andrew-healey commited on
Commit
fb67e73
·
verified ·
1 Parent(s): 5dd274d

Upload folder using huggingface_hub

Browse files
lr16e-4_total_batch_size61440_baseline_seed1339/args.json ADDED
@@ -0,0 +1 @@
 
 
1
+ {"hellaswag": true, "attention_kind": "selective", "log_dir": "wider_is_better_11/lr16e-4_total_batch_size61440_baseline_seed1339", "resume_checkpoint": null, "resume_optimizer": false, "add_a_head": false, "add_head_to_start": true, "new_head_init": "normal", "n_heads": 4, "protect_bos_token": true, "prevent_from_masking_myself": true, "max_steps": 8750, "warmup_steps": 500, "group": "wider_is_better_11", "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": 120, "total_batch_size": 61440, "protection_head_scaling_factor": 1.0, "protection_head_bias": 0.0, "n_sliced_masks": null, "n_latent_masks": null, "mask_layernorm": false, "residual_attention_masks": false, "compute_base_shapes": false, "base_shapes_savefile": null, "mup": true, "disable_selection": false, "mup_enable_coord_check_logging": false, "max_lr": 0.0016, "decay_lr": true, "readout_zero_init": false, "query_zero_init": false, "l1_loss": false, "debugpy": false, "key": "16e-4_61440", "n_embd": 256}
lr16e-4_total_batch_size61440_baseline_seed1339/dataloader_08749.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:953385078aa3787b69fc6857dfd48b0a2cd2f4d27c6f8892e01211aca53d07f5
3
+ size 964
lr16e-4_total_batch_size61440_baseline_seed1339/log2.txt ADDED
@@ -0,0 +1,1054 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ max_steps: 8750
2
+ 0 val loss 11.2592
3
+ 0 val perplexity 77590.4141
4
+ 0 train 11.272774 (lr=3.2000e-06) (hash(x)=150724848)
5
+ 10 train 10.025131 (lr=3.5200e-05) (hash(x)=149663361)
6
+ 20 train 9.585683 (lr=6.7200e-05) (hash(x)=153483789)
7
+ 30 train 9.031727 (lr=9.9200e-05) (hash(x)=145733176)
8
+ 40 train 8.260723 (lr=1.3120e-04) (hash(x)=145287265)
9
+ 50 train 7.810789 (lr=1.6320e-04) (hash(x)=149217061)
10
+ 60 train 7.627074 (lr=1.9520e-04) (hash(x)=150157569)
11
+ 70 train 7.564350 (lr=2.2720e-04) (hash(x)=140292841)
12
+ 80 train 7.589521 (lr=2.5920e-04) (hash(x)=152157598)
13
+ 90 train 7.370678 (lr=2.9120e-04) (hash(x)=149756968)
14
+ 100 val loss 7.3177
15
+ 100 val perplexity 1506.7437
16
+ 100 train 7.273450 (lr=3.2320e-04) (hash(x)=149910534)
17
+ 110 train 7.170372 (lr=3.5520e-04) (hash(x)=159020328)
18
+ 120 train 7.115211 (lr=3.8720e-04) (hash(x)=150731273)
19
+ 130 train 6.934540 (lr=4.1920e-04) (hash(x)=156727355)
20
+ 140 train 6.951688 (lr=4.5120e-04) (hash(x)=166473281)
21
+ 150 train 6.660701 (lr=4.8320e-04) (hash(x)=142185643)
22
+ 160 train 6.654724 (lr=5.1520e-04) (hash(x)=151371145)
23
+ 170 train 6.730270 (lr=5.4720e-04) (hash(x)=151601659)
24
+ 180 train 6.543693 (lr=5.7920e-04) (hash(x)=153368917)
25
+ 190 train 6.446299 (lr=6.1120e-04) (hash(x)=145008196)
26
+ 200 val loss 6.5278
27
+ 200 val perplexity 683.8601
28
+ 200 train 6.411638 (lr=6.4320e-04) (hash(x)=148123706)
29
+ 210 train 6.313298 (lr=6.7520e-04) (hash(x)=145714762)
30
+ 220 train 6.317086 (lr=7.0720e-04) (hash(x)=150622583)
31
+ 230 train 6.074991 (lr=7.3920e-04) (hash(x)=146512648)
32
+ 240 train 6.135986 (lr=7.7120e-04) (hash(x)=151535384)
33
+ 250 train 6.073927 (lr=8.0320e-04) (hash(x)=149952383)
34
+ 260 train 5.980754 (lr=8.3520e-04) (hash(x)=148726088)
35
+ 270 train 5.923471 (lr=8.6720e-04) (hash(x)=152776695)
36
+ 280 train 5.953705 (lr=8.9920e-04) (hash(x)=157260375)
37
+ 290 train 5.997379 (lr=9.3120e-04) (hash(x)=151559106)
38
+ 300 val loss 6.0856
39
+ 300 val perplexity 439.4800
40
+ 300 train 5.790976 (lr=9.6320e-04) (hash(x)=146678221)
41
+ 310 train 5.913116 (lr=9.9520e-04) (hash(x)=156713527)
42
+ 320 train 5.950741 (lr=1.0272e-03) (hash(x)=144947971)
43
+ 330 train 6.127790 (lr=1.0592e-03) (hash(x)=158684944)
44
+ 340 train 6.001911 (lr=1.0912e-03) (hash(x)=162318756)
45
+ 350 train 5.798350 (lr=1.1232e-03) (hash(x)=156180736)
46
+ 360 train 5.963525 (lr=1.1552e-03) (hash(x)=178373797)
47
+ 370 train 5.880543 (lr=1.1872e-03) (hash(x)=154389449)
48
+ 380 train 5.719584 (lr=1.2192e-03) (hash(x)=150369080)
49
+ 390 train 5.867765 (lr=1.2512e-03) (hash(x)=159840654)
50
+ 400 val loss 5.7576
51
+ 400 val perplexity 316.5767
52
+ 400 train 5.684437 (lr=1.2832e-03) (hash(x)=151700982)
53
+ 410 train 5.632832 (lr=1.3152e-03) (hash(x)=160467274)
54
+ 420 train 5.555092 (lr=1.3472e-03) (hash(x)=148930522)
55
+ 430 train 5.776380 (lr=1.3792e-03) (hash(x)=154565303)
56
+ 440 train 5.775686 (lr=1.4112e-03) (hash(x)=148079227)
57
+ 450 train 5.477000 (lr=1.4432e-03) (hash(x)=147288467)
58
+ 460 train 5.531157 (lr=1.4752e-03) (hash(x)=145272617)
59
+ 470 train 5.558499 (lr=1.5072e-03) (hash(x)=123126729)
60
+ 480 train 5.472197 (lr=1.5392e-03) (hash(x)=156287490)
61
+ 490 train 5.605177 (lr=1.5712e-03) (hash(x)=161700145)
62
+ 500 val loss 5.5186
63
+ 500 val perplexity 249.2834
64
+ 500 train 5.635873 (lr=1.6000e-03) (hash(x)=156182087)
65
+ 510 train 5.244022 (lr=1.6000e-03) (hash(x)=131674504)
66
+ 520 train 5.570564 (lr=1.6000e-03) (hash(x)=162758254)
67
+ 530 train 5.366342 (lr=1.6000e-03) (hash(x)=133510409)
68
+ 540 train 5.354335 (lr=1.5999e-03) (hash(x)=157363914)
69
+ 550 train 5.298860 (lr=1.5999e-03) (hash(x)=147861550)
70
+ 560 train 5.233920 (lr=1.5998e-03) (hash(x)=149449280)
71
+ 570 train 5.251096 (lr=1.5997e-03) (hash(x)=143899904)
72
+ 580 train 5.214754 (lr=1.5997e-03) (hash(x)=149266312)
73
+ 590 train 5.475574 (lr=1.5996e-03) (hash(x)=164352774)
74
+ 600 val loss 5.3670
75
+ 600 val perplexity 214.2096
76
+ 600 train 5.220910 (lr=1.5995e-03) (hash(x)=149318660)
77
+ 610 train 5.094209 (lr=1.5994e-03) (hash(x)=146101781)
78
+ 620 train 5.077780 (lr=1.5992e-03) (hash(x)=137184623)
79
+ 630 train 4.959744 (lr=1.5991e-03) (hash(x)=145791873)
80
+ 640 train 4.981156 (lr=1.5990e-03) (hash(x)=143999212)
81
+ 650 train 5.079996 (lr=1.5988e-03) (hash(x)=151631103)
82
+ 660 train 4.841329 (lr=1.5987e-03) (hash(x)=138360108)
83
+ 670 train 5.241291 (lr=1.5985e-03) (hash(x)=156637679)
84
+ 680 train 4.882954 (lr=1.5983e-03) (hash(x)=143752601)
85
+ 690 train 5.022505 (lr=1.5981e-03) (hash(x)=158368903)
86
+ 700 val loss 5.2393
87
+ 700 val perplexity 188.5335
88
+ 700 train 5.219721 (lr=1.5979e-03) (hash(x)=150482428)
89
+ 710 train 5.175689 (lr=1.5977e-03) (hash(x)=158085060)
90
+ 720 train 5.257870 (lr=1.5975e-03) (hash(x)=159461069)
91
+ 730 train 5.096364 (lr=1.5972e-03) (hash(x)=148795562)
92
+ 740 train 5.107410 (lr=1.5970e-03) (hash(x)=149967984)
93
+ 750 train 5.095672 (lr=1.5967e-03) (hash(x)=144458617)
94
+ 760 train 5.115508 (lr=1.5965e-03) (hash(x)=155106202)
95
+ 770 train 4.907781 (lr=1.5962e-03) (hash(x)=144506191)
96
+ 780 train 5.048966 (lr=1.5959e-03) (hash(x)=155825158)
97
+ 790 train 5.099349 (lr=1.5956e-03) (hash(x)=158835381)
98
+ 800 val loss 5.0581
99
+ 800 val perplexity 157.2875
100
+ 800 train 5.035565 (lr=1.5953e-03) (hash(x)=143268605)
101
+ 810 train 4.949813 (lr=1.5950e-03) (hash(x)=139072755)
102
+ 820 train 5.055183 (lr=1.5947e-03) (hash(x)=179384792)
103
+ 830 train 4.859488 (lr=1.5943e-03) (hash(x)=151086962)
104
+ 840 train 5.471696 (lr=1.5940e-03) (hash(x)=142711393)
105
+ 850 train 4.947808 (lr=1.5936e-03) (hash(x)=150817011)
106
+ 860 train 4.892491 (lr=1.5932e-03) (hash(x)=138495801)
107
+ 870 train 4.891448 (lr=1.5929e-03) (hash(x)=155474016)
108
+ 880 train 4.915212 (lr=1.5925e-03) (hash(x)=151951501)
109
+ 890 train 4.779245 (lr=1.5921e-03) (hash(x)=143699524)
110
+ 900 val loss 4.9319
111
+ 900 val perplexity 138.6464
112
+ 900 train 4.915530 (lr=1.5917e-03) (hash(x)=152322423)
113
+ 910 train 4.870879 (lr=1.5912e-03) (hash(x)=151010509)
114
+ 920 train 4.677186 (lr=1.5908e-03) (hash(x)=148720342)
