231048 (lr=2.1497e-03) (hash(x)=82604581)
203
+ 1670 train 4.079371 (lr=2.1427e-03) (hash(x)=68482265)
204
+ 1680 train 4.197592 (lr=2.1356e-03) (hash(x)=75088835)
205
+ 1690 train 4.143118 (lr=2.1285e-03) (hash(x)=66060989)
206
+ 1700 val loss 4.3196
207
+ 1700 val perplexity 75.1622
208
+ 1700 train 4.310461 (lr=2.1213e-03) (hash(x)=79986754)
209
+ 1710 train 4.294552 (lr=2.1140e-03) (hash(x)=83657930)
210
+ 1720 train 4.239735 (lr=2.1067e-03) (hash(x)=81754135)
211
+ 1730 train 4.323257 (lr=2.0994e-03) (hash(x)=78904427)
212
+ 1740 train 4.227790 (lr=2.0920e-03) (hash(x)=85920177)
213
+ 1750 train 4.279307 (lr=2.0845e-03) (hash(x)=86573211)
214
+ 1760 train 4.231834 (lr=2.0770e-03) (hash(x)=81737128)
215
+ 1770 train 4.222325 (lr=2.0694e-03) (hash(x)=77400968)
216
+ 1780 train 4.236928 (lr=2.0618e-03) (hash(x)=73545497)
217
+ 1790 train 4.482675 (lr=2.0541e-03) (hash(x)=71641943)
218
+ 1800 val loss 4.2716
219
+ 1800 val perplexity 71.6386
220
+ 1800 train 4.334870 (lr=2.0464e-03) (hash(x)=87819781)
221
+ 1810 train 4.191759 (lr=2.0386e-03) (hash(x)=86870770)
222
+ 1820 train 4.201089 (lr=2.0308e-03) (hash(x)=82522211)
223
+ 1830 train 4.288975 (lr=2.0229e-03) (hash(x)=61947437)
224
+ 1840 train 4.161946 (lr=2.0150e-03) (hash(x)=79865406)
225
+ 1850 train 4.133183 (lr=2.0070e-03) (hash(x)=79828721)
226
+ 1860 train 4.272957 (lr=1.9990e-03) (hash(x)=80869571)
227
+ 1870 train 4.259056 (lr=1.9910e-03) (hash(x)=73780971)
228
+ 1880 train 4.222263 (lr=1.9829e-03) (hash(x)=79249549)
229
+ 1890 train 4.280668 (lr=1.9747e-03) (hash(x)=81041904)
230
+ 1900 val loss 4.2519
231
+ 1900 val perplexity 70.2354
232
+ 1900 train 4.353182 (lr=1.9665e-03) (hash(x)=82456430)
233
+ 1910 train 4.116021 (lr=1.9583e-03) (hash(x)=82222135)
234
+ 1920 train 4.149905 (lr=1.9500e-03) (hash(x)=70033249)
235
+ 1930 train 4.136928 (lr=1.9417e-03) (hash(x)=72887360)
236
+ 1940 train 4.184105 (lr=1.9333e-03) (hash(x)=83251100)
237
+ 1950 train 4.226418 (lr=1.9249e-03) (hash(x)=79660266)
238
+ 1960 train 4.103600 (lr=1.9165e-03) (hash(x)=78919068)
239
+ 1970 train 4.201706 (lr=1.9080e-03) (hash(x)=94690431)
240
+ 1980 train 4.044730 (lr=1.8995e-03) (hash(x)=88153756)
241
+ 1990 train 4.092199 (lr=1.8909e-03) (hash(x)=77195688)
242
+ 2000 val loss 4.2283
243
+ 2000 val perplexity 68.5972
244
+ 2000 train 4.190606 (lr=1.8823e-03) (hash(x)=81308591)
245
+ 2010 train 4.242620 (lr=1.8737e-03) (hash(x)=80205479)
246
+ 2020 train 4.154243 (lr=1.8650e-03) (hash(x)=81008704)
247
+ 2030 train 4.221318 (lr=1.8563e-03) (hash(x)=82914358)
248
+ 2040 train 4.232865 (lr=1.8475e-03) (hash(x)=88064399)
249
+ 2050 train 4.235703 (lr=1.8387e-03) (hash(x)=82889390)
