andrew-healey commited on
Commit
a22bc25
·
verified ·
1 Parent(s): 968e9da

Upload folder using huggingface_hub

Browse files
12_head_baseline_lr_10e-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_10e-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": 0.001, "decay_lr": true, "readout_zero_init": false, "query_zero_init": false, "l1_loss": false, "debugpy": false, "key": "12_head_baseline_lr_10e-4_fixed"}
12_head_baseline_lr_10e-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_10e-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 89706.7812
4
+ 0 train 11.388923 (lr=1.3986e-06) (hash(x)=93356070)
5
+ 10 train 10.383682 (lr=1.5385e-05) (hash(x)=91253010)
6
+ 20 train 9.903660 (lr=2.9371e-05) (hash(x)=74934453)
7
+ 30 train 9.639926 (lr=4.3357e-05) (hash(x)=79278034)
8
+ 40 train 9.211796 (lr=5.7343e-05) (hash(x)=80107892)
9
+ 50 train 8.756122 (lr=7.1329e-05) (hash(x)=70745428)
10
+ 60 train 8.394462 (lr=8.5315e-05) (hash(x)=80077589)
11
+ 70 train 7.865565 (lr=9.9301e-05) (hash(x)=76213766)
12
+ 80 train 7.714816 (lr=1.1329e-04) (hash(x)=83218328)
13
+ 90 train 7.563947 (lr=1.2727e-04) (hash(x)=74855845)
14
+ 100 val loss 7.4883
15
+ 100 val perplexity 1787.0599
16
+ 100 train 7.345553 (lr=1.4126e-04) (hash(x)=82814902)
17
+ 110 train 7.291364 (lr=1.5524e-04) (hash(x)=84907741)
18
+ 120 train 7.182977 (lr=1.6923e-04) (hash(x)=82613223)
19
+ 130 train 7.011628 (lr=1.8322e-04) (hash(x)=83540876)
20
+ 140 train 6.975050 (lr=1.9720e-04) (hash(x)=75095216)
21
+ 150 train 7.027579 (lr=2.1119e-04) (hash(x)=97190944)
22
+ 160 train 6.913074 (lr=2.2517e-04) (hash(x)=82117809)
23
+ 170 train 6.804592 (lr=2.3916e-04) (hash(x)=70514724)
24
+ 180 train 6.720253 (lr=2.5315e-04) (hash(x)=81029624)
25
+ 190 train 6.641295 (lr=2.6713e-04) (hash(x)=78996841)
26
+ 200 val loss 6.5694
27
+ 200 val perplexity 712.9483
28
+ 200 train 6.573037 (lr=2.8112e-04) (hash(x)=79845097)
29
+ 210 train 6.446847 (lr=2.9510e-04) (hash(x)=78997683)
30
+ 220 train 6.337046 (lr=3.0909e-04) (hash(x)=74895865)
31
+ 230 train 6.502890 (lr=3.2308e-04) (hash(x)=80933276)
32
+ 240 train 6.211671 (lr=3.3706e-04) (hash(x)=77664606)
33
+ 250 train 6.249459 (lr=3.5105e-04) (hash(x)=75903930)
34
+ 260 train 6.158349 (lr=3.6503e-04) (hash(x)=82996853)
35
+ 270 train 6.076314 (lr=3.7902e-04) (hash(x)=73269568)
36
+ 280 train 6.017995 (lr=3.9301e-04) (hash(x)=82906427)
37
+ 290 train 5.940077 (lr=4.0699e-04) (hash(x)=71797895)
38
+ 300 val loss 6.0766
39
+ 300 val perplexity 435.5448
40
+ 300 train 5.799405 (lr=4.2098e-04) (hash(x)=85232249)
41
+ 310 train 5.733637 (lr=4.3497e-04) (hash(x)=75252489)
42
+ 320 train 5.829608 (lr=4.4895e-04) (hash(x)=74394644)
43
+ 330 train 5.897291 (lr=4.6294e-04) (hash(x)=82772910)
44
+ 340 train 5.933196 (lr=4.7692e-04) (hash(x)=81627464)
45
+ 350 train 5.929248 (lr=4.9091e-04) (hash(x)=84678053)
46
+ 360 train 5.865786 (lr=5.0490e-04) (hash(x)=81884128)
47
+ 370 train 5.858202 (lr=5.1888e-04) (hash(x)=76379242)
48
+ 380 train 5.637664 (lr=5.3287e-04) (hash(x)=98182875)
49
+ 390 train 5.636794 (lr=5.4685e-04) (hash(x)=79710436)
50
+ 400 val loss 5.6822
51
+ 400 val perplexity 293.6012
52
+ 400 train 5.631232 (lr=5.6084e-04) (hash(x)=79841071)
53
+ 410 train 5.556250 (lr=5.7483e-04) (hash(x)=75844151)
54
+ 420 train 5.575057 (lr=5.8881e-04) (hash(x)=73125036)
55
+ 430 train 5.602938 (lr=6.0280e-04) (hash(x)=84214858)
56
+ 440 train 5.491806 (lr=6.1678e-04) (hash(x)=80456994)
57
+ 450 train 5.459967 (lr=6.3077e-04) (hash(x)=76735962)
58
+ 460 train 5.397115 (lr=6.4476e-04) (hash(x)=81845446)
59
+ 470 train 5.199123 (lr=6.5874e-04) (hash(x)=76094689)
60
+ 480 train 5.313327 (lr=6.7273e-04) (hash(x)=83806686)
61
+ 490 train 5.144130 (lr=6.8671e-04) (hash(x)=84690227)
62
+ 500 val loss 5.4644
63