115
+ 930 train 4.646551 (lr=1.5904e-03) (hash(x)=125869480)
116
+ 940 train 4.870290 (lr=1.5899e-03) (hash(x)=150349503)
117
+ 950 train 4.638201 (lr=1.5895e-03) (hash(x)=143574126)
118
+ 960 train 4.847713 (lr=1.5890e-03) (hash(x)=146292562)
119
+ 970 train 4.783146 (lr=1.5885e-03) (hash(x)=162256596)
120
+ 980 train 4.745920 (lr=1.5880e-03) (hash(x)=137640259)
121
+ 990 train 4.602564 (lr=1.5875e-03) (hash(x)=141767794)
122
+ 1000 val loss 4.8883
123
+ 1000 val perplexity 132.7329
124
+ 1000 train 4.836267 (lr=1.5870e-03) (hash(x)=147904298)
125
+ 1010 train 4.752292 (lr=1.5865e-03) (hash(x)=162239186)
126
+ 1020 train 4.532690 (lr=1.5859e-03) (hash(x)=140706537)
127
+ 1030 train 4.661152 (lr=1.5854e-03) (hash(x)=155556720)
128
+ 1040 train 4.547434 (lr=1.5848e-03) (hash(x)=144746445)
129
+ 1050 train 4.589172 (lr=1.5843e-03) (hash(x)=144719400)
130
+ 1060 train 4.466649 (lr=1.5837e-03) (hash(x)=153065665)
131
+ 1070 train 4.570581 (lr=1.5831e-03) (hash(x)=136468722)
132
+ 1080 train 4.567636 (lr=1.5825e-03) (hash(x)=157674756)
133
+ 1090 train 4.773908 (lr=1.5819e-03) (hash(x)=155882884)
134
+ 1100 val loss 4.8150
135
+ 1100 val perplexity 123.3526
136
+ 1100 train 5.016587 (lr=1.5813e-03) (hash(x)=154343147)
137
+ 1110 train 4.834621 (lr=1.5807e-03) (hash(x)=158582939)
138
+ 1120 train 4.631239 (lr=1.5800e-03) (hash(x)=162479784)
139
+ 1130 train 4.825483 (lr=1.5794e-03) (hash(x)=155704345)
140
+ 1140 train 4.717195 (lr=1.5787e-03) (hash(x)=146934299)
141
+ 1150 train 4.647765 (lr=1.5781e-03) (hash(x)=154372635)
142
+ 1160 train 4.662500 (lr=1.5774e-03) (hash(x)=145536832)
143
+ 1170 train 4.947711 (lr=1.5767e-03) (hash(x)=158785901)
144
+ 1180 train 4.741705 (lr=1.5760e-03) (hash(x)=158271614)
145
+ 1190 train 4.785951 (lr=1.5753e-03) (hash(x)=157365589)
146
+ 1200 val loss 4.7723
147
+ 1200 val perplexity 118.1870
148
+ 1200 train 4.831985 (lr=1.5746e-03) (hash(x)=141843115)
149
+ 1210 train 4.709829 (lr=1.5738e-03) (hash(x)=146807973)
150
+ 1220 train 4.697108 (lr=1.5731e-03) (hash(x)=144149947)
151
+ 1230 train 4.628326 (lr=1.5724e-03) (hash(x)=147869063)
152
+ 1240 train 4.725892 (lr=1.5716e-03) (hash(x)=152217389)
153
+ 1250 train 4.713710 (lr=1.5708e-03) (hash(x)=149417679)
154
+ 1260 train 4.747297 (lr=1.5701e-03) (hash(x)=144029489)
155
+ 1270 train 4.877210 (lr=1.5693e-03) (hash(x)=154328303)
156
+ 1280 train 4.733044 (lr=1.5685e-03) (hash(x)=151625838)
157
+ 1290 train 4.591934 (lr=1.5677e-03) (hash(x)=140270541)
158
+ 1300 val loss 4.7242
159
+ 1300 val perplexity 112.6376
160
+ 1300 train 4.649772 (lr=1.5668e-03) (hash(x)=145279030)
161
+ 1310 train 4.681239 (lr=1.5660e-03) (hash(x)=140279630)
162
+ 1320 train 4.406396 (lr=1.5652e-03) (hash(x)=133642994)
163
+ 1330 train 4.570854 (lr=1.5643e-03) (hash(x)=146634776)
164
+ 1340 train 4.667644 (lr=1.5635e-03) (hash(x)=155919973)
165
+ 1350 train 4.698894 (lr=1.5626e-03) (hash(x)=146363911)
166
+ 1360 train 4.581388 (lr=1.5617e-03) (hash(x)=150613636)
167
+ 1370 train 4.578727 (lr=1.5608e-03) (hash(x)=148536643)
168
+ 1380 train 4.585372 (lr=1.5600e-03) (hash(x)=152803906)
169
+ 1390 train 4.517068 (lr=1.5590e-03) (hash(x)=150383305)
170
+ 1400 val loss 4.7034
171
+ 1400 val perplexity 110.3265
172
+ 1400 train 4.611593 (lr=1.5581e-03) (hash(x)=152507639)
173
+ 1410 train 4.369914 (lr=1.5572e-03) (hash(x)=161841408)
174
+ 1420 train 4.477463 (lr=1.5563e-03) (hash(x)=167824828)
175
+ 1430 train 4.436874 (lr=1.5553e-03) (hash(x)=147696374)
176
+ 1440 train 4.399590 (lr=1.5544e-03) (hash(x)=150085494)
177
+ 1450 train 4.358628 (lr=1.5534e-03) (hash(x)=139296191)
178
+ 1460 train 4.277649 (lr=1.5524e-03) (hash(x)=142956546)
179
+ 1470 train 4.334372 (lr=1.5514e-03) (hash(x)=155908628)
180
+ 1480 train 4.346934 (lr=1.5504e-03) (hash(x)=150893345)
181
+ 1490 train 4.353087 (lr=1.5494e-03) (hash(x)=143548902)
182
+ 1500 val loss 4.6530
183
+ 1500 val perplexity 104.9005
184
+ 1500 train 4.607267 (lr=1.5484e-03) (hash(x)=148473774)
185
+ 1510 train 4.696157 (lr=1.5474e-03) (hash(x)=161631000)
186
+ 1520 train 4.619637 (lr=1.5464e-03) (hash(x)=166217855)
187
+ 1530 train 4.582170 (lr=1.5453e-03) (hash(x)=149440046)
188
+ 1540 train 4.531698 (lr=1.5443e-03) (hash(x)=155013278)
189
+ 1550 train 4.559169 (lr=1.5432e-03) (hash(x)=154130527)
190
+ 1560 train 4.584047 (lr=1.5421e-03) (hash(x)=152138735)
191
+ 1570 train 4.486287 (lr=1.5411e-03) (hash(x)=148128305)
192
+ 1580 train 4.621866 (lr=1.5400e-03) (hash(x)=157176546)
193
+ 1590 train 4.738033 (lr=1.5389e-03) (hash(x)=152935467)
194
+ 1600 val loss 4.5835
195
+ 1600 val perplexity 97.8593
196
+ 1600 train 4.561612 (lr=1.5378e-03) (hash(x)=151117002)
197
+ 1610 train 4.560181 (lr=1.5366e-03) (hash(x)=149883384)
198
+ 1620 train 4.600738 (lr=1.5355e-03) (hash(x)=145015354)
199
+ 1630 train 4.530306 (lr=1.5344e-03) (hash(x)=146717645)
200
+ 1640 train 4.732592 (lr=1.5332e-03) (hash(x)=156372727)
201
+ 1650 train 4.653121 (lr=1.5321e-03) (hash(x)=140261555)
202
+ 1660 train 4.479080 (lr=1.5309e-03) (hash(x)=150088650)
203
+ 1670 train 4.409728 (lr=1.5297e-03) (hash(x)=135943513)
204
+ 1680 train 4.632160 (lr=1.5285e-03) (hash(x)=153529653)
205
+ 1690 train 4.685378 (lr=1.5273e-03) (hash(x)=148930360)
206
+ 1700 val loss 4.5981
207
+ 1700 val perplexity 99.2958
208
+ 1700 train 4.567412 (lr=1.5261e-03) (hash(x)=138011335)
209
+ 1710 train 4.525795 (lr=1.5249e-03) (hash(x)=141383599)
210
+ 1720 train 4.717627 (lr=1.5237e-03) (hash(x)=161449652)
211
+ 1730 train 4.502111 (lr=1.5225e-03) (hash(x)=153923561)
212
+ 1740 train 4.440037 (lr=1.5212e-03) (hash(x)=146043253)
213
+ 1750 train 4.460851 (lr=1.5200e-03) (hash(x)=155720883)
214
+ 1760 train 4.458522 (lr=1.5187e-03) (hash(x)=148222148)
215
+ 1770 train 4.505610 (lr=1.5174e-03) (hash(x)=155712380)
216
+ 1780 train 4.403464 (lr=1.5162e-03) (hash(x)=149767859)
217
+ 1790 train 4.645420 (lr=1.5149e-03) (hash(x)=167214230)
218
+ 1800 val loss 4.5576
219
+ 1800 val perplexity 95.3506
220
+ 1800 train 4.531645 (lr=1.5136e-03) (hash(x)=171180926)
221
+ 1810 train 4.557944 (lr=1.5123e-03) (hash(x)=162375834)
222
+ 1820 train 4.673844 (lr=1.5109e-03) (hash(x)=151495752)
223
+ 1830 train 4.573869 (lr=1.5096e-03) (hash(x)=146532257)
224
+ 1840 train 4.650194 (lr=1.5083e-03) (hash(x)=145694434)
225
+ 1850 train 4.543767 (lr=1.5069e-03) (hash(x)=144823488)
226
+ 1860 train 4.641824 (lr=1.5056e-03) (hash(x)=149836286)
227
+ 1870 train 4.613108 (lr=1.5042e-03) (hash(x)=143935049)
228
+ 1880 train 4.649473 (lr=1.5029e-03) (hash(x)=151786371)
229
+ 1890 train 4.531791 (lr=1.5015e-03) (hash(x)=152892567)
230
+ 1900 val loss 4.5230
231
+ 1900 val perplexity 92.1087
232
+ 1900 train 4.560646 (lr=1.5001e-03) (hash(x)=141769419)
233
+ 1910 train 4.592002 (lr=1.4987e-03) (hash(x)=156914416)
234
+ 1920 train 4.442892 (lr=1.4973e-03) (hash(x)=151573900)
235
+ 1930 train 4.538916 (lr=1.4959e-03) (hash(x)=145121701)
236
+ 1940 train 4.477026 (lr=1.4944e-03) (hash(x)=150896348)
237
+ 1950 train 4.580748 (lr=1.4930e-03) (hash(x)=168797189)
238
+ 1960 train 4.455036 (lr=1.4916e-03) (hash(x)=146711430)
239
+ 1970 train 4.317023 (lr=1.4901e-03) (hash(x)=152180946)
240
+ 1980 train 4.444354 (lr=1.4886e-03) (hash(x)=155542268)
241
+ 1990 train 4.411927 (lr=1.4872e-03) (hash(x)=144668660)
242
+ 2000 val loss 4.4857
243
+ 2000 val perplexity 88.7403
244
+ 2000 train 4.472634 (lr=1.4857e-03) (hash(x)=151963443)
245
+ 2010 train 4.463285 (lr=1.4842e-03) (hash(x)=150810352)
246
+ 2020 train 4.481060 (lr=1.4827e-03) (hash(x)=157539832)
247
+ 2030 train 4.692491 (lr=1.4812e-03) (hash(x)=137858786)
248
+ 2040 train 4.436897 (lr=1.4797e-03) (hash(x)=154270900)
249
+ 2050 train 4.294101 (lr=1.4782e-03) (hash(x)=156252579)
