andrew-healey commited on
Commit
c8c45dc
·
verified ·
1 Parent(s): 280e188

Upload folder using huggingface_hub

Browse files
logs/64_latent_masks_degenerate/degen_compile_seed_1340/args.json ADDED
@@ -0,0 +1 @@
 
 
1
+ {"hellaswag": true, "attention_kind": "selective", "log_dir": "logs/64_latent_masks_degenerate/degen_compile_seed_1340", "resume_checkpoint": null, "resume_optimizer": false, "add_a_head": false, "add_head_to_start": true, "new_head_init": "normal", "n_heads": 12, "n_embd": 768, "head_dim": 64, "protect_bos_token": true, "prevent_from_masking_myself": true, "max_steps": 4375, "warmup_steps": 250, "group": "64_latent_masks_degenerate", "use_wandb": true, "kill_self_after_run": false, "random_seed": 1340, "memory_penalty_epsilon": 0.1, "selection_head_linear_combo": "n_latent_masks", "selection_head_linear_combo_scale": 0.0, "disable_selection_head_linear_combo_bias": true, "assert_latent_matches_no_head": false, "n_latent_masks": 1, "init_latent_masks_to_identity": true, "init_latent_masks_to_inverse": false, "latent_mask_scale": null, "latent_mask_runtime_multiplier": null, "latent_mask_sigmoid": false, "S_layernorm": false, "one_head_per_latent_mask": false, "protection_kind": "none", "leaky_relu_alpha": null, "leaky_relu_bias": null, "use_compile": true, "use_mini_model": false, "upload_to_hf": true, "seq_len": 1024, "batch_size": 4, "total_batch_size": 131072, "protection_head_scaling_factor": 1.0, "protection_head_bias": 0.0, "n_sliced_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": 0.003, "decay_lr": true, "readout_zero_init": false, "query_zero_init": false, "l1_loss": false, "debugpy": false, "key": "degen_compile_seed_1340", "latent_mask_precision": "float32"}
logs/64_latent_masks_degenerate/degen_compile_seed_1340/dataloader_04374.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:994809a5ef065b3b39ff8d94ca69f209f3eedb103954655f09fa348a0918b20e
3
+ size 964
logs/64_latent_masks_degenerate/degen_compile_seed_1340/log2.txt ADDED
@@ -0,0 +1,529 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ max_steps: 4375
2
+ 0 val loss 10.9948
3
+ 0 val perplexity 59564.3750
4
+ 0 train 10.992901 (lr=4.1958e-06) (hash(x)=19978355)
5
+ 10 train 9.624754 (lr=4.6154e-05) (hash(x)=24395949)
6
+ 20 train 9.126051 (lr=8.8112e-05) (hash(x)=20203571)
7
+ 30 train 8.426197 (lr=1.3007e-04) (hash(x)=22227211)
8
+ 40 train 7.793196 (lr=1.7203e-04) (hash(x)=16492366)
9
+ 50 train 7.600268 (lr=2.1399e-04) (hash(x)=17690059)
10
+ 60 train 7.510891 (lr=2.5594e-04) (hash(x)=19389177)
11
+ 70 train 7.470484 (lr=2.9790e-04) (hash(x)=16938017)
12
+ 80 train 7.318397 (lr=3.3986e-04) (hash(x)=18013370)
13
+ 90 train 7.093542 (lr=3.8182e-04) (hash(x)=22273301)
14
+ 100 val loss 7.0181
15
+ 100 val perplexity 1116.6989
16
+ 100 train 6.983054 (lr=4.2378e-04) (hash(x)=21867969)
17
+ 110 train 6.864524 (lr=4.6573e-04) (hash(x)=17559385)