+ 500 val perplexity 236.1261
64
+ 500 train 5.221138 (lr=7.0070e-04) (hash(x)=71851938)
65
+ 510 train 5.310715 (lr=7.1469e-04) (hash(x)=77159346)
66
+ 520 train 5.371479 (lr=7.2867e-04) (hash(x)=80755753)
67
+ 530 train 5.469559 (lr=7.4266e-04) (hash(x)=75379679)
68
+ 540 train 5.375363 (lr=7.5664e-04) (hash(x)=82458619)
69
+ 550 train 5.357809 (lr=7.7063e-04) (hash(x)=86366396)
70
+ 560 train 5.358559 (lr=7.8462e-04) (hash(x)=85095044)
71
+ 570 train 5.316364 (lr=7.9860e-04) (hash(x)=78385159)
72
+ 580 train 5.267788 (lr=8.1259e-04) (hash(x)=79342394)
73
+ 590 train 5.271762 (lr=8.2657e-04) (hash(x)=70782192)
74
+ 600 val loss 5.2663
75
+ 600 val perplexity 193.6914
76
+ 600 train 5.247325 (lr=8.4056e-04) (hash(x)=76130353)
77
+ 610 train 5.156121 (lr=8.5455e-04) (hash(x)=74778440)
78
+ 620 train 5.141376 (lr=8.6853e-04) (hash(x)=79129709)
79
+ 630 train 5.036836 (lr=8.8252e-04) (hash(x)=76469962)
80
+ 640 train 5.035790 (lr=8.9650e-04) (hash(x)=78288049)
81
+ 650 train 4.992970 (lr=9.1049e-04) (hash(x)=76641388)
82
+ 660 train 4.908565 (lr=9.2448e-04) (hash(x)=75937906)
83
+ 670 train 4.910959 (lr=9.3846e-04) (hash(x)=74807157)
84
+ 680 train 4.832582 (lr=9.5245e-04) (hash(x)=77490144)
85
+ 690 train 4.855207 (lr=9.6643e-04) (hash(x)=72900124)
86
+ 700 val loss 5.1190
87
+ 700 val perplexity 167.1615
88
+ 700 train 4.814493 (lr=9.8042e-04) (hash(x)=73424218)
89
+ 710 train 5.102498 (lr=9.9441e-04) (hash(x)=86052345)
90
+ 720 train 5.059681 (lr=1.0000e-03) (hash(x)=77462613)
91
+ 730 train 5.070732 (lr=9.9996e-04) (hash(x)=79027471)
92
+ 740 train 5.038265 (lr=9.9990e-04) (hash(x)=78149992)
93
+ 750 train 5.009299 (lr=9.9980e-04) (hash(x)=89147499)
94
+ 760 train 5.005920 (lr=9.9966e-04) (hash(x)=74931538)
95
+ 770 train 4.954458 (lr=9.9950e-04) (hash(x)=89648204)
96
+ 780 train 5.006054 (lr=9.9930e-04) (hash(x)=83708293)
97
+ 790 train 4.997158 (lr=9.9907e-04) (hash(x)=84336512)
98
+ 800 val loss 4.9588
99
+ 800 val perplexity 142.4243
100
+ 800 train 4.877936 (lr=9.9880e-04) (hash(x)=75025285)
101
+ 810 train 4.836957 (lr=9.9850e-04) (hash(x)=79436536)
102
+ 820 train 4.803548 (lr=9.9817e-04) (hash(x)=90467390)
103
+ 830 train 4.843698 (lr=9.9781e-04) (hash(x)=77292020)
104
+ 840 train 4.743368 (lr=9.9741e-04) (hash(x)=75568927)
105
+ 850 train 4.768930 (lr=9.9698e-04) (hash(x)=79671781)
106
+ 860 train 4.853292 (lr=9.9652e-04) (hash(x)=84280943)
107
+ 870 train 4.833578 (lr=9.9602e-04) (hash(x)=80901294)
108
+ 880 train 4.948204 (lr=9.9549e-04) (hash(x)=81437584)
109
+ 890 train 4.796063 (lr=9.9493e-04) (hash(x)=73893778)
110
+ 900 val loss 4.8156
111
+ 900 val perplexity 123.4218
112
+ 900 train 4.738143 (lr=9.9434e-04) (hash(x)=74987794)
113
+ 910 train 4.758990 (lr=9.9371e-04) (hash(x)=93721374)
114
+ 920 train 4.739633 (lr=9.9305e-04) (hash(x)=79149678)
115
+ 930 train 4.687983 (lr=9.9236e-04) (hash(x)=83179387)
116
+ 940 train 4.693848 (lr=9.9163e-04) (hash(x)=89009978)
117
+ 950 train 4.659063 (lr=9.9088e-04) (hash(x)=76483735)
118
+ 960 train 4.593154 (lr=9.9009e-04) (hash(x)=78638850)
119
+ 970 train 4.624710 (lr=9.8926e-04) (hash(x)=90415206)
120
+ 980 train 4.561686 (lr=9.8841e-04) (hash(x)=90819195)
121
+ 990 train 4.685452 (lr=9.8752e-04) (hash(x)=85938852)
122
+ 1000 val loss 4.6702
123
+ 1000 val perplexity 106.7176
124
+ 1000 train 4.720173 (lr=9.8660e-04) (hash(x)=82436789)
125
+ 1010 train 4.748746 (lr=9.8565e-04) (hash(x)=63022149)
126
+ 1020 train 4.685092 (lr=9.8467e-04) (hash(x)=83501199)
127
+ 1030 train 4.660248 (lr=9.8365e-04) (hash(x)=77093326)
128
+ 1040 train 4.532311 (lr=9.8260e-04) (hash(x)=73437559)
129
+ 1050 train 4.626408 (lr=9.8152e-04) (hash(x)=84550388)
130
+ 1060 train 4.619750 (lr=9.8041e-04) (hash(x)=86184566)
131
+ 1070 train 4.612539 (lr=9.7927e-04) (hash(x)=85644922)