250
+ 2060 train 4.224921 (lr=1.4767e-03) (hash(x)=143285549)
251
+ 2070 train 4.439602 (lr=1.4751e-03) (hash(x)=150542438)
252
+ 2080 train 4.343332 (lr=1.4736e-03) (hash(x)=146290330)
253
+ 2090 train 4.442492 (lr=1.4720e-03) (hash(x)=146486756)
254
+ 2100 val loss 4.5209
255
+ 2100 val perplexity 91.9223
256
+ 2100 train 4.456211 (lr=1.4704e-03) (hash(x)=162947470)
257
+ 2110 train 4.363920 (lr=1.4689e-03) (hash(x)=150305391)
258
+ 2120 train 4.513997 (lr=1.4673e-03) (hash(x)=156452783)
259
+ 2130 train 4.482985 (lr=1.4657e-03) (hash(x)=167481875)
260
+ 2140 train 4.470562 (lr=1.4641e-03) (hash(x)=141782256)
261
+ 2150 train 4.493862 (lr=1.4625e-03) (hash(x)=145035419)
262
+ 2160 train 4.545048 (lr=1.4609e-03) (hash(x)=142000400)
263
+ 2170 train 4.495989 (lr=1.4593e-03) (hash(x)=144229380)
264
+ 2180 train 4.549059 (lr=1.4576e-03) (hash(x)=166204176)
265
+ 2190 train 4.473619 (lr=1.4560e-03) (hash(x)=150712057)
266
+ 2200 val loss 4.4486
267
+ 2200 val perplexity 85.5039
268
+ 2200 train 4.352681 (lr=1.4543e-03) (hash(x)=154954810)
269
+ 2210 train 4.567746 (lr=1.4527e-03) (hash(x)=155668760)
270
+ 2220 train 4.402070 (lr=1.4510e-03) (hash(x)=144834708)
271
+ 2230 train 4.377214 (lr=1.4493e-03) (hash(x)=159744566)
272
+ 2240 train 4.352819 (lr=1.4476e-03) (hash(x)=158543377)
273
+ 2250 train 4.356544 (lr=1.4460e-03) (hash(x)=151004281)
274
+ 2260 train 4.468034 (lr=1.4443e-03) (hash(x)=151242482)
275
+ 2270 train 4.415412 (lr=1.4426e-03) (hash(x)=152056559)
276
+ 2280 train 4.352321 (lr=1.4408e-03) (hash(x)=148864076)
277
+ 2290 train 4.468650 (lr=1.4391e-03) (hash(x)=144294586)
278
+ 2300 val loss 4.4439
279
+ 2300 val perplexity 85.1048
280
+ 2300 train 4.376667 (lr=1.4374e-03) (hash(x)=151878111)
281
+ 2310 train 4.384327 (lr=1.4356e-03) (hash(x)=154139457)
282
+ 2320 train 4.321778 (lr=1.4339e-03) (hash(x)=155571592)
283
+ 2330 train 4.360383 (lr=1.4321e-03) (hash(x)=154823764)
284
+ 2340 train 4.415033 (lr=1.4304e-03) (hash(x)=153252177)
285
+ 2350 train 4.353195 (lr=1.4286e-03) (hash(x)=136051830)
286
+ 2360 train 4.392235 (lr=1.4268e-03) (hash(x)=153814175)
287
+ 2370 train 4.393646 (lr=1.4250e-03) (hash(x)=146391078)
288
+ 2380 train 4.364965 (lr=1.4232e-03) (hash(x)=150782698)
289
+ 2390 train 4.430039 (lr=1.4214e-03) (hash(x)=142752256)
290
+ 2400 val loss 4.4618
291
+ 2400 val perplexity 86.6414
292
+ 2400 train 4.344165 (lr=1.4196e-03) (hash(x)=158661057)
293
+ 2410 train 4.409606 (lr=1.4178e-03) (hash(x)=153611303)
294
+ 2420 train 4.583669 (lr=1.4160e-03) (hash(x)=159508684)
295
+ 2430 train 4.375042 (lr=1.4141e-03) (hash(x)=149501192)
296
+ 2440 train 4.052652 (lr=1.4123e-03) (hash(x)=181177177)
297
+ 2450 train 4.502018 (lr=1.4105e-03) (hash(x)=163664732)
298
+ 2460 train 4.185614 (lr=1.4086e-03) (hash(x)=194527955)
299
+ 2470 train 4.431728 (lr=1.4067e-03) (hash(x)=158088521)
300
+ 2480 train 4.381139 (lr=1.4049e-03) (hash(x)=152865678)
301
+ 2490 train 4.367587 (lr=1.4030e-03) (hash(x)=165703871)
302
+ 2500 val loss 4.4001
303
+ 2500 val perplexity 81.4582
304
+ 2500 train 4.447060 (lr=1.4011e-03) (hash(x)=150925584)
305
+ 2510 train 4.351169 (lr=1.3992e-03) (hash(x)=144406177)
306
+ 2520 train 4.256191 (lr=1.3973e-03) (hash(x)=120133043)
307
+ 2530 train 4.446594 (lr=1.3954e-03) (hash(x)=167900247)
308
+ 2540 train 4.338045 (lr=1.3935e-03) (hash(x)=162599203)
309
+ 2550 train 4.407025 (lr=1.3915e-03) (hash(x)=149328180)
310
+ 2560 train 4.288121 (lr=1.3896e-03) (hash(x)=150049968)
311
+ 2570 train 4.442749 (lr=1.3877e-03) (hash(x)=162876159)
312
+ 2580 train 4.249069 (lr=1.3857e-03) (hash(x)=148260649)
313
+ 2590 train 4.401249 (lr=1.3838e-03) (hash(x)=148998018)
314
+ 2600 val loss 4.3918
315
+ 2600 val perplexity 80.7822
316
+ 2600 train 4.311572 (lr=1.3818e-03) (hash(x)=144515755)
317
+ 2610 train 4.424881 (lr=1.3798e-03) (hash(x)=154455224)
318
+ 2620 train 4.224558 (lr=1.3778e-03) (hash(x)=142920048)
319
+ 2630 train 4.383066 (lr=1.3759e-03) (hash(x)=143931236)
320
+ 2640 train 4.243068 (lr=1.3739e-03) (hash(x)=150983323)
321
+ 2650 train 4.337639 (lr=1.3719e-03) (hash(x)=139360405)
322
+ 2660 train 4.254861 (lr=1.3699e-03) (hash(x)=148850470)
323
+ 2670 train 4.510337 (lr=1.3679e-03) (hash(x)=149929606)
324
+ 2680 train 4.480824 (lr=1.3658e-03) (hash(x)=151673000)
325
+ 2690 train 4.354672 (lr=1.3638e-03) (hash(x)=140631266)
326
+ 2700 val loss 4.3888
327
+ 2700 val perplexity 80.5476
328
+ 2700 train 4.319871 (lr=1.3618e-03) (hash(x)=153109144)
329
+ 2710 train 4.209406 (lr=1.3597e-03) (hash(x)=135332836)
330
+ 2720 train 4.571719 (lr=1.3577e-03) (hash(x)=131611182)
331
+ 2730 train 4.403709 (lr=1.3556e-03) (hash(x)=142909388)
332
+ 2740 train 4.443554 (lr=1.3536e-03) (hash(x)=145593419)
333
+ 2750 train 4.327682 (lr=1.3515e-03) (hash(x)=158225769)
334
+ 2760 train 4.295289 (lr=1.3494e-03) (hash(x)=170509938)
335
+ 2770 train 4.309628 (lr=1.3473e-03) (hash(x)=150785805)
336
+ 2780 train 4.260721 (lr=1.3453e-03) (hash(x)=167281736)
337
+ 2790 train 4.437510 (lr=1.3432e-03) (hash(x)=161660187)
338
+ 2800 val loss 4.3609
339
+ 2800 val perplexity 78.3259
340
+ 2800 train 4.397021 (lr=1.3411e-03) (hash(x)=151152897)
341
+ 2810 train 4.366120 (lr=1.3389e-03) (hash(x)=143172067)
342
+ 2820 train 4.373993 (lr=1.3368e-03) (hash(x)=141569169)
343
+ 2830 train 4.253506 (lr=1.3347e-03) (hash(x)=159709681)
344
+ 2840 train 4.307917 (lr=1.3326e-03) (hash(x)=162552294)
345
+ 2850 train 4.443490 (lr=1.3304e-03) (hash(x)=146762690)
346
+ 2860 train 4.313388 (lr=1.3283e-03) (hash(x)=145729422)
347
+ 2870 train 4.389859 (lr=1.3262e-03) (hash(x)=144360628)
348
+ 2880 train 4.196325 (lr=1.3240e-03) (hash(x)=147737098)
349
+ 2890 train 4.309228 (lr=1.3218e-03) (hash(x)=146849057)
350
+ 2900 val loss 4.3465
351
+ 2900 val perplexity 77.2098
352
+ 2900 train 4.375220 (lr=1.3197e-03) (hash(x)=145800210)
353
+ 2910 train 4.177730 (lr=1.3175e-03) (hash(x)=147316313)
354
+ 2920 train 4.228498 (lr=1.3153e-03) (hash(x)=152595415)
355
+ 2930 train 4.383525 (lr=1.3131e-03) (hash(x)=156542647)
356
+ 2940 train 4.245067 (lr=1.3109e-03) (hash(x)=141183896)
357
+ 2950 train 4.344598 (lr=1.3087e-03) (hash(x)=156329036)
358
+ 2960 train 4.392339 (lr=1.3065e-03) (hash(x)=156717793)
359
+ 2970 train 4.247432 (lr=1.3043e-03) (hash(x)=143949931)
360
+ 2980 train 4.312513 (lr=1.3021e-03) (hash(x)=155067290)
361
+ 2990 train 4.283980 (lr=1.2999e-03) (hash(x)=148774729)
362
+ 3000 val loss 4.3668
363
+ 3000 val perplexity 78.7895
364
+ 3000 train 4.272683 (lr=1.2976e-03) (hash(x)=141997485)
365
+ 3010 train 4.244236 (lr=1.2954e-03) (hash(x)=160762953)
366
+ 3020 train 4.191498 (lr=1.2932e-03) (hash(x)=170341665)
367
+ 3030 train 4.466419 (lr=1.2909e-03) (hash(x)=150722836)
368
+ 3040 train 4.258226 (lr=1.2887e-03) (hash(x)=154243307)
369
+ 3050 train 4.318624 (lr=1.2864e-03) (hash(x)=146077234)
370
+ 3060 train 4.271032 (lr=1.2841e-03) (hash(x)=152125542)
371
+ 3070 train 4.448186 (lr=1.2819e-03) (hash(x)=144149387)
372
+ 3080 train 4.267434 (lr=1.2796e-03) (hash(x)=147508712)
373
+ 3090 train 4.267958 (lr=1.2773e-03) (hash(x)=143583369)
374
+ 3100 val loss 4.3211
375
+ 3100 val perplexity 75.2730
376
+ 3100 train 4.293116 (lr=1.2750e-03) (hash(x)=154049740)
377
+ 3110 train 4.357312 (lr=1.2727e-03) (hash(x)=147054835)
378
+ 3120 train 4.333012 (lr=1.2704e-03) (hash(x)=152468568)
379
+ 3130 train 4.294221 (lr=1.2681e-03) (hash(x)=151635194)
380
+ 3140 train 4.254244 (lr=1.2658e-03) (hash(x)=150396274)
381
+ 3150 train 4.299094 (lr=1.2635e-03) (hash(x)=141306267)
382
+ 3160 train 4.253588 (lr=1.2612e-03) (hash(x)=151110762)
383
+ 3170 train 4.332781 (lr=1.2588e-03) (hash(x)=146518875)
384
+ 3180 train 4.215192 (lr=1.2565e-03) (hash(x)=139914401)