18
+ 120 train 6.773450 (lr=5.0769e-04) (hash(x)=21415593)
19
+ 130 train 6.621806 (lr=5.4965e-04) (hash(x)=17493363)
20
+ 140 train 6.506111 (lr=5.9161e-04) (hash(x)=19060392)
21
+ 150 train 6.457766 (lr=6.3357e-04) (hash(x)=18669634)
22
+ 160 train 6.323212 (lr=6.7552e-04) (hash(x)=19128903)
23
+ 170 train 6.361216 (lr=7.1748e-04) (hash(x)=21173132)
24
+ 180 train 6.345321 (lr=7.5944e-04) (hash(x)=21482016)
25
+ 190 train 6.112868 (lr=8.0140e-04) (hash(x)=19274288)
26
+ 200 val loss 6.0914
27
+ 200 val perplexity 442.0237
28
+ 200 train 6.047045 (lr=8.4336e-04) (hash(x)=17153997)
29
+ 210 train 5.953410 (lr=8.8531e-04) (hash(x)=18659881)
30
+ 220 train 5.909381 (lr=9.2727e-04) (hash(x)=13041762)
31
+ 230 train 5.899319 (lr=9.6923e-04) (hash(x)=19089409)
32
+ 240 train 5.962750 (lr=1.0112e-03) (hash(x)=21588907)
33
+ 250 train 5.898584 (lr=1.0531e-03) (hash(x)=22853384)
34
+ 260 train 5.842453 (lr=1.0951e-03) (hash(x)=21258918)
35
+ 270 train 5.697768 (lr=1.1371e-03) (hash(x)=18064675)
36
+ 280 train 5.684769 (lr=1.1790e-03) (hash(x)=17783654)
37
+ 290 train 5.618798 (lr=1.2210e-03) (hash(x)=16206303)
38
+ 300 val loss 5.6017
39
+ 300 val perplexity 270.8839
40
+ 300 train 5.620321 (lr=1.2629e-03) (hash(x)=23548851)
41
+ 310 train 5.464726 (lr=1.3049e-03) (hash(x)=17695462)
42
+ 320 train 5.506824 (lr=1.3469e-03) (hash(x)=16887671)
43
+ 330 train 5.462076 (lr=1.3888e-03) (hash(x)=19956923)
44
+ 340 train 5.466666 (lr=1.4308e-03) (hash(x)=16428364)
45
+ 350 train 5.347406 (lr=1.4727e-03) (hash(x)=20437634)
46
+ 360 train 5.374386 (lr=1.5147e-03) (hash(x)=19622578)
47
+ 370 train 5.358817 (lr=1.5566e-03) (hash(x)=21294724)
48
+ 380 train 5.343548 (lr=1.5986e-03) (hash(x)=18869667)
49
+ 390 train 5.295751 (lr=1.6406e-03) (hash(x)=22256474)
50
+ 400 val loss 5.2871
51
+ 400 val perplexity 197.7736
52
+ 400 train 5.357620 (lr=1.6825e-03) (hash(x)=20048534)
53
+ 410 train 5.274522 (lr=1.7245e-03) (hash(x)=23042068)
54
+ 420 train 5.264373 (lr=1.7664e-03) (hash(x)=19937603)
55
+ 430 train 5.288309 (lr=1.8084e-03) (hash(x)=22150243)
56
+ 440 train 5.252538 (lr=1.8503e-03) (hash(x)=23468794)
57
+ 450 train 5.173120 (lr=1.8923e-03) (hash(x)=18432856)
58
+ 460 train 5.247069 (lr=1.9343e-03) (hash(x)=21494081)
59
+ 470 train 5.124979 (lr=1.9762e-03) (hash(x)=18494966)
60
+ 480 train 5.042526 (lr=2.0182e-03) (hash(x)=20099891)
61
+ 490 train 5.107101 (lr=2.0601e-03) (hash(x)=21778663)
62
+ 500 val loss 5.0668
63
+ 500 val perplexity 158.6695
64
+ 500 train 4.993870 (lr=2.1021e-03) (hash(x)=21164498)
65
+ 510 train 4.990623 (lr=2.1441e-03) (hash(x)=20617181)
66
+ 520 train 5.036397 (lr=2.1860e-03) (hash(x)=24124812)
67
+ 530 train 4.977835 (lr=2.2280e-03) (hash(x)=18270550)
68
+ 540 train 5.001924 (lr=2.2699e-03) (hash(x)=17643805)
69
+ 550 train 4.875770 (lr=2.3119e-03) (hash(x)=19633450)
70