132
+ 1080 train 4.547601 (lr=9.7809e-04) (hash(x)=86179801)
133
+ 1090 train 4.516680 (lr=9.7689e-04) (hash(x)=82868303)
134
+ 1100 val loss 4.5865
135
+ 1100 val perplexity 98.1462
136
+ 1100 train 4.620056 (lr=9.7565e-04) (hash(x)=96780388)
137
+ 1110 train 4.555966 (lr=9.7438e-04) (hash(x)=87223122)
138
+ 1120 train 4.521392 (lr=9.7308e-04) (hash(x)=77292786)
139
+ 1130 train 4.539902 (lr=9.7175e-04) (hash(x)=88761375)
140
+ 1140 train 4.455670 (lr=9.7039e-04) (hash(x)=80956468)
141
+ 1150 train 4.634190 (lr=9.6899e-04) (hash(x)=84725462)
142
+ 1160 train 4.606533 (lr=9.6757e-04) (hash(x)=76098113)
143
+ 1170 train 4.580040 (lr=9.6611e-04) (hash(x)=84387685)
144
+ 1180 train 4.525800 (lr=9.6463e-04) (hash(x)=76120321)
145
+ 1190 train 4.488889 (lr=9.6311e-04) (hash(x)=85974065)
146
+ 1200 val loss 4.5049
147
+ 1200 val perplexity 90.4565
148
+ 1200 train 4.388789 (lr=9.6157e-04) (hash(x)=79340644)
149
+ 1210 train 4.485209 (lr=9.5999e-04) (hash(x)=88345025)
150
+ 1220 train 4.392861 (lr=9.5838e-04) (hash(x)=86406230)
151
+ 1230 train 4.377093 (lr=9.5674e-04) (hash(x)=80608975)
152
+ 1240 train 4.429608 (lr=9.5508e-04) (hash(x)=81635225)
153
+ 1250 train 4.267586 (lr=9.5338e-04) (hash(x)=82126376)
154
+ 1260 train 4.444512 (lr=9.5165e-04) (hash(x)=83737972)
155
+ 1270 train 4.352499 (lr=9.4990e-04) (hash(x)=81132345)
156
+ 1280 train 4.439312 (lr=9.4811e-04) (hash(x)=85850404)
157
+ 1290 train 4.445881 (lr=9.4629e-04) (hash(x)=83490640)
158
+ 1300 val loss 4.4533
159
+ 1300 val perplexity 85.9066
160
+ 1300 train 4.422691 (lr=9.4445e-04) (hash(x)=77545187)
161
+ 1310 train 4.474619 (lr=9.4258e-04) (hash(x)=86412685)
162
+ 1320 train 4.367522 (lr=9.4067e-04) (hash(x)=86429640)
163
+ 1330 train 4.506609 (lr=9.3874e-04) (hash(x)=79645281)
164
+ 1340 train 4.457325 (lr=9.3678e-04) (hash(x)=82668541)
165
+ 1350 train 4.318626 (lr=9.3479e-04) (hash(x)=73627248)
166
+ 1360 train 4.415100 (lr=9.3278e-04) (hash(x)=74845133)
167
+ 1370 train 4.324305 (lr=9.3073e-04) (hash(x)=77850497)
168
+ 1380 train 4.349428 (lr=9.2866e-04) (hash(x)=76143954)
169
+ 1390 train 4.363151 (lr=9.2656e-04) (hash(x)=80675544)
170
+ 1400 val loss 4.4256
171
+ 1400 val perplexity 83.5655
172
+ 1400 train 4.333576 (lr=9.2443e-04) (hash(x)=76010938)
173
+ 1410 train 4.382585 (lr=9.2227e-04) (hash(x)=81260300)
174
+ 1420 train 4.491401 (lr=9.2009e-04) (hash(x)=83803629)
175
+ 1430 train 4.328485 (lr=9.1788e-04) (hash(x)=83840110)
176
+ 1440 train 4.522609 (lr=9.1564e-04) (hash(x)=97711831)
177
+ 1450 train 4.393065 (lr=9.1338e-04) (hash(x)=81110139)
178
+ 1460 train 4.313437 (lr=9.1108e-04) (hash(x)=90942839)
179
+ 1470 train 4.351507 (lr=9.0877e-04) (hash(x)=80721564)
180
+ 1480 train 4.447525 (lr=9.0642e-04) (hash(x)=77852059)
181
+ 1490 train 4.340530 (lr=9.0405e-04) (hash(x)=80011365)
182
+ 1500 val loss 4.3747
183
+ 1500 val perplexity 79.4159
184
+ 1500 train 4.362991 (lr=9.0165e-04) (hash(x)=84496142)
185
+ 1510 train 4.303360 (lr=8.9923e-04) (hash(x)=70186729)
186
+ 1520 train 4.330505 (lr=8.9678e-04) (hash(x)=74854227)
187
+ 1530 train 4.233740 (lr=8.9431e-04) (hash(x)=77468161)
188
+ 1540 train 4.245420 (lr=8.9181e-04) (hash(x)=87853059)
189
+ 1550 train 4.183795 (lr=8.8928e-04) (hash(x)=71225436)
190
+ 1560 train 4.281275 (lr=8.8673e-04) (hash(x)=84072783)
191
+ 1570 train 4.331686 (lr=8.8416e-04) (hash(x)=72723098)
192
+ 1580 train 4.350638 (lr=8.8156e-04) (hash(x)=90409866)
193
+ 1590 train 4.420529 (lr=8.7893e-04) (hash(x)=87481378)
194
+ 1600 val loss 4.3485
195
+ 1600 val perplexity 77.3634
196
+ 1600 train 4.406886 (lr=8.7629e-04) (hash(x)=77643862)
197
+ 1610 train 4.294946 (lr=8.7361e-04) (hash(x)=88862575)
198
+ 1620 train 4.307903 (lr=8.7092e-04) (hash(x)=84612581)
199