385
+ 3190 train 4.214882 (lr=1.2542e-03) (hash(x)=161631880)
386
+ 3200 val loss 4.3384
387
+ 3200 val perplexity 76.5814
388
+ 3200 train 4.248756 (lr=1.2518e-03) (hash(x)=150471842)
389
+ 3210 train 4.315307 (lr=1.2495e-03) (hash(x)=147355138)
390
+ 3220 train 4.118887 (lr=1.2471e-03) (hash(x)=138879678)
391
+ 3230 train 4.211522 (lr=1.2447e-03) (hash(x)=123929679)
392
+ 3240 train 4.272280 (lr=1.2424e-03) (hash(x)=157263734)
393
+ 3250 train 4.283725 (lr=1.2400e-03) (hash(x)=148065539)
394
+ 3260 train 4.212765 (lr=1.2376e-03) (hash(x)=138020742)
395
+ 3270 train 4.076176 (lr=1.2352e-03) (hash(x)=142053611)
396
+ 3280 train 4.230538 (lr=1.2329e-03) (hash(x)=155444784)
397
+ 3290 train 4.068761 (lr=1.2305e-03) (hash(x)=148500315)
398
+ 3300 val loss 4.3257
399
+ 3300 val perplexity 75.6204
400
+ 3300 train 4.285111 (lr=1.2281e-03) (hash(x)=149048126)
401
+ 3310 train 4.241220 (lr=1.2257e-03) (hash(x)=141680674)
402
+ 3320 train 4.178150 (lr=1.2233e-03) (hash(x)=143310642)
403
+ 3330 train 4.295571 (lr=1.2208e-03) (hash(x)=149891496)
404
+ 3340 train 4.252459 (lr=1.2184e-03) (hash(x)=151244609)
405
+ 3350 train 4.255222 (lr=1.2160e-03) (hash(x)=157158479)
406
+ 3360 train 4.119131 (lr=1.2136e-03) (hash(x)=151709876)
407
+ 3370 train 4.266103 (lr=1.2111e-03) (hash(x)=145297090)
408
+ 3380 train 4.279053 (lr=1.2087e-03) (hash(x)=154767881)
409
+ 3390 train 4.256643 (lr=1.2063e-03) (hash(x)=152600366)
410
+ 3400 val loss 4.3019
411
+ 3400 val perplexity 73.8435
412
+ 3400 train 4.494072 (lr=1.2038e-03) (hash(x)=161261339)
413
+ 3410 train 4.249516 (lr=1.2014e-03) (hash(x)=146859162)
414
+ 3420 train 4.333557 (lr=1.1989e-03) (hash(x)=149276224)
415
+ 3430 train 4.355667 (lr=1.1964e-03) (hash(x)=155866933)
416
+ 3440 train 4.222234 (lr=1.1940e-03) (hash(x)=143191904)
417
+ 3450 train 4.435290 (lr=1.1915e-03) (hash(x)=156910661)
418
+ 3460 train 4.227069 (lr=1.1890e-03) (hash(x)=140850879)
419
+ 3470 train 4.446159 (lr=1.1866e-03) (hash(x)=162140445)
420
+ 3480 train 4.188616 (lr=1.1841e-03) (hash(x)=145607495)
421
+ 3490 train 4.205326 (lr=1.1816e-03) (hash(x)=152496121)
422
+ 3500 val loss 4.2973
423
+ 3500 val perplexity 73.4984
424
+ 3500 train 4.175090 (lr=1.1791e-03) (hash(x)=157495564)
425
+ 3510 train 4.290747 (lr=1.1766e-03) (hash(x)=147657004)
426
+ 3520 train 4.266298 (lr=1.1741e-03) (hash(x)=152978060)
427
+ 3530 train 4.206063 (lr=1.1716e-03) (hash(x)=146800180)
428
+ 3540 train 4.287113 (lr=1.1691e-03) (hash(x)=195450444)
429
+ 3550 train 4.264196 (lr=1.1666e-03) (hash(x)=151114129)
430
+ 3560 train 4.244889 (lr=1.1641e-03) (hash(x)=143833190)
431
+ 3570 train 4.060926 (lr=1.1615e-03) (hash(x)=146039971)
432
+ 3580 train 4.111614 (lr=1.1590e-03) (hash(x)=145172249)
433
+ 3590 train 4.032078 (lr=1.1565e-03) (hash(x)=159164125)
434
+ 3600 val loss 4.2983
435
+ 3600 val perplexity 73.5740
436
+ 3600 train 4.280652 (lr=1.1539e-03) (hash(x)=144352932)
437
+ 3610 train 4.240660 (lr=1.1514e-03) (hash(x)=150550670)
438
+ 3620 train 4.363882 (lr=1.1489e-03) (hash(x)=147170125)
439
+ 3630 train 4.170075 (lr=1.1463e-03) (hash(x)=139473935)
440
+ 3640 train 4.157699 (lr=1.1438e-03) (hash(x)=159294424)
441
+ 3650 train 4.294442 (lr=1.1412e-03) (hash(x)=145943278)
442
+ 3660 train 4.337271 (lr=1.1387e-03) (hash(x)=141647872)
443
+ 3670 train 4.310481 (lr=1.1361e-03) (hash(x)=147109351)
444
+ 3680 train 4.346969 (lr=1.1335e-03) (hash(x)=140683368)
445
+ 3690 train 4.246592 (lr=1.1310e-03) (hash(x)=156276284)
446
+ 3700 val loss 4.2821
447
+ 3700 val perplexity 72.3911
448
+ 3700 train 4.191104 (lr=1.1284e-03) (hash(x)=149389012)
449
+ 3710 train 4.327221 (lr=1.1258e-03) (hash(x)=142107929)
450
+ 3720 train 4.298885 (lr=1.1232e-03) (hash(x)=147828843)
451
+ 3730 train 4.358304 (lr=1.1207e-03) (hash(x)=141784679)
452
+ 3740 train 4.266986 (lr=1.1181e-03) (hash(x)=153375464)
453
+ 3750 train 4.206771 (lr=1.1155e-03) (hash(x)=153716001)
454
+ 3760 train 4.163213 (lr=1.1129e-03) (hash(x)=146848822)
455
+ 3770 train 4.120928 (lr=1.1103e-03) (hash(x)=148396782)
456
+ 3780 train 4.195003 (lr=1.1077e-03) (hash(x)=153563244)
457
+ 3790 train 4.116428 (lr=1.1051e-03) (hash(x)=160808734)
458
+ 3800 val loss 4.2742
459
+ 3800 val perplexity 71.8228
460
+ 3800 train 4.301201 (lr=1.1025e-03) (hash(x)=146607620)
461
+ 3810 train 4.270831 (lr=1.0999e-03) (hash(x)=152580677)
462
+ 3820 train 4.622921 (lr=1.0973e-03) (hash(x)=144860879)
463
+ 3830 train 4.281878 (lr=1.0947e-03) (hash(x)=162099851)
464
+ 3840 train 4.238841 (lr=1.0920e-03) (hash(x)=159506855)
465
+ 3850 train 4.157506 (lr=1.0894e-03) (hash(x)=140022686)
466
+ 3860 train 4.237765 (lr=1.0868e-03) (hash(x)=152137850)
467
+ 3870 train 4.354002 (lr=1.0842e-03) (hash(x)=142666785)
468
+ 3880 train 4.213869 (lr=1.0815e-03) (hash(x)=148285868)
469
+ 3890 train 4.173796 (lr=1.0789e-03) (hash(x)=147910067)
470
+ 3900 val loss 4.2721
471
+ 3900 val perplexity 71.6692
472
+ 3900 train 4.214138 (lr=1.0763e-03) (hash(x)=143754617)
473
+ 3910 train 4.198431 (lr=1.0736e-03) (hash(x)=154779949)
474
+ 3920 train 4.048389 (lr=1.0710e-03) (hash(x)=137824061)
475
+ 3930 train 4.023436 (lr=1.0683e-03) (hash(x)=136933373)
476
+ 3940 train 4.087575 (lr=1.0657e-03) (hash(x)=164054337)
477
+ 3950 train 4.254358 (lr=1.0630e-03) (hash(x)=155752984)
478
+ 3960 train 4.399663 (lr=1.0604e-03) (hash(x)=160957673)
479
+ 3970 train 4.278500 (lr=1.0577e-03) (hash(x)=157158690)
480
+ 3980 train 4.281799 (lr=1.0551e-03) (hash(x)=153282027)
481
+ 3990 train 4.305571 (lr=1.0524e-03) (hash(x)=152183510)
482
+ 4000 val loss 4.2447
483
+ 4000 val perplexity 69.7374
484
+ 4000 train 4.196651 (lr=1.0497e-03) (hash(x)=156930722)
485
+ 4010 train 4.191834 (lr=1.0471e-03) (hash(x)=150798740)
486
+ 4020 train 4.390197 (lr=1.0444e-03) (hash(x)=142632400)
487
+ 4030 train 4.298959 (lr=1.0417e-03) (hash(x)=173326287)
488
+ 4040 train 4.207175 (lr=1.0391e-03) (hash(x)=128797443)
489
+ 4050 train 4.203255 (lr=1.0364e-03) (hash(x)=146937792)
490
+ 4060 train 4.307610 (lr=1.0337e-03) (hash(x)=164579030)
491
+ 4070 train 4.167141 (lr=1.0310e-03) (hash(x)=149459820)
492
+ 4080 train 4.134647 (lr=1.0284e-03) (hash(x)=160159595)
493
+ 4090 train 4.073075 (lr=1.0257e-03) (hash(x)=141507853)
494
+ 4100 val loss 4.2436
495
+ 4100 val perplexity 69.6616
496
+ 4100 train 4.212739 (lr=1.0230e-03) (hash(x)=147500519)
497
+ 4110 train 4.188548 (lr=1.0203e-03) (hash(x)=153297081)
498
+ 4120 train 4.197169 (lr=1.0176e-03) (hash(x)=142560288)
499
+ 4130 train 4.216734 (lr=1.0149e-03) (hash(x)=153380118)
500
+ 4140 train 4.122244 (lr=1.0122e-03) (hash(x)=143150720)
501
+ 4150 train 4.158964 (lr=1.0095e-03) (hash(x)=143459564)
502
+ 4160 train 4.193592 (lr=1.0068e-03) (hash(x)=153955741)
503
+ 4170 train 4.300733 (lr=1.0041e-03) (hash(x)=152199253)
504
+ 4180 train 4.063235 (lr=1.0014e-03) (hash(x)=165957668)
505
+ 4190 train 4.109220 (lr=9.9872e-04) (hash(x)=144866923)
506
+ 4200 val loss 4.2544
507
+ 4200 val perplexity 70.4147
508
+ 4200 train 4.125908 (lr=9.9602e-04) (hash(x)=143232237)
509
+ 4210 train 4.036789 (lr=9.9331e-04) (hash(x)=151907163)
510
+ 4220 train 4.144682 (lr=9.9060e-04) (hash(x)=145917587)
511
+ 4230 train 3.977664 (lr=9.8789e-04) (hash(x)=147258846)
512
+ 4240 train 4.126507 (lr=9.8518e-04) (hash(x)=143449843)
513
+ 4250 train 4.086202 (lr=9.8247e-04) (hash(x)=184883715)
514
+ 4260 train 4.188468 (lr=9.7975e-04) (hash(x)=156958761)
515
+ 4270 train 4.251468 (lr=9.7704e-04) (hash(x)=168870544)
516
+ 4280 train 4.186996 (lr=9.7432e-04) (hash(x)=179446735)
517
+ 4290 train 4.222748 (lr=9.7160e-04) (hash(x)=145649386)
518
+ 4300 val loss 4.2391
519
+ 4300 val perplexity 69.3469
520
+ 4300 train 4.073118 (lr=9.6888e-04) (hash(x)=146811670)