250
+ 2060 train 4.312901 (lr=1.8299e-03) (hash(x)=80913554)
251
+ 2070 train 4.144819 (lr=1.8211e-03) (hash(x)=78121791)
252
+ 2080 train 4.282660 (lr=1.8122e-03) (hash(x)=75356657)
253
+ 2090 train 4.108078 (lr=1.8033e-03) (hash(x)=74048412)
254
+ 2100 val loss 4.2060
255
+ 2100 val perplexity 67.0881
256
+ 2100 train 4.028151 (lr=1.7943e-03) (hash(x)=68928225)
257
+ 2110 train 4.081586 (lr=1.7854e-03) (hash(x)=93734745)
258
+ 2120 train 4.078258 (lr=1.7764e-03) (hash(x)=72878164)
259
+ 2130 train 4.211380 (lr=1.7673e-03) (hash(x)=71508226)
260
+ 2140 train 4.171084 (lr=1.7583e-03) (hash(x)=84168671)
261
+ 2150 train 4.099793 (lr=1.7492e-03) (hash(x)=74673639)
262
+ 2160 train 4.252820 (lr=1.7400e-03) (hash(x)=75470031)
263
+ 2170 train 4.206608 (lr=1.7309e-03) (hash(x)=74307890)
264
+ 2180 train 4.322135 (lr=1.7217e-03) (hash(x)=77214245)
265
+ 2190 train 4.105399 (lr=1.7125e-03) (hash(x)=88628359)
266
+ 2200 val loss 4.1833
267
+ 2200 val perplexity 65.5842
268
+ 2200 train 4.117539 (lr=1.7033e-03) (hash(x)=74779126)
269
+ 2210 train 4.158056 (lr=1.6941e-03) (hash(x)=79817976)
270
+ 2220 train 4.112379 (lr=1.6848e-03) (hash(x)=75258996)
271
+ 2230 train 4.083318 (lr=1.6755e-03) (hash(x)=88164047)
272
+ 2240 train 4.068811 (lr=1.6662e-03) (hash(x)=78580686)
273
+ 2250 train 4.005162 (lr=1.6568e-03) (hash(x)=76510617)
274
+ 2260 train 4.083836 (lr=1.6475e-03) (hash(x)=76043966)
275
+ 2270 train 4.062353 (lr=1.6381e-03) (hash(x)=73062098)
276
+ 2280 train 4.286774 (lr=1.6287e-03) (hash(x)=81885909)
277
+ 2290 train 4.223174 (lr=1.6193e-03) (hash(x)=81481741)
278
+ 2300 val loss 4.1627
279
+ 2300 val perplexity 64.2464
280
+ 2300 train 4.085193 (lr=1.6098e-03) (hash(x)=82104275)
281
+ 2310 train 4.129548 (lr=1.6004e-03) (hash(x)=82137309)
282
+ 2320 train 4.108154 (lr=1.5909e-03) (hash(x)=86741167)
283
+ 2330 train 4.118347 (lr=1.5814e-03) (hash(x)=84796263)
284
+ 2340 train 4.090961 (lr=1.5719e-03) (hash(x)=169966529)
285
+ 2350 train 3.972981 (lr=1.5624e-03) (hash(x)=80006061)
286
+ 2360 train 4.028340 (lr=1.5529e-03) (hash(x)=71090999)
287
+ 2370 train 4.160749 (lr=1.5434e-03) (hash(x)=86641850)
288
+ 2380 train 4.057519 (lr=1.5338e-03) (hash(x)=84573253)
289
+ 2390 train 4.094037 (lr=1.5242e-03) (hash(x)=74483764)
290
+ 2400 val loss 4.1554
291
+ 2400 val perplexity 63.7806
292
+ 2400 train 4.062446 (lr=1.5147e-03) (hash(x)=78327659)
293
+ 2410 train 4.126875 (lr=1.5051e-03) (hash(x)=77273627)
294
+ 2420 train 3.991289 (lr=1.4955e-03) (hash(x)=76938049)
295
+ 2430 train 3.995062 (lr=1.4859e-03) (hash(x)=98449442)
296
+ 2440 train 3.854595 (lr=1.4763e-03) (hash(x)=78157797)
297
+ 2450 train 3.897134 (lr=1.4666e-03) (hash(x)=80637582)