+ 1630 train 4.332493 (lr=8.6820e-04) (hash(x)=87075989)
200
+ 1640 train 4.223543 (lr=8.6546e-04) (hash(x)=88277361)
201
+ 1650 train 4.338336 (lr=8.6269e-04) (hash(x)=78750236)
202
+ 1660 train 4.263496 (lr=8.5990e-04) (hash(x)=82604581)
203
+ 1670 train 4.101541 (lr=8.5709e-04) (hash(x)=68482265)
204
+ 1680 train 4.222331 (lr=8.5425e-04) (hash(x)=75088835)
205
+ 1690 train 4.170703 (lr=8.5140e-04) (hash(x)=66060989)
206
+ 1700 val loss 4.3409
207
+ 1700 val perplexity 76.7748
208
+ 1700 train 4.337801 (lr=8.4852e-04) (hash(x)=79986754)
209
+ 1710 train 4.324283 (lr=8.4562e-04) (hash(x)=83657930)
210
+ 1720 train 4.272106 (lr=8.4269e-04) (hash(x)=81754135)
211
+ 1730 train 4.348814 (lr=8.3975e-04) (hash(x)=78904427)
212
+ 1740 train 4.248496 (lr=8.3678e-04) (hash(x)=85920177)
213
+ 1750 train 4.304587 (lr=8.3379e-04) (hash(x)=86573211)
214
+ 1760 train 4.255859 (lr=8.3079e-04) (hash(x)=81737128)
215
+ 1770 train 4.247238 (lr=8.2776e-04) (hash(x)=77400968)
216
+ 1780 train 4.265047 (lr=8.2471e-04) (hash(x)=73545497)
217
+ 1790 train 4.511281 (lr=8.2164e-04) (hash(x)=71641943)
218
+ 1800 val loss 4.2918
219
+ 1800 val perplexity 73.0970
220
+ 1800 train 4.350615 (lr=8.1855e-04) (hash(x)=87819781)
221
+ 1810 train 4.213409 (lr=8.1544e-04) (hash(x)=86870770)
222
+ 1820 train 4.224694 (lr=8.1231e-04) (hash(x)=82522211)
223
+ 1830 train 4.308731 (lr=8.0916e-04) (hash(x)=61947437)
224
+ 1840 train 4.182499 (lr=8.0600e-04) (hash(x)=79865406)
225
+ 1850 train 4.153718 (lr=8.0281e-04) (hash(x)=79828721)
226
+ 1860 train 4.294019 (lr=7.9961e-04) (hash(x)=80869571)
227
+ 1870 train 4.278061 (lr=7.9638e-04) (hash(x)=73780971)
228
+ 1880 train 4.241172 (lr=7.9314e-04) (hash(x)=79249549)
229
+ 1890 train 4.301267 (lr=7.8988e-04) (hash(x)=81041904)
230
+ 1900 val loss 4.2693
231
+ 1900 val perplexity 71.4708
232
+ 1900 train 4.360258 (lr=7.8660e-04) (hash(x)=82456430)
233
+ 1910 train 4.132666 (lr=7.8331e-04) (hash(x)=82222135)
234
+ 1920 train 4.170662 (lr=7.8000e-04) (hash(x)=70033249)
235
+ 1930 train 4.151264 (lr=7.7667e-04) (hash(x)=72887360)
236
+ 1940 train 4.198902 (lr=7.7333e-04) (hash(x)=83251100)
237
+ 1950 train 4.243005 (lr=7.6996e-04) (hash(x)=79660266)
238
+ 1960 train 4.121330 (lr=7.6659e-04) (hash(x)=78919068)
239
+ 1970 train 4.224911 (lr=7.6319e-04) (hash(x)=94690431)
240
+ 1980 train 4.061162 (lr=7.5978e-04) (hash(x)=88153756)
241
+ 1990 train 4.108451 (lr=7.5636e-04) (hash(x)=77195688)
242
+ 2000 val loss 4.2426
243
+ 2000 val perplexity 69.5871
244
+ 2000 train 4.205563 (lr=7.5292e-04) (hash(x)=81308591)
245
+ 2010 train 4.265396 (lr=7.4946e-04) (hash(x)=80205479)
246
+ 2020 train 4.175561 (lr=7.4599e-04) (hash(x)=81008704)
247
+ 2030 train 4.239628 (lr=7.4251e-04) (hash(x)=82914358)
248
+ 2040 train 4.252584 (lr=7.3901e-04) (hash(x)=88064399)
249
+ 2050 train 4.254251 (lr=7.3550e-04) (hash(x)=82889390)
250
+ 2060 train 4.321682 (lr=7.3197e-04) (hash(x)=80913554)
251
+ 2070 train 4.166076 (lr=7.2843e-04) (hash(x)=78121791)
252
+ 2080 train 4.301443 (lr=7.2488e-04) (hash(x)=75356657)
253
+ 2090 train 4.126113 (lr=7.2132e-04) (hash(x)=74048412)
254
+ 2100 val loss 4.2185
255
+ 2100 val perplexity 67.9346
256
+ 2100 train 4.039075 (lr=7.1774e-04) (hash(x)=68928225)
257
+ 2110 train 4.096006 (lr=7.1415e-04) (hash(x)=93734745)
258
+ 2120 train 4.095790 (lr=7.1054e-04) (hash(x)=72878164)
259
+ 2130 train 4.221679 (lr=7.0693e-04) (hash(x)=71508226)
260
+ 2140 train 4.183282 (lr=7.0330e-04) (hash(x)=84168671)
261
+ 2150 train 4.117144 (lr=6.9967e-04) (hash(x)=74673639)
262
+ 2160 train 4.267021 (lr=6.9602e-04) (hash(x)=75470031)
263
+ 2170 train 4.225623 (lr=6.9236e-04) (hash(x)=74307890)
264
+ 2180 train 4.332872 (lr=6.8869e-04) (hash(x)=77214245)
265
+ 2190 train 4.122440 (lr=6.8501e-04) (hash(x)=88628359)
266