+ 560 train 4.898574 (lr=2.3538e-03) (hash(x)=18215455)
71
+ 570 train 5.013833 (lr=2.3958e-03) (hash(x)=23240966)
72
+ 580 train 4.871814 (lr=2.4378e-03) (hash(x)=20144929)
73
+ 590 train 4.884547 (lr=2.4797e-03) (hash(x)=16296841)
74
+ 600 val loss 4.7838
75
+ 600 val perplexity 119.5526
76
+ 600 train 4.813649 (lr=2.5217e-03) (hash(x)=19836338)
77
+ 610 train 5.086088 (lr=2.5636e-03) (hash(x)=19928421)
78
+ 620 train 4.699828 (lr=2.6056e-03) (hash(x)=17727783)
79
+ 630 train 4.755662 (lr=2.6476e-03) (hash(x)=20940923)
80
+ 640 train 4.699132 (lr=2.6895e-03) (hash(x)=29374503)
81
+ 650 train 4.694619 (lr=2.7315e-03) (hash(x)=18000892)
82
+ 660 train 4.729548 (lr=2.7734e-03) (hash(x)=20632048)
83
+ 670 train 4.725777 (lr=2.8154e-03) (hash(x)=16874750)
84
+ 680 train 4.832220 (lr=2.8573e-03) (hash(x)=16996745)
85
+ 690 train 4.615730 (lr=2.8993e-03) (hash(x)=17242267)
86
+ 700 val loss 4.6048
87
+ 700 val perplexity 99.9616
88
+ 700 train 4.531688 (lr=2.9413e-03) (hash(x)=4532237)
89
+ 710 train 4.852465 (lr=2.9832e-03) (hash(x)=18455888)
90
+ 720 train 4.596185 (lr=3.0000e-03) (hash(x)=22389104)
91
+ 730 train 4.576728 (lr=2.9999e-03) (hash(x)=19048232)
92
+ 740 train 4.522534 (lr=2.9997e-03) (hash(x)=21823621)
93
+ 750 train 4.531976 (lr=2.9994e-03) (hash(x)=19834992)
94
+ 760 train 4.419338 (lr=2.9990e-03) (hash(x)=20804217)
95
+ 770 train 4.327791 (lr=2.9985e-03) (hash(x)=21676803)
96
+ 780 train 4.284773 (lr=2.9979e-03) (hash(x)=20436543)
97
+ 790 train 4.332989 (lr=2.9972e-03) (hash(x)=24049409)
98
+ 800 val loss 4.4645
99
+ 800 val perplexity 86.8809
100
+ 800 train 4.208374 (lr=2.9964e-03) (hash(x)=21101828)
101
+ 810 train 4.423259 (lr=2.9955e-03) (hash(x)=21158051)
102
+ 820 train 4.384329 (lr=2.9945e-03) (hash(x)=17068830)
103
+ 830 train 4.411280 (lr=2.9934e-03) (hash(x)=20708816)
104
+ 840 train 4.458148 (lr=2.9922e-03) (hash(x)=23031486)
105
+ 850 train 4.476852 (lr=2.9909e-03) (hash(x)=17808332)
106
+ 860 train 4.357502 (lr=2.9896e-03) (hash(x)=19935633)
107
+ 870 train 4.393702 (lr=2.9881e-03) (hash(x)=21546448)
108
+ 880 train 4.408737 (lr=2.9865e-03) (hash(x)=18928060)
109
+ 890 train 4.301801 (lr=2.9848e-03) (hash(x)=19620142)
110
+ 900 val loss 4.3266
111
+ 900 val perplexity 75.6856
112
+ 900 train 4.333351 (lr=2.9830e-03) (hash(x)=15281589)
113
+ 910 train 4.265833 (lr=2.9811e-03) (hash(x)=19313761)
114
+ 920 train 4.275827 (lr=2.9792e-03) (hash(x)=21534098)
115
+ 930 train 4.277096 (lr=2.9771e-03) (hash(x)=22007834)
116
+ 940 train 4.375876 (lr=2.9749e-03) (hash(x)=19528278)
117
+ 950 train 4.288821 (lr=2.9726e-03) (hash(x)=18826968)
118
+ 960 train 4.260841 (lr=2.9703e-03) (hash(x)=24147820)
119
+ 970 train 4.093845 (lr=2.9678e-03) (hash(x)=22646220)
120
+ 980 train 4.090410 (lr=2.9652e-03) (hash(x)=21026041)
121
+ 990 train 4.103734 (lr=2.9626e-03) (hash(x)=17067127)
122
+ 1000 val loss 4.2633
123
+ 1000 val perplexity 71.0423