298
+ 2460 train 3.886410 (lr=1.4570e-03) (hash(x)=80023854)
299
+ 2470 train 3.944814 (lr=1.4474e-03) (hash(x)=86678884)
300
+ 2480 train 4.170742 (lr=1.4377e-03) (hash(x)=84086469)
301
+ 2490 train 4.194137 (lr=1.4281e-03) (hash(x)=83131141)
302
+ 2500 val loss 4.1302
303
+ 2500 val perplexity 62.1918
304
+ 2500 train 4.107495 (lr=1.4184e-03) (hash(x)=82583497)
305
+ 2510 train 4.170167 (lr=1.4088e-03) (hash(x)=79164326)
306
+ 2520 train 4.077123 (lr=1.3991e-03) (hash(x)=76456503)
307
+ 2530 train 4.135890 (lr=1.3895e-03) (hash(x)=82001228)
308
+ 2540 train 4.171318 (lr=1.3798e-03) (hash(x)=75615595)
309
+ 2550 train 4.115598 (lr=1.3702e-03) (hash(x)=78397869)
310
+ 2560 train 4.117532 (lr=1.3605e-03) (hash(x)=75547032)
311
+ 2570 train 4.030212 (lr=1.3509e-03) (hash(x)=89201025)
312
+ 2580 train 4.016969 (lr=1.3412e-03) (hash(x)=89856704)
313
+ 2590 train 4.028861 (lr=1.3316e-03) (hash(x)=82175682)
314
+ 2600 val loss 4.1159
315
+ 2600 val perplexity 61.3101
316
+ 2600 train 4.032916 (lr=1.3219e-03) (hash(x)=78312826)
317
+ 2610 train 4.028169 (lr=1.3123e-03) (hash(x)=77066588)
318
+ 2620 train 3.868958 (lr=1.3026e-03) (hash(x)=78666061)
319
+ 2630 train 3.865645 (lr=1.2930e-03) (hash(x)=93762143)
320
+ 2640 train 3.881858 (lr=1.2834e-03) (hash(x)=83191587)
321
+ 2650 train 3.807027 (lr=1.2737e-03) (hash(x)=87169585)
322
+ 2660 train 4.022814 (lr=1.2641e-03) (hash(x)=86426388)
323
+ 2670 train 4.158422 (lr=1.2545e-03) (hash(x)=76692638)
324
+ 2680 train 4.069583 (lr=1.2449e-03) (hash(x)=77446063)
325
+ 2690 train 4.072866 (lr=1.2353e-03) (hash(x)=79809050)
326
+ 2700 val loss 4.1070
327
+ 2700 val perplexity 60.7624
328
+ 2700 train 3.998014 (lr=1.2258e-03) (hash(x)=83116823)
329
+ 2710 train 4.114951 (lr=1.2162e-03) (hash(x)=75622148)
330
+ 2720 train 4.040291 (lr=1.2066e-03) (hash(x)=80690512)
331
+ 2730 train 4.111761 (lr=1.1971e-03) (hash(x)=78009984)
332
+ 2740 train 4.065063 (lr=1.1876e-03) (hash(x)=74205488)
333
+ 2750 train 4.069279 (lr=1.1781e-03) (hash(x)=91013332)
334
+ 2760 train 3.992820 (lr=1.1686e-03) (hash(x)=76954961)
335
+ 2770 train 3.998701 (lr=1.1591e-03) (hash(x)=79390317)
336
+ 2780 train 3.872392 (lr=1.1496e-03) (hash(x)=70168783)
337
+ 2790 train 4.032706 (lr=1.1402e-03) (hash(x)=76028417)
338
+ 2800 val loss 4.0878
339
+ 2800 val perplexity 59.6094
340
+ 2800 train 3.859599 (lr=1.1307e-03) (hash(x)=77656050)
341
+ 2810 train 3.905592 (lr=1.1213e-03) (hash(x)=94248216)
342
+ 2820 train 3.968545 (lr=1.1119e-03) (hash(x)=78305078)
343
+ 2830 train 3.881601 (lr=1.1025e-03) (hash(x)=79948848)
344
+ 2840 train 3.745872 (lr=1.0932e-03) (hash(x)=85341024)
345