+ 2200 val loss 4.2005
267
+ 2200 val perplexity 66.7225
268
+ 2200 train 4.137580 (lr=6.8132e-04) (hash(x)=74779126)
269
+ 2210 train 4.173217 (lr=6.7762e-04) (hash(x)=79817976)
270
+ 2220 train 4.127203 (lr=6.7391e-04) (hash(x)=75258996)
271
+ 2230 train 4.101178 (lr=6.7020e-04) (hash(x)=88164047)
272
+ 2240 train 4.080373 (lr=6.6647e-04) (hash(x)=78580686)
273
+ 2250 train 4.023821 (lr=6.6273e-04) (hash(x)=76510617)
274
+ 2260 train 4.099797 (lr=6.5899e-04) (hash(x)=76043966)
275
+ 2270 train 4.074233 (lr=6.5524e-04) (hash(x)=73062098)
276
+ 2280 train 4.301982 (lr=6.5148e-04) (hash(x)=81885909)
277
+ 2290 train 4.236540 (lr=6.4771e-04) (hash(x)=81481741)
278
+ 2300 val loss 4.1776
279
+ 2300 val perplexity 65.2069
280
+ 2300 train 4.099236 (lr=6.4394e-04) (hash(x)=82104275)
281
+ 2310 train 4.146192 (lr=6.4016e-04) (hash(x)=82137309)
282
+ 2320 train 4.122744 (lr=6.3637e-04) (hash(x)=86741167)
283
+ 2330 train 4.139983 (lr=6.3258e-04) (hash(x)=84796263)
284
+ 2340 train 4.116372 (lr=6.2878e-04) (hash(x)=169966529)
285
+ 2350 train 3.984068 (lr=6.2497e-04) (hash(x)=80006061)
286
+ 2360 train 4.041545 (lr=6.2116e-04) (hash(x)=71090999)
287
+ 2370 train 4.176558 (lr=6.1734e-04) (hash(x)=86641850)
288
+ 2380 train 4.075324 (lr=6.1352e-04) (hash(x)=84573253)
289
+ 2390 train 4.109580 (lr=6.0969e-04) (hash(x)=74483764)
290
+ 2400 val loss 4.1720
291
+ 2400 val perplexity 64.8457
292
+ 2400 train 4.082648 (lr=6.0586e-04) (hash(x)=78327659)
293
+ 2410 train 4.147253 (lr=6.0203e-04) (hash(x)=77273627)
294
+ 2420 train 4.007768 (lr=5.9819e-04) (hash(x)=76938049)
295
+ 2430 train 4.017433 (lr=5.9435e-04) (hash(x)=98449442)
296
+ 2440 train 3.871452 (lr=5.9050e-04) (hash(x)=78157797)
297
+ 2450 train 3.914529 (lr=5.8665e-04) (hash(x)=80637582)
298
+ 2460 train 3.901491 (lr=5.8280e-04) (hash(x)=80023854)
299
+ 2470 train 3.963451 (lr=5.7895e-04) (hash(x)=86678884)
300
+ 2480 train 4.183783 (lr=5.7509e-04) (hash(x)=84086469)
301
+ 2490 train 4.208554 (lr=5.7124e-04) (hash(x)=83131141)
302
+ 2500 val loss 4.1439
303
+ 2500 val perplexity 63.0511
304
+ 2500 train 4.121066 (lr=5.6738e-04) (hash(x)=82583497)
305
+ 2510 train 4.184245 (lr=5.6352e-04) (hash(x)=79164326)
306
+ 2520 train 4.093334 (lr=5.5966e-04) (hash(x)=76456503)
307
+ 2530 train 4.148188 (lr=5.5579e-04) (hash(x)=82001228)
308
+ 2540 train 4.185853 (lr=5.5193e-04) (hash(x)=75615595)
309
+ 2550 train 4.130745 (lr=5.4807e-04) (hash(x)=78397869)
310
+ 2560 train 4.133675 (lr=5.4421e-04) (hash(x)=75547032)
311
+ 2570 train 4.049031 (lr=5.4034e-04) (hash(x)=89201025)
312
+ 2580 train 4.033206 (lr=5.3648e-04) (hash(x)=89856704)
313
+ 2590 train 4.039921 (lr=5.3262e-04) (hash(x)=82175682)
314
+ 2600 val loss 4.1279
315
+ 2600 val perplexity 62.0475
316
+ 2600 train 4.048053 (lr=5.2876e-04) (hash(x)=78312826)
317
+ 2610 train 4.039161 (lr=5.2491e-04) (hash(x)=77066588)
318
+ 2620 train 3.884663 (lr=5.2105e-04) (hash(x)=78666061)
319
+ 2630 train 3.881917 (lr=5.1720e-04) (hash(x)=93762143)
320
+ 2640 train 3.895318 (lr=5.1335e-04) (hash(x)=83191587)
321
+ 2650 train 3.821294 (lr=5.0950e-04) (hash(x)=87169585)
322
+ 2660 train 4.035018 (lr=5.0565e-04) (hash(x)=86426388)
323
+ 2670 train 4.169761 (lr=5.0181e-04) (hash(x)=76692638)
324
+ 2680 train 4.087070 (lr=4.9797e-04) (hash(x)=77446063)
325
+ 2690 train 4.090132 (lr=4.9414e-04) (hash(x)=79809050)
326
+ 2700 val loss 4.1195
327
+ 2700 val perplexity 61.5311
328
+ 2700 train 4.016455 (lr=4.9031e-04) (hash(x)=83116823)
329
+ 2710 train 4.130111 (lr=4.8648e-04) (hash(x)=75622148)
330
+ 2720 train 4.057587 (lr=4.8266e-04) (hash(x)=80690512)
331
+ 2730 train 4.128715 (lr=4.7884e-04) (hash(x)=78009984)
332
+ 2740 train 4.083000 (lr=4.7503e-04) (hash(x)=74205488)
333
+ 2750 train 4.085611 (lr=4.7122e-04) (hash(x)=91013332)