124
+ 1000 train 4.433141 (lr=2.9598e-03) (hash(x)=18557650)
125
+ 1010 train 4.316326 (lr=2.9570e-03) (hash(x)=20276417)
126
+ 1020 train 4.341493 (lr=2.9540e-03) (hash(x)=25973944)
127
+ 1030 train 4.350534 (lr=2.9510e-03) (hash(x)=21253298)
128
+ 1040 train 4.197713 (lr=2.9478e-03) (hash(x)=24702365)
129
+ 1050 train 4.277829 (lr=2.9446e-03) (hash(x)=19691634)
130
+ 1060 train 4.214283 (lr=2.9412e-03) (hash(x)=16659615)
131
+ 1070 train 4.184013 (lr=2.9378e-03) (hash(x)=18095563)
132
+ 1080 train 4.174944 (lr=2.9343e-03) (hash(x)=18443383)
133
+ 1090 train 4.206981 (lr=2.9307e-03) (hash(x)=18330349)
134
+ 1100 val loss 4.1713
135
+ 1100 val perplexity 64.8017
136
+ 1100 train 4.158901 (lr=2.9270e-03) (hash(x)=19031432)
137
+ 1110 train 4.301467 (lr=2.9231e-03) (hash(x)=27724663)
138
+ 1120 train 4.116494 (lr=2.9192e-03) (hash(x)=20063068)
139
+ 1130 train 4.045102 (lr=2.9152e-03) (hash(x)=18456567)
140
+ 1140 train 4.071517 (lr=2.9112e-03) (hash(x)=21274378)
141
+ 1150 train 4.336835 (lr=2.9070e-03) (hash(x)=20166104)
142
+ 1160 train 3.878881 (lr=2.9027e-03) (hash(x)=19016072)
143
+ 1170 train 3.933783 (lr=2.8983e-03) (hash(x)=20983879)
144
+ 1180 train 3.904728 (lr=2.8939e-03) (hash(x)=22773283)
145
+ 1190 train 4.154860 (lr=2.8893e-03) (hash(x)=19469721)
146
+ 1200 val loss 4.1179
147
+ 1200 val perplexity 61.4316
148
+ 1200 train 4.150585 (lr=2.8847e-03) (hash(x)=21239510)
149
+ 1210 train 4.078050 (lr=2.8800e-03) (hash(x)=21515094)
150
+ 1220 train 4.057166 (lr=2.8751e-03) (hash(x)=16618399)
151
+ 1230 train 4.174292 (lr=2.8702e-03) (hash(x)=19402906)
152
+ 1240 train 4.247060 (lr=2.8652e-03) (hash(x)=21224237)
153
+ 1250 train 4.024019 (lr=2.8601e-03) (hash(x)=16261179)
154
+ 1260 train 4.146606 (lr=2.8550e-03) (hash(x)=19045671)
155
+ 1270 train 4.084397 (lr=2.8497e-03) (hash(x)=13880880)
156
+ 1280 train 4.143082 (lr=2.8443e-03) (hash(x)=22126650)
157
+ 1290 train 3.968640 (lr=2.8389e-03) (hash(x)=22005339)
158
+ 1300 val loss 4.0424
159
+ 1300 val perplexity 56.9606
160
+ 1300 train 4.046616 (lr=2.8333e-03) (hash(x)=22293195)
161
+ 1310 train 4.075191 (lr=2.8277e-03) (hash(x)=19908750)
162
+ 1320 train 3.939063 (lr=2.8220e-03) (hash(x)=24199231)
163
+ 1330 train 3.921196 (lr=2.8162e-03) (hash(x)=20450695)
164
+ 1340 train 3.858055 (lr=2.8103e-03) (hash(x)=17483865)
165
+ 1350 train 3.877633 (lr=2.8044e-03) (hash(x)=20351158)
166
+ 1360 train 3.871216 (lr=2.7983e-03) (hash(x)=23798487)
167
+ 1370 train 3.823091 (lr=2.7922e-03) (hash(x)=20090485)
168
+ 1380 train 3.971621 (lr=2.7860e-03) (hash(x)=24781076)
169
+ 1390 train 4.046624 (lr=2.7797e-03) (hash(x)=21234884)
170
+ 1400 val loss 3.9947
171
+ 1400 val perplexity 54.3095
172
+ 1400 train 3.967570 (lr=2.7733e-03) (hash(x)=22226319)
173
+ 1410 train 4.214735 (lr=2.7668e-03) (hash(x)=25309506)
174
+ 1420 train 4.073005 (lr=2.7603e-03) (hash(x)=22351162)
175