+ 2850 train 4.055525 (lr=1.0838e-03) (hash(x)=78735170)
346
+ 2860 train 4.305941 (lr=1.0745e-03) (hash(x)=71616419)
347
+ 2870 train 4.077571 (lr=1.0652e-03) (hash(x)=78656517)
348
+ 2880 train 4.089402 (lr=1.0559e-03) (hash(x)=80073987)
349
+ 2890 train 4.019093 (lr=1.0467e-03) (hash(x)=76894809)
350
+ 2900 val loss 4.0640
351
+ 2900 val perplexity 58.2050
352
+ 2900 train 3.999931 (lr=1.0375e-03) (hash(x)=80499838)
353
+ 2910 train 4.046417 (lr=1.0283e-03) (hash(x)=72673354)
354
+ 2920 train 4.026885 (lr=1.0191e-03) (hash(x)=84265768)
355
+ 2930 train 4.056563 (lr=1.0100e-03) (hash(x)=79612060)
356
+ 2940 train 3.912519 (lr=1.0008e-03) (hash(x)=74970087)
357
+ 2950 train 4.067937 (lr=9.9174e-04) (hash(x)=84166818)
358
+ 2960 train 4.059937 (lr=9.8268e-04) (hash(x)=89410221)
359
+ 2970 train 3.996826 (lr=9.7364e-04) (hash(x)=75672566)
360
+ 2980 train 3.852962 (lr=9.6463e-04) (hash(x)=81760314)
361
+ 2990 train 3.825138 (lr=9.5566e-04) (hash(x)=80605200)
362
+ 3000 val loss 4.0617
363
+ 3000 val perplexity 58.0734
364
+ 3000 train 3.820753 (lr=9.4671e-04) (hash(x)=83804735)
365
+ 3010 train 3.749279 (lr=9.3780e-04) (hash(x)=77015303)
366
+ 3020 train 3.817938 (lr=9.2892e-04) (hash(x)=81464523)
367
+ 3030 train 4.014338 (lr=9.2007e-04) (hash(x)=80168230)
368
+ 3040 train 3.959165 (lr=9.1125e-04) (hash(x)=84817006)
369
+ 3050 train 4.064665 (lr=9.0247e-04) (hash(x)=71601811)
370
+ 3060 train 3.982501 (lr=8.9373e-04) (hash(x)=85499733)
371
+ 3070 train 4.042049 (lr=8.8502e-04) (hash(x)=78670408)
372
+ 3080 train 3.936097 (lr=8.7634e-04) (hash(x)=77120468)
373
+ 3090 train 4.041956 (lr=8.6771e-04) (hash(x)=77927426)
374
+ 3100 val loss 4.0409
375
+ 3100 val perplexity 56.8759
376
+ 3100 train 4.056437 (lr=8.5911e-04) (hash(x)=83998606)
377
+ 3110 train 3.957352 (lr=8.5054e-04) (hash(x)=77990218)
378
+ 3120 train 3.870935 (lr=8.4202e-04) (hash(x)=81623970)
379
+ 3130 train 3.842823 (lr=8.3353e-04) (hash(x)=75014781)
380
+ 3140 train 3.901370 (lr=8.2509e-04) (hash(x)=72591250)
381
+ 3150 train 3.909689 (lr=8.1669e-04) (hash(x)=81421847)
382
+ 3160 train 3.957351 (lr=8.0832e-04) (hash(x)=75112631)
383
+ 3170 train 4.057568 (lr=8.0000e-04) (hash(x)=87518033)
384
+ 3180 train 3.904346 (lr=7.9172e-04) (hash(x)=78648348)
385
+ 3190 train 4.022759 (lr=7.8349e-04) (hash(x)=78270029)
386
+ 3200 val loss 4.0208
387
+ 3200 val perplexity 55.7464
388
+ 3200 train 4.094899 (lr=7.7530e-04) (hash(x)=83380714)
389
+ 3210 train 4.277036 (lr=7.6715e-04) (hash(x)=67804991)
390
+ 3220 train 4.008252 (lr=7.5905e-04) (hash(x)=76968804)
391
+ 3230 train 3.938192 (lr=7.5099e-04) (hash(x)=75798670)
392
+ 3240 train 4.015611 (lr=7.4298e-04) (hash(x)=83192811)