334
+ 2760 train 4.010113 (lr=4.6742e-04) (hash(x)=76954961)
335
+ 2770 train 4.016261 (lr=4.6363e-04) (hash(x)=79390317)
336
+ 2780 train 3.881595 (lr=4.5984e-04) (hash(x)=70168783)
337
+ 2790 train 4.043994 (lr=4.5606e-04) (hash(x)=76028417)
338
+ 2800 val loss 4.1017
339
+ 2800 val perplexity 60.4402
340
+ 2800 train 3.876253 (lr=4.5229e-04) (hash(x)=77656050)
341
+ 2810 train 3.922241 (lr=4.4852e-04) (hash(x)=94248216)
342
+ 2820 train 3.983012 (lr=4.4476e-04) (hash(x)=78305078)
343
+ 2830 train 3.898071 (lr=4.4101e-04) (hash(x)=79948848)
344
+ 2840 train 3.768102 (lr=4.3727e-04) (hash(x)=85341024)
345
+ 2850 train 4.068344 (lr=4.3353e-04) (hash(x)=78735170)
346
+ 2860 train 4.319623 (lr=4.2980e-04) (hash(x)=71616419)
347
+ 2870 train 4.095048 (lr=4.2609e-04) (hash(x)=78656517)
348
+ 2880 train 4.101239 (lr=4.2238e-04) (hash(x)=80073987)
349
+ 2890 train 4.034958 (lr=4.1868e-04) (hash(x)=76894809)
350
+ 2900 val loss 4.0780
351
+ 2900 val perplexity 59.0267
352
+ 2900 train 4.012119 (lr=4.1499e-04) (hash(x)=80499838)
353
+ 2910 train 4.060806 (lr=4.1131e-04) (hash(x)=72673354)
354
+ 2920 train 4.047205 (lr=4.0764e-04) (hash(x)=84265768)
355
+ 2930 train 4.078131 (lr=4.0398e-04) (hash(x)=79612060)
356
+ 2940 train 3.930761 (lr=4.0033e-04) (hash(x)=74970087)
357
+ 2950 train 4.084149 (lr=3.9670e-04) (hash(x)=84166818)
358
+ 2960 train 4.071276 (lr=3.9307e-04) (hash(x)=89410221)
359
+ 2970 train 4.012252 (lr=3.8946e-04) (hash(x)=75672566)
360
+ 2980 train 3.867584 (lr=3.8585e-04) (hash(x)=81760314)
361
+ 2990 train 3.845665 (lr=3.8226e-04) (hash(x)=80605200)
362
+ 3000 val loss 4.0774
363
+ 3000 val perplexity 58.9894
364
+ 3000 train 3.840361 (lr=3.7868e-04) (hash(x)=83804735)
365
+ 3010 train 3.767772 (lr=3.7512e-04) (hash(x)=77015303)
366
+ 3020 train 3.837901 (lr=3.7157e-04) (hash(x)=81464523)
367
+ 3030 train 4.030494 (lr=3.6803e-04) (hash(x)=80168230)
368
+ 3040 train 3.973297 (lr=3.6450e-04) (hash(x)=84817006)
369
+ 3050 train 4.085903 (lr=3.6099e-04) (hash(x)=71601811)
370
+ 3060 train 4.005211 (lr=3.5749e-04) (hash(x)=85499733)
371
+ 3070 train 4.059131 (lr=3.5401e-04) (hash(x)=78670408)
372
+ 3080 train 3.955578 (lr=3.5054e-04) (hash(x)=77120468)
373
+ 3090 train 4.063004 (lr=3.4708e-04) (hash(x)=77927426)
374
+ 3100 val loss 4.0582
375
+ 3100 val perplexity 57.8691
376
+ 3100 train 4.069612 (lr=3.4364e-04) (hash(x)=83998606)
377
+ 3110 train 3.977296 (lr=3.4022e-04) (hash(x)=77990218)
378
+ 3120 train 3.884751 (lr=3.3681e-04) (hash(x)=81623970)
379
+ 3130 train 3.860155 (lr=3.3341e-04) (hash(x)=75014781)
380
+ 3140 train 3.916252 (lr=3.3004e-04) (hash(x)=72591250)
381
+ 3150 train 3.927227 (lr=3.2667e-04) (hash(x)=81421847)
382
+ 3160 train 3.974535 (lr=3.2333e-04) (hash(x)=75112631)
383
+ 3170 train 4.079255 (lr=3.2000e-04) (hash(x)=87518033)
384
+ 3180 train 3.947504 (lr=3.1669e-04) (hash(x)=78648348)
385
+ 3190 train 4.044508 (lr=3.1340e-04) (hash(x)=78270029)
386
+ 3200 val loss 4.0401
387
+ 3200 val perplexity 56.8310
388
+ 3200 train 4.115208 (lr=3.1012e-04) (hash(x)=83380714)
389
+ 3210 train 4.291146 (lr=3.0686e-04) (hash(x)=67804991)
390
+ 3220 train 4.024065 (lr=3.0362e-04) (hash(x)=76968804)
391
+ 3230 train 3.957221 (lr=3.0039e-04) (hash(x)=75798670)
392
+ 3240 train 4.036116 (lr=2.9719e-04) (hash(x)=83192811)
393
+ 3250 train 3.970036 (lr=2.9400e-04) (hash(x)=81659789)
394
+ 3260 train 4.013213 (lr=2.9084e-04) (hash(x)=80622502)
395
+ 3270 train 3.983583 (lr=2.8769e-04) (hash(x)=85436511)
396
+ 3280 train 3.926018 (lr=2.8456e-04) (hash(x)=80589180)
397
+ 3290 train 4.067779 (lr=2.8145e-04) (hash(x)=85747193)
398
+ 3300 val loss 4.0368
399
+ 3300 val perplexity 56.6448
400
+ 3300 train 3.937358 (lr=2.7836e-04) (hash(x)=74635692)
401