+ 1430 train 3.941210 (lr=2.7536e-03) (hash(x)=20164493)
176
+ 1440 train 3.992082 (lr=2.7469e-03) (hash(x)=16571343)
177
+ 1450 train 4.097606 (lr=2.7401e-03) (hash(x)=17493318)
178
+ 1460 train 4.057437 (lr=2.7333e-03) (hash(x)=19324896)
179
+ 1470 train 3.979571 (lr=2.7263e-03) (hash(x)=19804956)
180
+ 1480 train 4.006355 (lr=2.7193e-03) (hash(x)=19628407)
181
+ 1490 train 3.948212 (lr=2.7121e-03) (hash(x)=19770129)
182
+ 1500 val loss 3.9603
183
+ 1500 val perplexity 52.4754
184
+ 1500 train 3.940845 (lr=2.7050e-03) (hash(x)=16945773)
185
+ 1510 train 3.957130 (lr=2.6977e-03) (hash(x)=18780367)
186
+ 1520 train 3.778744 (lr=2.6903e-03) (hash(x)=19997660)
187
+ 1530 train 4.054967 (lr=2.6829e-03) (hash(x)=17063727)
188
+ 1540 train 3.970093 (lr=2.6754e-03) (hash(x)=20300139)
189
+ 1550 train 3.859628 (lr=2.6678e-03) (hash(x)=21541648)
190
+ 1560 train 3.931530 (lr=2.6602e-03) (hash(x)=25169762)
191
+ 1570 train 3.922239 (lr=2.6525e-03) (hash(x)=18020155)
192
+ 1580 train 3.853418 (lr=2.6447e-03) (hash(x)=21439415)
193
+ 1590 train 3.880779 (lr=2.6368e-03) (hash(x)=20345952)
194
+ 1600 val loss 3.9230
195
+ 1600 val perplexity 50.5520
196
+ 1600 train 3.788270 (lr=2.6289e-03) (hash(x)=19677331)
197
+ 1610 train 3.907609 (lr=2.6208e-03) (hash(x)=23142852)
198
+ 1620 train 3.886032 (lr=2.6128e-03) (hash(x)=18712875)
199
+ 1630 train 3.769138 (lr=2.6046e-03) (hash(x)=17940151)
200
+ 1640 train 3.739172 (lr=2.5964e-03) (hash(x)=18928868)
201
+ 1650 train 3.897854 (lr=2.5881e-03) (hash(x)=21562811)
202
+ 1660 train 3.978559 (lr=2.5797e-03) (hash(x)=19917546)
203
+ 1670 train 3.984498 (lr=2.5713e-03) (hash(x)=19862190)
204
+ 1680 train 3.968380 (lr=2.5628e-03) (hash(x)=19510122)
205
+ 1690 train 3.854952 (lr=2.5542e-03) (hash(x)=18819854)
206
+ 1700 val loss 3.8848
207
+ 1700 val perplexity 48.6569
208
+ 1700 train 3.834157 (lr=2.5455e-03) (hash(x)=20098069)
209
+ 1710 train 3.841183 (lr=2.5368e-03) (hash(x)=19180845)
210
+ 1720 train 3.892183 (lr=2.5281e-03) (hash(x)=22610840)
211
+ 1730 train 3.894111 (lr=2.5192e-03) (hash(x)=17804171)
212
+ 1740 train 3.859748 (lr=2.5103e-03) (hash(x)=18275747)
213
+ 1750 train 3.787904 (lr=2.5014e-03) (hash(x)=20884937)
214
+ 1760 train 3.792976 (lr=2.4924e-03) (hash(x)=18783179)
215
+ 1770 train 3.839625 (lr=2.4833e-03) (hash(x)=20225710)
216
+ 1780 train 3.773698 (lr=2.4741e-03) (hash(x)=19910233)
217
+ 1790 train 3.877656 (lr=2.4649e-03) (hash(x)=19722753)
218
+ 1800 val loss 3.8615
219
+ 1800 val perplexity 47.5366
220
+ 1800 train 3.867217 (lr=2.4556e-03) (hash(x)=21871844)
221
+ 1810 train 3.873319 (lr=2.4463e-03) (hash(x)=19497156)
222
+ 1820 train 3.905986 (lr=2.4369e-03) (hash(x)=18788042)
223
+ 1830 train 3.857513 (lr=2.4275e-03) (hash(x)=19961057)
224
+ 1840 train 3.816167 (lr=2.4180e-03) (hash(x)=17863548)
225
+ 1850 train 3.905554 (lr=2.4084e-03) (hash(x)=17751881)
226