393
+ 3250 train 3.953701 (lr=7.3501e-04) (hash(x)=81659789)
394
+ 3260 train 3.997344 (lr=7.2709e-04) (hash(x)=80622502)
395
+ 3270 train 3.970113 (lr=7.1922e-04) (hash(x)=85436511)
396
+ 3280 train 3.908877 (lr=7.1140e-04) (hash(x)=80589180)
397
+ 3290 train 4.046679 (lr=7.0363e-04) (hash(x)=85747193)
398
+ 3300 val loss 4.0184
399
+ 3300 val perplexity 55.6148
400
+ 3300 train 3.917713 (lr=6.9590e-04) (hash(x)=74635692)
401
+ 3310 train 3.901635 (lr=6.8823e-04) (hash(x)=82045455)
402
+ 3320 train 3.769691 (lr=6.8061e-04) (hash(x)=73249173)
403
+ 3330 train 3.729814 (lr=6.7303e-04) (hash(x)=71492338)
404
+ 3340 train 4.064158 (lr=6.6551e-04) (hash(x)=76374071)
405
+ 3350 train 4.036660 (lr=6.5805e-04) (hash(x)=80474064)
406
+ 3360 train 3.945580 (lr=6.5063e-04) (hash(x)=84390892)
407
+ 3370 train 3.988594 (lr=6.4327e-04) (hash(x)=83399949)
408
+ 3380 train 4.039264 (lr=6.3596e-04) (hash(x)=77648059)
409
+ 3390 train 4.013268 (lr=6.2871e-04) (hash(x)=80152701)
410
+ 3400 val loss 3.9962
411
+ 3400 val perplexity 54.3912
412
+ 3400 train 4.065991 (lr=6.2151e-04) (hash(x)=81952545)
413
+ 3410 train 4.020948 (lr=6.1437e-04) (hash(x)=80908993)
414
+ 3420 train 4.063770 (lr=6.0728e-04) (hash(x)=79610037)
415
+ 3430 train 3.935652 (lr=6.0025e-04) (hash(x)=87624382)
416
+ 3440 train 3.988576 (lr=5.9328e-04) (hash(x)=82336381)
417
+ 3450 train 3.966711 (lr=5.8636e-04) (hash(x)=87687835)
418
+ 3460 train 3.969498 (lr=5.7950e-04) (hash(x)=76376135)
419
+ 3470 train 3.876685 (lr=5.7271e-04) (hash(x)=79331391)
420
+ 3480 train 3.786352 (lr=5.6597e-04) (hash(x)=107002681)
421
+ 3490 train 3.865254 (lr=5.5929e-04) (hash(x)=77465514)
422
+ 3500 val loss 3.9933
423
+ 3500 val perplexity 54.2329
424
+ 3500 train 3.825722 (lr=5.5267e-04) (hash(x)=88237229)
425
+ 3510 train 3.898677 (lr=5.4611e-04) (hash(x)=81011739)
426
+ 3520 train 3.947072 (lr=5.3961e-04) (hash(x)=64643427)
427
+ 3530 train 4.224852 (lr=5.3317e-04) (hash(x)=78029539)
428
+ 3540 train 4.054970 (lr=5.2680e-04) (hash(x)=83188968)
429
+ 3550 train 4.075386 (lr=5.2048e-04) (hash(x)=86104185)
430
+ 3560 train 3.993020 (lr=5.1424e-04) (hash(x)=81993629)
431
+ 3570 train 3.967967 (lr=5.0805e-04) (hash(x)=72719368)
432
+ 3580 train 3.976884 (lr=5.0193e-04) (hash(x)=90091487)
433
+ 3590 train 3.864897 (lr=4.9587e-04) (hash(x)=77393152)
434
+ 3600 val loss 3.9733
435
+ 3600 val perplexity 53.1623
436
+ 3600 train 3.974158 (lr=4.8988e-04) (hash(x)=73103504)
437
+ 3610 train 3.983311 (lr=4.8395e-04) (hash(x)=84429400)
438
+ 3620 train 3.953668 (lr=4.7809e-04) (hash(x)=77021795)
439
+ 3630 train 3.908778 (lr=4.7229e-04) (hash(x)=92830605)
440