521
+ 4310 train 4.340371 (lr=9.6616e-04) (hash(x)=151907075)
522
+ 4320 train 4.142743 (lr=9.6344e-04) (hash(x)=138628365)
523
+ 4330 train 4.194609 (lr=9.6071e-04) (hash(x)=151608240)
524
+ 4340 train 4.136559 (lr=9.5799e-04) (hash(x)=160111319)
525
+ 4350 train 4.282754 (lr=9.5526e-04) (hash(x)=144958652)
526
+ 4360 train 4.074617 (lr=9.5253e-04) (hash(x)=149176222)
527
+ 4370 train 4.165335 (lr=9.4980e-04) (hash(x)=216221371)
528
+ 4380 train 4.069569 (lr=9.4708e-04) (hash(x)=141760648)
529
+ 4390 train 4.164903 (lr=9.4435e-04) (hash(x)=151710283)
530
+ 4400 val loss 4.2267
531
+ 4400 val perplexity 68.4919
532
+ 4400 train 4.281494 (lr=9.4161e-04) (hash(x)=158418746)
533
+ 4410 train 4.142350 (lr=9.3888e-04) (hash(x)=142331301)
534
+ 4420 train 4.221125 (lr=9.3615e-04) (hash(x)=146507180)
535
+ 4430 train 4.316234 (lr=9.3342e-04) (hash(x)=166506937)
536
+ 4440 train 4.255503 (lr=9.3068e-04) (hash(x)=151209447)
537
+ 4450 train 4.215953 (lr=9.2795e-04) (hash(x)=149186822)
538
+ 4460 train 4.198572 (lr=9.2521e-04) (hash(x)=144348412)
539
+ 4470 train 4.082734 (lr=9.2247e-04) (hash(x)=156494284)
540
+ 4480 train 4.115755 (lr=9.1974e-04) (hash(x)=147238744)
541
+ 4490 train 4.059837 (lr=9.1700e-04) (hash(x)=150746185)
542
+ 4500 val loss 4.2183
543
+ 4500 val perplexity 67.9155
544
+ 4500 train 4.150872 (lr=9.1426e-04) (hash(x)=156695778)
545
+ 4510 train 3.992307 (lr=9.1152e-04) (hash(x)=135988958)
546
+ 4520 train 4.166739 (lr=9.0878e-04) (hash(x)=148272110)
547
+ 4530 train 4.080204 (lr=9.0604e-04) (hash(x)=136459794)
548
+ 4540 train 4.117513 (lr=9.0330e-04) (hash(x)=147624954)
549
+ 4550 train 4.125357 (lr=9.0056e-04) (hash(x)=154947199)
550
+ 4560 train 4.178821 (lr=8.9782e-04) (hash(x)=147869166)
551
+ 4570 train 4.032541 (lr=8.9508e-04) (hash(x)=148415463)
552
+ 4580 train 4.194614 (lr=8.9234e-04) (hash(x)=152104066)
553
+ 4590 train 4.097000 (lr=8.8960e-04) (hash(x)=147060040)
554
+ 4600 val loss 4.2165
555
+ 4600 val perplexity 67.7980
556
+ 4600 train 4.167636 (lr=8.8685e-04) (hash(x)=147791497)
557
+ 4610 train 4.363392 (lr=8.8411e-04) (hash(x)=155595556)
558
+ 4620 train 4.209068 (lr=8.8137e-04) (hash(x)=147178858)
559
+ 4630 train 4.256698 (lr=8.7863e-04) (hash(x)=152617112)
560
+ 4640 train 4.264138 (lr=8.7589e-04) (hash(x)=161766869)
561
+ 4650 train 4.256812 (lr=8.7315e-04) (hash(x)=158075561)
562
+ 4660 train 4.284628 (lr=8.7040e-04) (hash(x)=145360002)
563
+ 4670 train 4.168013 (lr=8.6766e-04) (hash(x)=144538791)
564
+ 4680 train 4.205245 (lr=8.6492e-04) (hash(x)=142711273)
565
+ 4690 train 4.357350 (lr=8.6218e-04) (hash(x)=193907362)
566
+ 4700 val loss 4.1988
567
+ 4700 val perplexity 66.6086
568
+ 4700 train 4.174163 (lr=8.5944e-04) (hash(x)=155533088)
569
+ 4710 train 4.119929 (lr=8.5670e-04) (hash(x)=148618528)
570
+ 4720 train 4.230325 (lr=8.5396e-04) (hash(x)=150502624)
571
+ 4730 train 4.103704 (lr=8.5122e-04) (hash(x)=189205847)
572
+ 4740 train 4.161226 (lr=8.4848e-04) (hash(x)=167322800)
573
+ 4750 train 4.280221 (lr=8.4574e-04) (hash(x)=162169095)
574
+ 4760 train 3.989284 (lr=8.4300e-04) (hash(x)=143459469)
575
+ 4770 train 4.067940 (lr=8.4026e-04) (hash(x)=144386593)
576
+ 4780 train 4.024207 (lr=8.3753e-04) (hash(x)=143721184)
577
+ 4790 train 4.093266 (lr=8.3479e-04) (hash(x)=146077222)
578
+ 4800 val loss 4.2027
579
+ 4800 val perplexity 66.8676
580
+ 4800 train 4.071481 (lr=8.3205e-04) (hash(x)=138350044)
581
+ 4810 train 4.075161 (lr=8.2932e-04) (hash(x)=154960112)
582
+ 4820 train 3.989809 (lr=8.2658e-04) (hash(x)=146950685)
583
+ 4830 train 3.989542 (lr=8.2385e-04) (hash(x)=140224554)
584
+ 4840 train 3.986736 (lr=8.2112e-04) (hash(x)=156640495)
585
+ 4850 train 4.203962 (lr=8.1839e-04) (hash(x)=153409150)
586
+ 4860 train 4.121429 (lr=8.1565e-04) (hash(x)=151647700)
587
+ 4870 train 4.518312 (lr=8.1292e-04) (hash(x)=155465603)
588
+ 4880 train 4.363112 (lr=8.1020e-04) (hash(x)=153404400)
589
+ 4890 train 4.249812 (lr=8.0747e-04) (hash(x)=170409281)
590
+ 4900 val loss 4.1858
591
+ 4900 val perplexity 65.7439
592
+ 4900 train 4.103229 (lr=8.0474e-04) (hash(x)=143735284)
593
+ 4910 train 4.244550 (lr=8.0201e-04) (hash(x)=150784329)
594
+ 4920 train 4.129103 (lr=7.9929e-04) (hash(x)=152353574)
595
+ 4930 train 4.081770 (lr=7.9656e-04) (hash(x)=152677008)
596
+ 4940 train 4.148868 (lr=7.9384e-04) (hash(x)=140553861)
597
+ 4950 train 4.248311 (lr=7.9112e-04) (hash(x)=152869181)
598
+ 4960 train 4.199991 (lr=7.8840e-04) (hash(x)=153011079)
599
+ 4970 train 4.155344 (lr=7.8568e-04) (hash(x)=152796617)
600
+ 4980 train 4.262026 (lr=7.8296e-04) (hash(x)=146868417)
601
+ 4990 train 4.148071 (lr=7.8025e-04) (hash(x)=140168188)
602
+ 5000 val loss 4.1764
603
+ 5000 val perplexity 65.1294
604
+ 5000 train 4.122110 (lr=7.7753e-04) (hash(x)=154976463)
605
+ 5010 train 4.184629 (lr=7.7482e-04) (hash(x)=148582690)
606
+ 5020 train 4.196929 (lr=7.7211e-04) (hash(x)=166715464)
607
+ 5030 train 4.162450 (lr=7.6940e-04) (hash(x)=150616397)
608
+ 5040 train 4.206439 (lr=7.6669e-04) (hash(x)=150775548)
609
+ 5050 train 4.195756 (lr=7.6398e-04) (hash(x)=150575367)
610
+ 5060 train 4.112638 (lr=7.6128e-04) (hash(x)=152352082)
611
+ 5070 train 4.150111 (lr=7.5858e-04) (hash(x)=150269272)
612
+ 5080 train 4.211859 (lr=7.5587e-04) (hash(x)=146499361)
613
+ 5090 train 3.900180 (lr=7.5317e-04) (hash(x)=148734454)
614
+ 5100 val loss 4.1795
615
+ 5100 val perplexity 65.3314
616
+ 5100 train 4.080353 (lr=7.5048e-04) (hash(x)=149894982)
617
+ 5110 train 4.131424 (lr=7.4778e-04) (hash(x)=154487751)
618
+ 5120 train 4.084700 (lr=7.4509e-04) (hash(x)=144169582)
619
+ 5130 train 4.061710 (lr=7.4239e-04) (hash(x)=152217630)
620
+ 5140 train 4.077796 (lr=7.3970e-04) (hash(x)=158233831)
621
+ 5150 train 4.150780 (lr=7.3701e-04) (hash(x)=142746426)
622
+ 5160 train 4.134115 (lr=7.3433e-04) (hash(x)=139184458)
623
+ 5170 train 4.224742 (lr=7.3164e-04) (hash(x)=140624857)
624
+ 5180 train 3.945129 (lr=7.2896e-04) (hash(x)=154107423)
625
+ 5190 train 3.890058 (lr=7.2628e-04) (hash(x)=144892610)
626
+ 5200 val loss 4.1769
627
+ 5200 val perplexity 65.1606
628
+ 5200 train 3.831846 (lr=7.2361e-04) (hash(x)=159326689)
629
+ 5210 train 4.010603 (lr=7.2093e-04) (hash(x)=151186017)
630
+ 5220 train 3.807011 (lr=7.1826e-04) (hash(x)=148996819)
631
+ 5230 train 3.756491 (lr=7.1559e-04) (hash(x)=150829917)
632
+ 5240 train 3.881472 (lr=7.1292e-04) (hash(x)=152758972)
633
+ 5250 train 3.844984 (lr=7.1025e-04) (hash(x)=154345903)
634
+ 5260 train 3.837996 (lr=7.0759e-04) (hash(x)=150134947)
635
+ 5270 train 3.930423 (lr=7.0493e-04) (hash(x)=149689310)
636
+ 5280 train 4.178551 (lr=7.0227e-04) (hash(x)=147662223)
637
+ 5290 train 4.238178 (lr=6.9962e-04) (hash(x)=165770608)
638
+ 5300 val loss 4.1682
639
+ 5300 val perplexity 64.5985
640
+ 5300 train 4.215611 (lr=6.9696e-04) (hash(x)=159484800)
641
+ 5310 train 4.243243 (lr=6.9431e-04) (hash(x)=152672536)
642
+ 5320 train 4.071619 (lr=6.9166e-04) (hash(x)=152728812)
643
+ 5330 train 4.017704 (lr=6.8902e-04) (hash(x)=152778685)
644
+ 5340 train 4.117883 (lr=6.8638e-04) (hash(x)=149818849)
645
+ 5350 train 4.041136 (lr=6.8374e-04) (hash(x)=168239739)
646
+ 5360 train 4.187983 (lr=6.8110e-04) (hash(x)=139859423)
647
+ 5370 train 4.017087 (lr=6.7847e-04) (hash(x)=139315327)
648
+ 5380 train 4.226107 (lr=6.7584e-04) (hash(x)=163068509)
649
+ 5390 train 4.179071 (lr=6.7321e-04) (hash(x)=147645415)
650
+ 5400 val loss 4.1441
651
+ 5400 val perplexity 63.0620
652
+ 5400 train 4.135152 (lr=6.7059e-04) (hash(x)=140385615)
653
+ 5410 train 3.996602 (lr=6.6796e-04) (hash(x)=151381862)
654
+ 5420 train 4.138696 (lr=6.6534e-04) (hash(x)=152445092)
655
+ 5430 train 4.157798 (lr=6.6273e-04) (hash(x)=142454479)
656