+ 3310 train 3.924999 (lr=2.7529e-04) (hash(x)=82045455)
402
+ 3320 train 3.792733 (lr=2.7224e-04) (hash(x)=73249173)
403
+ 3330 train 3.756548 (lr=2.6921e-04) (hash(x)=71492338)
404
+ 3340 train 4.081250 (lr=2.6621e-04) (hash(x)=76374071)
405
+ 3350 train 4.057066 (lr=2.6322e-04) (hash(x)=80474064)
406
+ 3360 train 3.967772 (lr=2.6025e-04) (hash(x)=84390892)
407
+ 3370 train 4.012055 (lr=2.5731e-04) (hash(x)=83399949)
408
+ 3380 train 4.062373 (lr=2.5438e-04) (hash(x)=77648059)
409
+ 3390 train 4.032550 (lr=2.5148e-04) (hash(x)=80152701)
410
+ 3400 val loss 4.0184
411
+ 3400 val perplexity 55.6105
412
+ 3400 train 4.078629 (lr=2.4860e-04) (hash(x)=81952545)
413
+ 3410 train 4.042744 (lr=2.4575e-04) (hash(x)=80908993)
414
+ 3420 train 4.085543 (lr=2.4291e-04) (hash(x)=79610037)
415
+ 3430 train 3.963293 (lr=2.4010e-04) (hash(x)=87624382)
416
+ 3440 train 4.010305 (lr=2.3731e-04) (hash(x)=82336381)
417
+ 3450 train 3.981266 (lr=2.3454e-04) (hash(x)=87687835)
418
+ 3460 train 3.988847 (lr=2.3180e-04) (hash(x)=76376135)
419
+ 3470 train 3.899115 (lr=2.2908e-04) (hash(x)=79331391)
420
+ 3480 train 3.811670 (lr=2.2639e-04) (hash(x)=107002681)
421
+ 3490 train 3.887200 (lr=2.2371e-04) (hash(x)=77465514)
422
+ 3500 val loss 4.0140
423
+ 3500 val perplexity 55.3698
424
+ 3500 train 3.851650 (lr=2.2107e-04) (hash(x)=88237229)
425
+ 3510 train 3.922492 (lr=2.1844e-04) (hash(x)=81011739)
426
+ 3520 train 3.964980 (lr=2.1584e-04) (hash(x)=64643427)
427
+ 3530 train 4.243170 (lr=2.1327e-04) (hash(x)=78029539)
428
+ 3540 train 4.081161 (lr=2.1072e-04) (hash(x)=83188968)
429
+ 3550 train 4.098843 (lr=2.0819e-04) (hash(x)=86104185)
430
+ 3560 train 4.024222 (lr=2.0569e-04) (hash(x)=81993629)
431
+ 3570 train 3.992031 (lr=2.0322e-04) (hash(x)=72719368)
432
+ 3580 train 4.006433 (lr=2.0077e-04) (hash(x)=90091487)
433
+ 3590 train 3.892180 (lr=1.9835e-04) (hash(x)=77393152)
434
+ 3600 val loss 3.9980
435
+ 3600 val perplexity 54.4894
436
+ 3600 train 4.000335 (lr=1.9595e-04) (hash(x)=73103504)
437
+ 3610 train 4.007365 (lr=1.9358e-04) (hash(x)=84429400)
438
+ 3620 train 3.977944 (lr=1.9123e-04) (hash(x)=77021795)
439
+ 3630 train 3.933591 (lr=1.8892e-04) (hash(x)=92830605)
440
+ 3640 train 4.128996 (lr=1.8662e-04) (hash(x)=78313175)
441
+ 3650 train 3.930659 (lr=1.8436e-04) (hash(x)=85395549)
442
+ 3660 train 3.721490 (lr=1.8212e-04) (hash(x)=78114459)
443
+ 3670 train 3.848449 (lr=1.7991e-04) (hash(x)=74968316)
444
+ 3680 train 3.775409 (lr=1.7773e-04) (hash(x)=73358737)
445
+ 3690 train 3.884755 (lr=1.7557e-04) (hash(x)=76399442)
446
+ 3700 val loss 3.9999
447
+ 3700 val perplexity 54.5953
448
+ 3700 train 3.886904 (lr=1.7344e-04) (hash(x)=74039273)
449
+ 3710 train 4.058114 (lr=1.7134e-04) (hash(x)=79299680)
450
+ 3720 train 3.973749 (lr=1.6927e-04) (hash(x)=94701498)
451
+ 3730 train 4.057297 (lr=1.6722e-04) (hash(x)=75352071)
452
+ 3740 train 3.932108 (lr=1.6521e-04) (hash(x)=77909487)
453
+ 3750 train 4.035911 (lr=1.6322e-04) (hash(x)=73986730)
454
+ 3760 train 3.994704 (lr=1.6126e-04) (hash(x)=79325763)
455
+ 3770 train 4.044218 (lr=1.5933e-04) (hash(x)=72457818)
456
+ 3780 train 4.178916 (lr=1.5742e-04) (hash(x)=71775590)
457
+ 3790 train 4.066329 (lr=1.5555e-04) (hash(x)=82638943)
458
+ 3800 val loss 3.9841
459
+ 3800 val perplexity 53.7372
460
+ 3800 train 4.034616 (lr=1.5371e-04) (hash(x)=79965893)
461
+ 3810 train 3.972143 (lr=1.5189e-04) (hash(x)=72598235)
462
+ 3820 train 3.953705 (lr=1.5010e-04) (hash(x)=83113889)
463
+ 3830 train 4.016311 (lr=1.4835e-04) (hash(x)=74434590)
464
+ 3840 train 3.930218 (lr=1.4662e-04) (hash(x)=82860348)
465
+ 3850 train 3.870552 (lr=1.4492e-04) (hash(x)=78067565)
466
+ 3860 train 3.940069 (lr=1.4326e-04) (hash(x)=82592498)
467
+ 3870 train 3.873112 (lr=1.4162e-04) (hash(x)=81820733)
468