+ 3640 train 4.104006 (lr=4.6656e-04) (hash(x)=78313175)
441
+ 3650 train 3.905580 (lr=4.6090e-04) (hash(x)=85395549)
442
+ 3660 train 3.699963 (lr=4.5530e-04) (hash(x)=78114459)
443
+ 3670 train 3.828334 (lr=4.4978e-04) (hash(x)=74968316)
444
+ 3680 train 3.753552 (lr=4.4432e-04) (hash(x)=73358737)
445
+ 3690 train 3.856151 (lr=4.3893e-04) (hash(x)=76399442)
446
+ 3700 val loss 3.9746
447
+ 3700 val perplexity 53.2277
448
+ 3700 train 3.857806 (lr=4.3360e-04) (hash(x)=74039273)
449
+ 3710 train 4.036342 (lr=4.2835e-04) (hash(x)=79299680)
450
+ 3720 train 3.946940 (lr=4.2317e-04) (hash(x)=94701498)
451
+ 3730 train 4.029010 (lr=4.1806e-04) (hash(x)=75352071)
452
+ 3740 train 3.907064 (lr=4.1302e-04) (hash(x)=77909487)
453
+ 3750 train 4.016834 (lr=4.0804e-04) (hash(x)=73986730)
454
+ 3760 train 3.970569 (lr=4.0314e-04) (hash(x)=79325763)
455
+ 3770 train 4.018567 (lr=3.9832e-04) (hash(x)=72457818)
456
+ 3780 train 4.146557 (lr=3.9356e-04) (hash(x)=71775590)
457
+ 3790 train 4.031501 (lr=3.8888e-04) (hash(x)=82638943)
458
+ 3800 val loss 3.9566
459
+ 3800 val perplexity 52.2792
460
+ 3800 train 4.011085 (lr=3.8426e-04) (hash(x)=79965893)
461
+ 3810 train 3.943778 (lr=3.7973e-04) (hash(x)=72598235)
462
+ 3820 train 3.929773 (lr=3.7526e-04) (hash(x)=83113889)
463
+ 3830 train 3.990464 (lr=3.7087e-04) (hash(x)=74434590)
464
+ 3840 train 3.901638 (lr=3.6655e-04) (hash(x)=82860348)
465
+ 3850 train 3.842507 (lr=3.6231e-04) (hash(x)=78067565)
466
+ 3860 train 3.910461 (lr=3.5814e-04) (hash(x)=82592498)
467
+ 3870 train 3.841162 (lr=3.5405e-04) (hash(x)=81820733)
468
+ 3880 train 3.858931 (lr=3.5003e-04) (hash(x)=87709040)
469
+ 3890 train 3.903772 (lr=3.4609e-04) (hash(x)=70379093)
470
+ 3900 val loss 3.9505
471
+ 3900 val perplexity 51.9632
472
+ 3900 train 4.027209 (lr=3.4222e-04) (hash(x)=76597431)
473
+ 3910 train 3.908327 (lr=3.3843e-04) (hash(x)=90490716)
474
+ 3920 train 4.019783 (lr=3.3471e-04) (hash(x)=81970659)
475
+ 3930 train 3.999418 (lr=3.3108e-04) (hash(x)=81496334)
476
+ 3940 train 3.975254 (lr=3.2752e-04) (hash(x)=75717605)
477
+ 3950 train 3.924225 (lr=3.2403e-04) (hash(x)=82749357)
478
+ 3960 train 3.949820 (lr=3.2063e-04) (hash(x)=89355157)
479
+ 3970 train 3.906679 (lr=3.1730e-04) (hash(x)=78980403)
480
+ 3980 train 3.880381 (lr=3.1405e-04) (hash(x)=76627217)
481
+ 3990 train 3.884381 (lr=3.1087e-04) (hash(x)=72412879)
482
+ 4000 val loss 3.9451
483
+ 4000 val perplexity 51.6834
484
+ 4000 train 3.897163 (lr=3.0778e-04) (hash(x)=83018142)
485
+ 4010 train 3.912729 (lr=3.0476e-04) (hash(x)=81272436)
486
+ 4020 train 3.840531 (lr=3.0183e-04) (hash(x)=85497482)
487
+ 4030 train 3.911720 (lr=2.9897e-04) (hash(x)=85241734)
488