+ 5440 train 4.108191 (lr=6.6012e-04) (hash(x)=151308771)
657
+ 5450 train 3.969954 (lr=6.5751e-04) (hash(x)=150886070)
658
+ 5460 train 4.069334 (lr=6.5490e-04) (hash(x)=146080408)
659
+ 5470 train 4.090832 (lr=6.5230e-04) (hash(x)=149828857)
660
+ 5480 train 4.070264 (lr=6.4970e-04) (hash(x)=139922034)
661
+ 5490 train 4.013256 (lr=6.4710e-04) (hash(x)=150262276)
662
+ 5500 val loss 4.1441
663
+ 5500 val perplexity 63.0584
664
+ 5500 train 3.982103 (lr=6.4451e-04) (hash(x)=148498335)
665
+ 5510 train 4.098257 (lr=6.4192e-04) (hash(x)=157036224)
666
+ 5520 train 3.985923 (lr=6.3934e-04) (hash(x)=134093409)
667
+ 5530 train 4.009608 (lr=6.3675e-04) (hash(x)=153126250)
668
+ 5540 train 4.017129 (lr=6.3417e-04) (hash(x)=136170985)
669
+ 5550 train 4.000783 (lr=6.3160e-04) (hash(x)=152841201)
670
+ 5560 train 4.103783 (lr=6.2903e-04) (hash(x)=147439014)
671
+ 5570 train 4.019072 (lr=6.2646e-04) (hash(x)=143476132)
672
+ 5580 train 3.790915 (lr=6.2390e-04) (hash(x)=139503437)
673
+ 5590 train 3.832694 (lr=6.2134e-04) (hash(x)=131664955)
674
+ 5600 val loss 4.1539
675
+ 5600 val perplexity 63.6831
676
+ 5600 train 3.928554 (lr=6.1878e-04) (hash(x)=151907614)
677
+ 5610 train 3.814938 (lr=6.1623e-04) (hash(x)=157132674)
678
+ 5620 train 4.016770 (lr=6.1368e-04) (hash(x)=146793775)
679
+ 5630 train 3.967438 (lr=6.1113e-04) (hash(x)=147237910)
680
+ 5640 train 3.876939 (lr=6.0859e-04) (hash(x)=141120024)
681
+ 5650 train 3.856092 (lr=6.0605e-04) (hash(x)=144275645)
682
+ 5660 train 3.886973 (lr=6.0352e-04) (hash(x)=159088002)
683
+ 5670 train 3.928365 (lr=6.0099e-04) (hash(x)=147224479)
684
+ 5680 train 4.286070 (lr=5.9846e-04) (hash(x)=147208743)
685
+ 5690 train 4.089945 (lr=5.9594e-04) (hash(x)=154384701)
686
+ 5700 val loss 4.1460
687
+ 5700 val perplexity 63.1782
688
+ 5700 train 4.124380 (lr=5.9342e-04) (hash(x)=155192267)
689
+ 5710 train 4.063839 (lr=5.9091e-04) (hash(x)=148458753)
690
+ 5720 train 4.177832 (lr=5.8840e-04) (hash(x)=159947204)
691
+ 5730 train 4.155349 (lr=5.8590e-04) (hash(x)=150746366)
692
+ 5740 train 4.151862 (lr=5.8340e-04) (hash(x)=155201884)
693
+ 5750 train 4.150684 (lr=5.8090e-04) (hash(x)=156432008)
694
+ 5760 train 4.135125 (lr=5.7841e-04) (hash(x)=185977746)
695
+ 5770 train 4.139095 (lr=5.7592e-04) (hash(x)=133620000)
696
+ 5780 train 4.140096 (lr=5.7344e-04) (hash(x)=153778044)
697
+ 5790 train 3.992345 (lr=5.7096e-04) (hash(x)=147713822)
698
+ 5800 val loss 4.1216
699
+ 5800 val perplexity 61.6607
700
+ 5800 train 4.140488 (lr=5.6849e-04) (hash(x)=153132158)
701
+ 5810 train 4.103373 (lr=5.6602e-04) (hash(x)=144293895)
702
+ 5820 train 4.075985 (lr=5.6355e-04) (hash(x)=151500590)
703
+ 5830 train 4.028651 (lr=5.6109e-04) (hash(x)=139308707)
704
+ 5840 train 4.139853 (lr=5.5864e-04) (hash(x)=153591278)
705
+ 5850 train 4.123959 (lr=5.5618e-04) (hash(x)=146077037)
706
+ 5860 train 4.085130 (lr=5.5374e-04) (hash(x)=149160733)
707
+ 5870 train 4.060245 (lr=5.5130e-04) (hash(x)=138482476)
708
+ 5880 train 4.082697 (lr=5.4886e-04) (hash(x)=152173137)
709
+ 5890 train 4.066964 (lr=5.4643e-04) (hash(x)=142387057)
710
+ 5900 val loss 4.1350
711
+ 5900 val perplexity 62.4895
712
+ 5900 train 4.139030 (lr=5.4400e-04) (hash(x)=161446764)
713
+ 5910 train 4.010157 (lr=5.4158e-04) (hash(x)=144512138)
714
+ 5920 train 4.067516 (lr=5.3916e-04) (hash(x)=152555618)
715
+ 5930 train 4.022644 (lr=5.3675e-04) (hash(x)=146129367)
716
+ 5940 train 3.967901 (lr=5.3434e-04) (hash(x)=154607527)
717
+ 5950 train 4.051786 (lr=5.3194e-04) (hash(x)=143827143)
718
+ 5960 train 4.182988 (lr=5.2954e-04) (hash(x)=151168510)
719
+ 5970 train 3.969777 (lr=5.2715e-04) (hash(x)=158961574)
720
+ 5980 train 3.807679 (lr=5.2476e-04) (hash(x)=148757787)
721
+ 5990 train 4.295764 (lr=5.2238e-04) (hash(x)=161719494)
722
+ 6000 val loss 4.1306
723
+ 6000 val perplexity 62.2129
724
+ 6000 train 3.799279 (lr=5.2000e-04) (hash(x)=151512446)
725
+ 6010 train 3.921787 (lr=5.1763e-04) (hash(x)=173500735)
726
+ 6020 train 4.055345 (lr=5.1526e-04) (hash(x)=157621031)
727
+ 6030 train 3.936944 (lr=5.1290e-04) (hash(x)=142662684)
728
+ 6040 train 3.931229 (lr=5.1054e-04) (hash(x)=146180940)
729
+ 6050 train 3.848628 (lr=5.0819e-04) (hash(x)=147654412)
730
+ 6060 train 3.720720 (lr=5.0585e-04) (hash(x)=156393157)
731
+ 6070 train 4.129524 (lr=5.0351e-04) (hash(x)=156947300)
732
+ 6080 train 4.119689 (lr=5.0117e-04) (hash(x)=163108626)
733
+ 6090 train 4.215795 (lr=4.9885e-04) (hash(x)=158654713)
734
+ 6100 val loss 4.1101
735
+ 6100 val perplexity 60.9507
736
+ 6100 train 3.950422 (lr=4.9652e-04) (hash(x)=188094053)
737
+ 6110 train 4.101242 (lr=4.9420e-04) (hash(x)=145132600)
738
+ 6120 train 3.979149 (lr=4.9189e-04) (hash(x)=146999707)
739
+ 6130 train 4.204317 (lr=4.8959e-04) (hash(x)=145719530)
740
+ 6140 train 4.138451 (lr=4.8729e-04) (hash(x)=153230722)
741
+ 6150 train 4.077563 (lr=4.8499e-04) (hash(x)=147455934)
742
+ 6160 train 4.125252 (lr=4.8270e-04) (hash(x)=146596242)
743
+ 6170 train 4.227705 (lr=4.8042e-04) (hash(x)=152029855)
744
+ 6180 train 4.082908 (lr=4.7814e-04) (hash(x)=151480034)
745
+ 6190 train 4.018599 (lr=4.7587e-04) (hash(x)=147381047)
746
+ 6200 val loss 4.0931
747
+ 6200 val perplexity 59.9239
748
+ 6200 train 4.005223 (lr=4.7360e-04) (hash(x)=149389789)
749
+ 6210 train 4.057927 (lr=4.7134e-04) (hash(x)=154894138)
750
+ 6220 train 4.091999 (lr=4.6909e-04) (hash(x)=150074302)
751
+ 6230 train 3.952763 (lr=4.6684e-04) (hash(x)=142881794)
752
+ 6240 train 3.945299 (lr=4.6460e-04) (hash(x)=164747461)
753
+ 6250 train 4.116146 (lr=4.6236e-04) (hash(x)=158378292)
754
+ 6260 train 4.119603 (lr=4.6013e-04) (hash(x)=147675938)
755
+ 6270 train 4.057930 (lr=4.5790e-04) (hash(x)=148197228)
756
+ 6280 train 4.017408 (lr=4.5569e-04) (hash(x)=135296747)
757
+ 6290 train 4.093952 (lr=4.5347e-04) (hash(x)=156439346)
758
+ 6300 val loss 4.0931
759
+ 6300 val perplexity 59.9279
760
+ 6300 train 3.915794 (lr=4.5127e-04) (hash(x)=138212820)
761
+ 6310 train 4.201530 (lr=4.4907e-04) (hash(x)=165985336)
762
+ 6320 train 4.074866 (lr=4.4688e-04) (hash(x)=145921897)
763
+ 6330 train 4.280838 (lr=4.4469e-04) (hash(x)=145502689)
764
+ 6340 train 4.019806 (lr=4.4251e-04) (hash(x)=141704937)
765
+ 6350 train 4.075778 (lr=4.4033e-04) (hash(x)=161789400)
766
+ 6360 train 3.807686 (lr=4.3817e-04) (hash(x)=145457780)
767
+ 6370 train 4.048119 (lr=4.3600e-04) (hash(x)=171628530)
768
+ 6380 train 3.840546 (lr=4.3385e-04) (hash(x)=153467923)
769
+ 6390 train 4.217614 (lr=4.3170e-04) (hash(x)=158066904)
770
+ 6400 val loss 4.0991
771
+ 6400 val perplexity 60.2846
772
+ 6400 train 3.864706 (lr=4.2956e-04) (hash(x)=146535423)
773
+ 6410 train 3.733438 (lr=4.2742e-04) (hash(x)=147264155)
774
+ 6420 train 3.834556 (lr=4.2529e-04) (hash(x)=145595485)
775
+ 6430 train 3.980077 (lr=4.2317e-04) (hash(x)=146929074)
776
+ 6440 train 3.913098 (lr=4.2105e-04) (hash(x)=147290878)
777
+ 6450 train 3.994871 (lr=4.1895e-04) (hash(x)=141154205)
778
+ 6460 train 3.843764 (lr=4.1684e-04) (hash(x)=154360901)
779
+ 6470 train 4.187998 (lr=4.1475e-04) (hash(x)=146350558)
780
+ 6480 train 4.143011 (lr=4.1266e-04) (hash(x)=153927683)
781
+ 6490 train 4.058379 (lr=4.1058e-04) (hash(x)=149337047)
782
+ 6500 val loss 4.0851
783
+ 6500 val perplexity 59.4472
784
+ 6500 train 3.986481 (lr=4.0850e-04) (hash(x)=145950843)
785
+ 6510 train 4.039566 (lr=4.0643e-04) (hash(x)=142577653)
786
+ 6520 train 4.104033 (lr=4.0437e-04) (hash(x)=165060689)
787
+ 6530 train 3.950519 (lr=4.0231e-04) (hash(x)=148367505)
788
+ 6540 train 4.043870 (lr=4.0027e-04) (hash(x)=154313651)
789
+ 6550 train 3.958898 (lr=3.9823e-04) (hash(x)=147477315)
790
+ 6560 train 4.050877 (lr=3.9619e-04) (hash(x)=139825403)
791