+ 3880 train 3.887015 (lr=1.4001e-04) (hash(x)=87709040)
469
+ 3890 train 3.929998 (lr=1.3843e-04) (hash(x)=70379093)
470
+ 3900 val loss 3.9786
471
+ 3900 val perplexity 53.4418
472
+ 3900 train 4.053595 (lr=1.3689e-04) (hash(x)=76597431)
473
+ 3910 train 3.939058 (lr=1.3537e-04) (hash(x)=90490716)
474
+ 3920 train 4.041865 (lr=1.3389e-04) (hash(x)=81970659)
475
+ 3930 train 4.025040 (lr=1.3243e-04) (hash(x)=81496334)
476
+ 3940 train 3.998601 (lr=1.3101e-04) (hash(x)=75717605)
477
+ 3950 train 3.948491 (lr=1.2961e-04) (hash(x)=82749357)
478
+ 3960 train 3.984629 (lr=1.2825e-04) (hash(x)=89355157)
479
+ 3970 train 3.935885 (lr=1.2692e-04) (hash(x)=78980403)
480
+ 3980 train 3.908915 (lr=1.2562e-04) (hash(x)=76627217)
481
+ 3990 train 3.910161 (lr=1.2435e-04) (hash(x)=72412879)
482
+ 4000 val loss 3.9736
483
+ 4000 val perplexity 53.1755
484
+ 4000 train 3.922434 (lr=1.2311e-04) (hash(x)=83018142)
485
+ 4010 train 3.937491 (lr=1.2191e-04) (hash(x)=81272436)
486
+ 4020 train 3.870353 (lr=1.2073e-04) (hash(x)=85497482)
487
+ 4030 train 3.942448 (lr=1.1959e-04) (hash(x)=85241734)
488
+ 4040 train 3.908227 (lr=1.1848e-04) (hash(x)=77925307)
489
+ 4050 train 3.904905 (lr=1.1740e-04) (hash(x)=84826179)
490
+ 4060 train 3.966163 (lr=1.1635e-04) (hash(x)=83606764)
491
+ 4070 train 3.881898 (lr=1.1533e-04) (hash(x)=80567590)
492
+ 4080 train 3.927043 (lr=1.1435e-04) (hash(x)=76860998)
493
+ 4090 train 3.967396 (lr=1.1340e-04) (hash(x)=74902328)
494
+ 4100 val loss 3.9636
495
+ 4100 val perplexity 52.6446
496
+ 4100 train 4.009918 (lr=1.1248e-04) (hash(x)=82832041)
497
+ 4110 train 3.894547 (lr=1.1159e-04) (hash(x)=79143262)
498
+ 4120 train 3.912313 (lr=1.1074e-04) (hash(x)=77038149)
499
+ 4130 train 4.098038 (lr=1.0991e-04) (hash(x)=86339074)
500
+ 4140 train 4.167376 (lr=1.0912e-04) (hash(x)=76686216)
501
+ 4150 train 3.941953 (lr=1.0837e-04) (hash(x)=70522682)
502
+ 4160 train 3.906858 (lr=1.0764e-04) (hash(x)=90958555)
503
+ 4170 train 3.932593 (lr=1.0695e-04) (hash(x)=91463532)
504
+ 4180 train 3.919674 (lr=1.0629e-04) (hash(x)=81959329)
505
+ 4190 train 3.874177 (lr=1.0566e-04) (hash(x)=83146752)
506
+ 4200 val loss 3.9668
507
+ 4200 val perplexity 52.8153
508
+ 4200 train 3.861100 (lr=1.0507e-04) (hash(x)=78361715)
509
+ 4210 train 4.004243 (lr=1.0451e-04) (hash(x)=87364889)
510
+ 4220 train 3.859195 (lr=1.0398e-04) (hash(x)=70465156)
511
+ 4230 train 4.010368 (lr=1.0348e-04) (hash(x)=84524081)
512
+ 4240 train 3.967828 (lr=1.0302e-04) (hash(x)=77824868)
513
+ 4250 train 3.981533 (lr=1.0259e-04) (hash(x)=81710711)
514
+ 4260 train 3.962546 (lr=1.0219e-04) (hash(x)=76362728)
515
+ 4270 train 3.951904 (lr=1.0183e-04) (hash(x)=83115208)
516
+ 4280 train 3.865821 (lr=1.0150e-04) (hash(x)=87218314)
517
+ 4290 train 4.033731 (lr=1.0120e-04) (hash(x)=74582673)
518
+ 4300 val loss 3.9612
519
+ 4300 val perplexity 52.5183
520
+ 4300 train 3.903133 (lr=1.0093e-04) (hash(x)=77379615)
521
+ 4310 train 3.876266 (lr=1.0070e-04) (hash(x)=78669579)
522
+ 4320 train 3.994651 (lr=1.0050e-04) (hash(x)=83066608)
523
+ 4330 train 3.960003 (lr=1.0034e-04) (hash(x)=83037340)
524
+ 4340 train 3.999604 (lr=1.0020e-04) (hash(x)=82849771)
525
+ 4350 train 3.933570 (lr=1.0010e-04) (hash(x)=76693985)
526
+ 4360 train 4.019527 (lr=1.0004e-04) (hash(x)=77745394)
527
+ 4370 train 3.984678 (lr=1.0000e-04) (hash(x)=79954388)
528
+ 4374 val loss 3.9551
529
+ 4374 val perplexity 52.2027
12_head_baseline_lr_10e-4_head_dim_22_fixed/model_04374.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:77699fd587770f71b71e8be8fee88933260e1e0e22f1055af86d8e7d6be2aa4f
3
+ size 96858242
12_head_baseline_lr_10e-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:ea2d108b32a81a5c6ef2c9fdc9ea41998c2bfb16acf856a18f59b4604248f0b5
3
+ size 187435910