+ 4040 train 3.875229 (lr=2.9619e-04) (hash(x)=77925307)
489
+ 4050 train 3.880402 (lr=2.9349e-04) (hash(x)=84826179)
490
+ 4060 train 3.937480 (lr=2.9087e-04) (hash(x)=83606764)
491
+ 4070 train 3.845355 (lr=2.8833e-04) (hash(x)=80567590)
492
+ 4080 train 3.900438 (lr=2.8587e-04) (hash(x)=76860998)
493
+ 4090 train 3.933043 (lr=2.8350e-04) (hash(x)=74902328)
494
+ 4100 val loss 3.9329
495
+ 4100 val perplexity 51.0568
496
+ 4100 train 3.978987 (lr=2.8120e-04) (hash(x)=82832041)
497
+ 4110 train 3.858107 (lr=2.7898e-04) (hash(x)=79143262)
498
+ 4120 train 3.888119 (lr=2.7684e-04) (hash(x)=77038149)
499
+ 4130 train 4.064957 (lr=2.7479e-04) (hash(x)=86339074)
500
+ 4140 train 4.137596 (lr=2.7281e-04) (hash(x)=76686216)
501
+ 4150 train 3.909303 (lr=2.7092e-04) (hash(x)=70522682)
502
+ 4160 train 3.870981 (lr=2.6910e-04) (hash(x)=90958555)
503
+ 4170 train 3.896600 (lr=2.6737e-04) (hash(x)=91463532)
504
+ 4180 train 3.888387 (lr=2.6572e-04) (hash(x)=81959329)
505
+ 4190 train 3.841862 (lr=2.6415e-04) (hash(x)=83146752)
506
+ 4200 val loss 3.9355
507
+ 4200 val perplexity 51.1868
508
+ 4200 train 3.826995 (lr=2.6267e-04) (hash(x)=78361715)
509
+ 4210 train 3.973566 (lr=2.6126e-04) (hash(x)=87364889)
510
+ 4220 train 3.831426 (lr=2.5994e-04) (hash(x)=70465156)
511
+ 4230 train 3.985258 (lr=2.5870e-04) (hash(x)=84524081)
512
+ 4240 train 3.935690 (lr=2.5754e-04) (hash(x)=77824868)
513
+ 4250 train 3.946224 (lr=2.5647e-04) (hash(x)=81710711)
514
+ 4260 train 3.934658 (lr=2.5548e-04) (hash(x)=76362728)
515
+ 4270 train 3.924897 (lr=2.5457e-04) (hash(x)=83115208)
516
+ 4280 train 3.837732 (lr=2.5374e-04) (hash(x)=87218314)
517
+ 4290 train 3.997554 (lr=2.5299e-04) (hash(x)=74582673)
518
+ 4300 val loss 3.9300
519
+ 4300 val perplexity 50.9092
520
+ 4300 train 3.872051 (lr=2.5233e-04) (hash(x)=77379615)
521
+ 4310 train 3.846852 (lr=2.5175e-04) (hash(x)=78669579)
522
+ 4320 train 3.963589 (lr=2.5125e-04) (hash(x)=83066608)
523
+ 4330 train 3.928256 (lr=2.5084e-04) (hash(x)=83037340)
524
+ 4340 train 3.964064 (lr=2.5051e-04) (hash(x)=82849771)
525
+ 4350 train 3.899836 (lr=2.5026e-04) (hash(x)=76693985)
526
+ 4360 train 3.991380 (lr=2.5009e-04) (hash(x)=77745394)
527
+ 4370 train 3.953964 (lr=2.5001e-04) (hash(x)=79954388)
528
+ 4374 val loss 3.9237
529
+ 4374 val perplexity 50.5886
12_head_baseline_lr_25e-4_head_dim_22_fixed/model_04374.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:737828245bd17af4e09dc6a15699e002d1da4b14f3613638fa4a21a0900da372
3
+ size 96858242
12_head_baseline_lr_25e-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:ed8a2d69f0c513df226046cd28ca6ed3faff48df7ac29604f6d381f8e03cb654
3
+ size 187435910