+ 6570 train 4.058730 (lr=3.9416e-04) (hash(x)=141482518)
792
+ 6580 train 4.103235 (lr=3.9214e-04) (hash(x)=144968144)
793
+ 6590 train 4.044857 (lr=3.9013e-04) (hash(x)=154319720)
794
+ 6600 val loss 4.0804
795
+ 6600 val perplexity 59.1689
796
+ 6600 train 4.093859 (lr=3.8813e-04) (hash(x)=141162902)
797
+ 6610 train 4.055049 (lr=3.8613e-04) (hash(x)=159490449)
798
+ 6620 train 3.937816 (lr=3.8414e-04) (hash(x)=151189772)
799
+ 6630 train 3.912310 (lr=3.8215e-04) (hash(x)=148451651)
800
+ 6640 train 4.040330 (lr=3.8017e-04) (hash(x)=152815550)
801
+ 6650 train 3.890241 (lr=3.7820e-04) (hash(x)=144316094)
802
+ 6660 train 4.224389 (lr=3.7624e-04) (hash(x)=143951240)
803
+ 6670 train 3.839010 (lr=3.7429e-04) (hash(x)=143203472)
804
+ 6680 train 3.880912 (lr=3.7234e-04) (hash(x)=132063806)
805
+ 6690 train 3.965489 (lr=3.7040e-04) (hash(x)=140739750)
806
+ 6700 val loss 4.0839
807
+ 6700 val perplexity 59.3756
808
+ 6700 train 4.027029 (lr=3.6847e-04) (hash(x)=153018737)
809
+ 6710 train 3.887612 (lr=3.6654e-04) (hash(x)=140107197)
810
+ 6720 train 3.995903 (lr=3.6462e-04) (hash(x)=144396559)
811
+ 6730 train 3.842317 (lr=3.6271e-04) (hash(x)=148038991)
812
+ 6740 train 4.058234 (lr=3.6081e-04) (hash(x)=145504746)
813
+ 6750 train 4.042772 (lr=3.5891e-04) (hash(x)=119579094)
814
+ 6760 train 4.056911 (lr=3.5702e-04) (hash(x)=161725430)
815
+ 6770 train 4.071033 (lr=3.5514e-04) (hash(x)=153646966)
816
+ 6780 train 4.086332 (lr=3.5327e-04) (hash(x)=153838463)
817
+ 6790 train 4.198997 (lr=3.5140e-04) (hash(x)=157891196)
818
+ 6800 val loss 4.0636
819
+ 6800 val perplexity 58.1836
820
+ 6800 train 3.918741 (lr=3.4955e-04) (hash(x)=155640155)
821
+ 6810 train 4.116039 (lr=3.4770e-04) (hash(x)=154913809)
822
+ 6820 train 4.028681 (lr=3.4585e-04) (hash(x)=159412493)
823
+ 6830 train 3.991987 (lr=3.4402e-04) (hash(x)=144756549)
824
+ 6840 train 4.200135 (lr=3.4219e-04) (hash(x)=143824638)
825
+ 6850 train 4.527844 (lr=3.4037e-04) (hash(x)=117180982)
826
+ 6860 train 4.055753 (lr=3.3856e-04) (hash(x)=141887150)
827
+ 6870 train 4.115954 (lr=3.3676e-04) (hash(x)=140094872)
828
+ 6880 train 3.953709 (lr=3.3496e-04) (hash(x)=141479953)
829
+ 6890 train 4.017599 (lr=3.3318e-04) (hash(x)=153166840)
830
+ 6900 val loss 4.0604
831
+ 6900 val perplexity 57.9970
832
+ 6900 train 4.043765 (lr=3.3140e-04) (hash(x)=153722115)
833
+ 6910 train 4.147164 (lr=3.2962e-04) (hash(x)=149578602)
834
+ 6920 train 4.004947 (lr=3.2786e-04) (hash(x)=149925108)
835
+ 6930 train 3.944329 (lr=3.2611e-04) (hash(x)=148589489)
836
+ 6940 train 3.855111 (lr=3.2436e-04) (hash(x)=144243794)
837
+ 6950 train 3.936490 (lr=3.2262e-04) (hash(x)=157743480)
838
+ 6960 train 3.924882 (lr=3.2089e-04) (hash(x)=139858247)
839
+ 6970 train 3.857079 (lr=3.1916e-04) (hash(x)=148727536)
840
+ 6980 train 3.970784 (lr=3.1745e-04) (hash(x)=153382595)
841
+ 6990 train 4.112020 (lr=3.1574e-04) (hash(x)=147529919)
842
+ 7000 val loss 4.0556
843
+ 7000 val perplexity 57.7201
844
+ 7000 train 4.074665 (lr=3.1404e-04) (hash(x)=146953450)
845
+ 7010 train 3.988949 (lr=3.1235e-04) (hash(x)=134561015)
846
+ 7020 train 4.097191 (lr=3.1067e-04) (hash(x)=148433515)
847
+ 7030 train 3.952348 (lr=3.0899e-04) (hash(x)=140394326)
848
+ 7040 train 3.957478 (lr=3.0733e-04) (hash(x)=136740763)
849
+ 7050 train 3.792797 (lr=3.0567e-04) (hash(x)=148637938)
850
+ 7060 train 3.962727 (lr=3.0402e-04) (hash(x)=147262147)
851
+ 7070 train 3.938805 (lr=3.0238e-04) (hash(x)=140191039)
852
+ 7080 train 3.945616 (lr=3.0075e-04) (hash(x)=148223305)
853
+ 7090 train 3.866401 (lr=2.9912e-04) (hash(x)=139984988)
854
+ 7100 val loss 4.0594
855
+ 7100 val perplexity 57.9406
856
+ 7100 train 3.979406 (lr=2.9751e-04) (hash(x)=137663885)
857
+ 7110 train 3.859655 (lr=2.9590e-04) (hash(x)=150955021)
858
+ 7120 train 3.775248 (lr=2.9430e-04) (hash(x)=148585682)
859
+ 7130 train 3.975757 (lr=2.9271e-04) (hash(x)=146055548)
860
+ 7140 train 4.011108 (lr=2.9113e-04) (hash(x)=155723541)
861
+ 7150 train 4.089570 (lr=2.8956e-04) (hash(x)=140465834)
862
+ 7160 train 4.061731 (lr=2.8799e-04) (hash(x)=163246922)
863
+ 7170 train 3.912742 (lr=2.8644e-04) (hash(x)=150754166)
864
+ 7180 train 4.458017 (lr=2.8489e-04) (hash(x)=142844151)
865
+ 7190 train 4.035649 (lr=2.8335e-04) (hash(x)=145167470)
866
+ 7200 val loss 4.0407
867
+ 7200 val perplexity 56.8663
868
+ 7200 train 4.227518 (lr=2.8182e-04) (hash(x)=146172950)
869
+ 7210 train 4.059411 (lr=2.8030e-04) (hash(x)=143155287)
870
+ 7220 train 4.010241 (lr=2.7878e-04) (hash(x)=151976669)
871
+ 7230 train 4.118203 (lr=2.7728e-04) (hash(x)=145979940)
872
+ 7240 train 4.091717 (lr=2.7578e-04) (hash(x)=155656034)
873
+ 7250 train 4.089636 (lr=2.7430e-04) (hash(x)=135610136)
874
+ 7260 train 4.169311 (lr=2.7282e-04) (hash(x)=148133023)
875
+ 7270 train 4.043428 (lr=2.7135e-04) (hash(x)=148006681)
876
+ 7280 train 4.076955 (lr=2.6989e-04) (hash(x)=151397627)
877
+ 7290 train 4.037013 (lr=2.6844e-04) (hash(x)=141696871)
878
+ 7300 val loss 4.0375
879
+ 7300 val perplexity 56.6827
880
+ 7300 train 3.971579 (lr=2.6700e-04) (hash(x)=150018163)
881
+ 7310 train 4.015401 (lr=2.6556e-04) (hash(x)=154900050)
882
+ 7320 train 3.909890 (lr=2.6414e-04) (hash(x)=142608283)
883
+ 7330 train 4.012889 (lr=2.6272e-04) (hash(x)=148977837)
884
+ 7340 train 4.030307 (lr=2.6132e-04) (hash(x)=141676907)
885
+ 7350 train 3.916158 (lr=2.5992e-04) (hash(x)=147846997)
886
+ 7360 train 4.151368 (lr=2.5853e-04) (hash(x)=143252181)
887
+ 7370 train 3.982061 (lr=2.5715e-04) (hash(x)=155081717)
888
+ 7380 train 4.065177 (lr=2.5578e-04) (hash(x)=149678451)
889
+ 7390 train 3.892082 (lr=2.5442e-04) (hash(x)=159341637)
890
+ 7400 val loss 4.0363
891
+ 7400 val perplexity 56.6143
892
+ 7400 train 4.185328 (lr=2.5306e-04) (hash(x)=145351166)
893
+ 7410 train 3.997351 (lr=2.5172e-04) (hash(x)=162274009)
894
+ 7420 train 3.845787 (lr=2.5038e-04) (hash(x)=143254858)
895
+ 7430 train 3.956005 (lr=2.4906e-04) (hash(x)=143049884)
896
+ 7440 train 3.880637 (lr=2.4774e-04) (hash(x)=141768538)
897
+ 7450 train 3.904248 (lr=2.4644e-04) (hash(x)=139895310)
898
+ 7460 train 3.769994 (lr=2.4514e-04) (hash(x)=155900365)
899
+ 7470 train 3.753873 (lr=2.4385e-04) (hash(x)=145048215)
900
+ 7480 train 3.898348 (lr=2.4257e-04) (hash(x)=136992624)
901
+ 7490 train 4.004033 (lr=2.4130e-04) (hash(x)=141424789)
902
+ 7500 val loss 4.0421
903
+ 7500 val perplexity 56.9442
904
+ 7500 train 3.860833 (lr=2.4004e-04) (hash(x)=145292116)
905
+ 7510 train 3.890233 (lr=2.3879e-04) (hash(x)=144818029)
906
+ 7520 train 4.068140 (lr=2.3754e-04) (hash(x)=152903515)
907
+ 7530 train 4.075236 (lr=2.3631e-04) (hash(x)=156934724)
908
+ 7540 train 4.299116 (lr=2.3509e-04) (hash(x)=156695926)
909
+ 7550 train 4.055875 (lr=2.3387e-04) (hash(x)=149115609)
910
+ 7560 train 4.033619 (lr=2.3267e-04) (hash(x)=143974224)
911
+ 7570 train 3.993920 (lr=2.3147e-04) (hash(x)=148204424)
912
+ 7580 train 4.073923 (lr=2.3029e-04) (hash(x)=150177910)
913
+ 7590 train 4.086332 (lr=2.2911e-04) (hash(x)=140325629)
914
+ 7600 val loss 4.0276
915
+ 7600 val perplexity 56.1262
916
+ 7600 train 3.970973 (lr=2.2794e-04) (hash(x)=150235132)
917
+ 7610 train 4.092599 (lr=2.2678e-04) (hash(x)=157594837)
918
+ 7620 train 4.103199 (lr=2.2564e-04) (hash(x)=146696749)
919
+ 7630 train 4.137468 (lr=2.2450e-04) (hash(x)=148463580)
920
+ 7640 train 3.960524 (lr=2.2337e-04) (hash(x)=140271046)
921
+ 7650 train 4.019231 (lr=2.2225e-04) (hash(x)=159952319)
922
+ 7660 train 4.019267 (lr=2.2114e-04) (hash(x)=146307392)
923
+ 7670 train 3.968556 (lr=2.2004e-04) (hash(x)=145027988)
924
+ 7680 train 4.031213 (lr=2.1894e-04) (hash(x)=147729698)
925
+ 7690 train 4.022460 (lr=2.1786e-04) (hash(x)=143401210)
926
+ 7700 val loss 4.0214
927
+ 7700 val perplexity 55.7777
928
+ 7700 train 3.950626 (lr=2.1679e-04) (hash(x)=154543455)
929
+ 7710 train 4.028580 (lr=2.1573e-04) (hash(x)=151600456)
930
+ 7720 train 3.942882 (lr=2.1468e-04) (hash(x)=165913803)
931
+ 7730 train 3.812745 (lr=2.1363e-04) (hash(x)=144942059)
932
+ 7740 train 4.036946 (lr=2.1260e-04) (hash(x)=150302103)
933
+ 7750 train 3.982657 (lr=2.1158e-04) (hash(x)=155782458)
934
+ 7760 train 4.021356 (lr=2.1056e-04) (hash(x)=147824384)
935
+ 7770 train 4.030849 (lr=2.0956e-04) (hash(x)=149177657)
936
+ 7780 train 4.047894 (lr=2.0856e-04) (hash(x)=164446738)
937
+ 7790 train 3.858728 (lr=2.0758e-04) (hash(x)=146417077)
938
+ 7800 val loss 4.0239
939
+ 7800 val perplexity 55.9198
940
+ 7800 train 3.998363 (lr=2.0660e-04) (hash(x)=142456852)
941
+ 7810 train 3.709593 (lr=2.0564e-04) (hash(x)=146221807)
942
+ 7820 train 3.890103 (lr=2.0468e-04) (hash(x)=154136533)
943
+ 7830 train 3.802536 (lr=2.0373e-04) (hash(x)=147492735)
944
+ 7840 train 3.963478 (lr=2.0280e-04) (hash(x)=143481291)
945
+ 7850 train 3.793144 (lr=2.0187e-04) (hash(x)=136749307)
946
+ 7860 train 3.813534 (lr=2.0096e-04) (hash(x)=128985347)
947
+ 7870 train 3.717208 (lr=2.0005e-04) (hash(x)=152897363)
948
+ 7880 train 3.925831 (lr=1.9915e-04) (hash(x)=151460360)
949
+ 7890 train 3.748500 (lr=1.9827e-04) (hash(x)=137481738)
950
+ 7900 val loss 4.0245
951
+ 7900 val perplexity 55.9535
952
+ 7900 train 3.819672 (lr=1.9739e-04) (hash(x)=147363479)
953
+ 7910 train 4.039794 (lr=1.9652e-04) (hash(x)=151165744)
954
+ 7920 train 4.219973 (lr=1.9566e-04) (hash(x)=146214226)
955
+ 7930 train 4.017499 (lr=1.9482e-04) (hash(x)=154457502)
956
+ 7940 train 3.927829 (lr=1.9398e-04) (hash(x)=162025201)
957
+ 7950 train 3.917063 (lr=1.9315e-04) (hash(x)=156488145)
958
+ 7960 train 4.194811 (lr=1.9233e-04) (hash(x)=154011584)
959
+ 7970 train 4.053802 (lr=1.9153e-04) (hash(x)=145330320)
960
+ 7980 train 3.955650 (lr=1.9073e-04) (hash(x)=148216312)
961
+ 7990 train 3.891703 (lr=1.8994e-04) (hash(x)=156819091)
962
+ 8000 val loss 4.0139
963
+ 8000 val perplexity 55.3614
964
+ 8000 train 4.051818 (lr=1.8917e-04) (hash(x)=156122973)
965
+ 8010 train 3.989439 (lr=1.8840e-04) (hash(x)=146933088)
966
+ 8020 train 4.093035 (lr=1.8764e-04) (hash(x)=151650746)
967
+ 8030 train 4.012912 (lr=1.8689e-04) (hash(x)=140692373)
968
+ 8040 train 4.204198 (lr=1.8616e-04) (hash(x)=147850094)
969
+ 8050 train 4.020403 (lr=1.8543e-04) (hash(x)=144454109)
970
+ 8060 train 3.879637 (lr=1.8471e-04) (hash(x)=152205124)
971
+ 8070 train 4.092819 (lr=1.8400e-04) (hash(x)=148918066)
972
+ 8080 train 4.026250 (lr=1.8331e-04) (hash(x)=146481024)
973
+ 8090 train 4.063261 (lr=1.8262e-04) (hash(x)=168785549)
974
+ 8100 val loss 4.0103
975
+ 8100 val perplexity 55.1649
976
+ 8100 train 4.087007 (lr=1.8194e-04) (hash(x)=156153179)
977
+ 8110 train 4.048580 (lr=1.8128e-04) (hash(x)=150498601)
978
+ 8120 train 4.040079 (lr=1.8062e-04) (hash(x)=142002714)
979
+ 8130 train 3.935630 (lr=1.7997e-04) (hash(x)=138450483)
980
+ 8140 train 4.018624 (lr=1.7934e-04) (hash(x)=145306351)
981
+ 8150 train 3.895334 (lr=1.7871e-04) (hash(x)=172502865)
982
+ 8160 train 3.999201 (lr=1.7810e-04) (hash(x)=156689649)
983
+ 8170 train 3.915990 (lr=1.7749e-04) (hash(x)=143766932)
984
+ 8180 train 4.056478 (lr=1.7689e-04) (hash(x)=150361321)
985
+ 8190 train 3.882954 (lr=1.7631e-04) (hash(x)=152664241)
986
+ 8200 val loss 4.0098
987
+ 8200 val perplexity 55.1364
988
+ 8200 train 4.168783 (lr=1.7573e-04) (hash(x)=146430698)
989
+ 8210 train 3.856670 (lr=1.7517e-04) (hash(x)=167362078)
990
+ 8220 train 4.070909 (lr=1.7461e-04) (hash(x)=157650942)
991
+ 8230 train 3.882697 (lr=1.7407e-04) (hash(x)=142102044)
992
+ 8240 train 3.924375 (lr=1.7354e-04) (hash(x)=153405900)
993
+ 8250 train 3.874124 (lr=1.7301e-04) (hash(x)=140222909)
994
+ 8260 train 4.004837 (lr=1.7250e-04) (hash(x)=137981854)
995
+ 8270 train 3.918634 (lr=1.7199e-04) (hash(x)=147576272)
996
+ 8280 train 3.938046 (lr=1.7150e-04) (hash(x)=157988195)
997
+ 8290 train 3.989228 (lr=1.7102e-04) (hash(x)=145715681)
998
+ 8300 val loss 4.0105
999
+ 8300 val perplexity 55.1747
1000
+ 8300 train 3.934768 (lr=1.7055e-04) (hash(x)=143507257)
1001
+ 8310 train 3.989935 (lr=1.7008e-04) (hash(x)=148930064)
1002
+ 8320 train 3.920303 (lr=1.6963e-04) (hash(x)=144134805)
1003
+ 8330 train 4.036876 (lr=1.6919e-04) (hash(x)=148418079)
1004
+ 8340 train 3.763815 (lr=1.6876e-04) (hash(x)=166518807)
1005
+ 8350 train 4.067300 (lr=1.6834e-04) (hash(x)=140042036)
1006
+ 8360 train 3.939672 (lr=1.6793e-04) (hash(x)=151427074)
1007
+ 8370 train 4.023929 (lr=1.6752e-04) (hash(x)=158389185)
1008
+ 8380 train 3.924505 (lr=1.6713e-04) (hash(x)=148193308)
1009
+ 8390 train 3.998194 (lr=1.6675e-04) (hash(x)=144289008)
1010
+ 8400 val loss 4.0011
1011
+ 8400 val perplexity 54.6579
1012
+ 8400 train 3.701216 (lr=1.6639e-04) (hash(x)=166272643)
1013
+ 8410 train 3.921342 (lr=1.6603e-04) (hash(x)=162244433)
1014
+ 8420 train 3.963708 (lr=1.6568e-04) (hash(x)=149271299)
1015
+ 8430 train 4.098791 (lr=1.6534e-04) (hash(x)=154135338)
1016
+ 8440 train 4.140888 (lr=1.6501e-04) (hash(x)=149661838)
1017
+ 8450 train 3.983823 (lr=1.6469e-04) (hash(x)=142205056)
1018
+ 8460 train 3.911945 (lr=1.6439e-04) (hash(x)=166652398)
1019
+ 8470 train 4.098294 (lr=1.6409e-04) (hash(x)=149029710)
1020
+ 8480 train 3.858336 (lr=1.6380e-04) (hash(x)=154217415)
1021
+ 8490 train 3.936216 (lr=1.6353e-04) (hash(x)=151526557)
1022
+ 8500 val loss 4.0015
1023
+ 8500 val perplexity 54.6824
1024
+ 8500 train 3.925799 (lr=1.6326e-04) (hash(x)=143887848)
1025
+ 8510 train 3.917280 (lr=1.6300e-04) (hash(x)=152324645)
1026
+ 8520 train 3.651400 (lr=1.6276e-04) (hash(x)=145124959)
1027
+ 8530 train 3.971555 (lr=1.6253e-04) (hash(x)=186319340)
1028
+ 8540 train 3.984204 (lr=1.6230e-04) (hash(x)=147388339)
1029
+ 8550 train 4.018062 (lr=1.6209e-04) (hash(x)=150118928)
1030
+ 8560 train 4.010619 (lr=1.6188e-04) (hash(x)=153985286)
1031
+ 8570 train 3.953693 (lr=1.6169e-04) (hash(x)=149582799)
1032
+ 8580 train 3.960850 (lr=1.6151e-04) (hash(x)=156867870)
1033
+ 8590 train 3.847930 (lr=1.6134e-04) (hash(x)=148773298)
1034
+ 8600 val loss 4.0042
1035
+ 8600 val perplexity 54.8302
1036
+ 8600 train 3.949675 (lr=1.6117e-04) (hash(x)=156900341)
1037
+ 8610 train 4.100622 (lr=1.6102e-04) (hash(x)=156790089)
1038
+ 8620 train 3.909747 (lr=1.6088e-04) (hash(x)=154680805)
1039
+ 8630 train 3.948655 (lr=1.6075e-04) (hash(x)=138107873)
1040
+ 8640 train 4.032140 (lr=1.6063e-04) (hash(x)=142554243)
1041
+ 8650 train 3.846131 (lr=1.6052e-04) (hash(x)=134360956)
1042
+ 8660 train 3.917255 (lr=1.6042e-04) (hash(x)=128599610)
1043
+ 8670 train 3.949328 (lr=1.6033e-04) (hash(x)=159628945)
1044
+ 8680 train 3.970998 (lr=1.6026e-04) (hash(x)=143161532)
1045
+ 8690 train 4.005953 (lr=1.6019e-04) (hash(x)=140104912)
1046
+ 8700 val loss 3.9955
1047
+ 8700 val perplexity 54.3516
1048
+ 8700 train 4.178931 (lr=1.6013e-04) (hash(x)=146417632)
1049
+ 8710 train 4.164752 (lr=1.6008e-04) (hash(x)=153671695)
1050
+ 8720 train 4.034028 (lr=1.6005e-04) (hash(x)=141680750)
1051
+ 8730 train 3.903031 (lr=1.6002e-04) (hash(x)=147279231)
1052
+ 8740 train 4.200152 (lr=1.6001e-04) (hash(x)=129729371)
1053
+ 8749 val loss 3.9956
1054
+ 8749 val perplexity 54.3587
lr16e-4_total_batch_size61440_baseline_seed1339/model_08749.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:9af83aed3203bd591636f2fbe8b7f6a539e73077e02dd067157be1808931a811
3
+ size 92843394
lr16e-4_total_batch_size61440_baseline_seed1339/optimizer_08749.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:209580948780dc1f985af57c7a03f33e5b4462677636853ec4cfe2fb05620bd1
3
+ size 179406214