+ 1860 train 3.789345 (lr=2.3988e-03) (hash(x)=19219734)
227
+ 1870 train 3.783065 (lr=2.3891e-03) (hash(x)=19455814)
228
+ 1880 train 3.826757 (lr=2.3794e-03) (hash(x)=20580620)
229
+ 1890 train 3.797682 (lr=2.3696e-03) (hash(x)=17077823)
230
+ 1900 val loss 3.8460
231
+ 1900 val perplexity 46.8068
232
+ 1900 train 3.808995 (lr=2.3598e-03) (hash(x)=17322645)
233
+ 1910 train 3.723638 (lr=2.3499e-03) (hash(x)=20740743)
234
+ 1920 train 3.906842 (lr=2.3400e-03) (hash(x)=20599034)
235
+ 1930 train 3.846735 (lr=2.3300e-03) (hash(x)=21358394)
236
+ 1940 train 3.916583 (lr=2.3200e-03) (hash(x)=19899662)
237
+ 1950 train 3.849501 (lr=2.3099e-03) (hash(x)=15623618)
238
+ 1960 train 3.717166 (lr=2.2998e-03) (hash(x)=19094692)
239
+ 1970 train 3.805766 (lr=2.2896e-03) (hash(x)=21521915)
240
+ 1980 train 3.764127 (lr=2.2793e-03) (hash(x)=18396717)
241
+ 1990 train 3.823514 (lr=2.2691e-03) (hash(x)=20176898)
242
+ 2000 val loss 3.8040
243
+ 2000 val perplexity 44.8813
244
+ 2000 train 3.765759 (lr=2.2588e-03) (hash(x)=17939830)
245
+ 2010 train 3.810364 (lr=2.2484e-03) (hash(x)=18313312)
246
+ 2020 train 3.714509 (lr=2.2380e-03) (hash(x)=22132416)
247
+ 2030 train 3.755854 (lr=2.2275e-03) (hash(x)=18403620)
248
+ 2040 train 3.650112 (lr=2.2170e-03) (hash(x)=19852232)
249
+ 2050 train 3.657945 (lr=2.2065e-03) (hash(x)=21010646)
250
+ 2060 train 3.846443 (lr=2.1959e-03) (hash(x)=21119229)
251
+ 2070 train 3.908116 (lr=2.1853e-03) (hash(x)=22447456)
252
+ 2080 train 3.792986 (lr=2.1746e-03) (hash(x)=19656298)
253
+ 2090 train 3.845864 (lr=2.1639e-03) (hash(x)=17300349)
254
+ 2100 val loss 3.7815
255
+ 2100 val perplexity 43.8835
256
+ 2100 train 3.741989 (lr=2.1532e-03) (hash(x)=19489545)
257
+ 2110 train 3.679748 (lr=2.1424e-03) (hash(x)=18592164)
258
+ 2120 train 3.762527 (lr=2.1316e-03) (hash(x)=18607056)
259
+ 2130 train 3.777504 (lr=2.1208e-03) (hash(x)=18652918)
260
+ 2140 train 3.796880 (lr=2.1099e-03) (hash(x)=21040103)
261
+ 2150 train 3.661484 (lr=2.0990e-03) (hash(x)=16399764)
262
+ 2160 train 3.715065 (lr=2.0881e-03) (hash(x)=17535477)
263
+ 2170 train 3.636092 (lr=2.0771e-03) (hash(x)=19890415)
264
+ 2180 train 3.754170 (lr=2.0661e-03) (hash(x)=16652534)
265
+ 2190 train 3.702074 (lr=2.0550e-03) (hash(x)=22131443)
266
+ 2200 val loss 3.7649
267
+ 2200 val perplexity 43.1597
268
+ 2200 train 3.807282 (lr=2.0440e-03) (hash(x)=19148364)
269
+ 2210 train 3.887057 (lr=2.0329e-03) (hash(x)=15715527)
270
+ 2220 train 3.731326 (lr=2.0217e-03) (hash(x)=23414699)
271
+ 2230 train 3.811237 (lr=2.0106e-03) (hash(x)=19689870)
272
+ 2240 train 3.822787 (lr=1.9994e-03) (hash(x)=22678814)
273
+ 2250 train 3.676825 (lr=1.9882e-03) (hash(x)=18976933)
274
+ 2260 train 3.771121 (lr=1.9770e-03) (hash(x)=20397863)
275
+ 2270 train 3.697568 (lr=1.9657e-03) (hash(x)=19308365)
276
+ 2280 train 3.702699 (lr=1.9544e-03) (hash(x)=18078666)
277
+ 2290 train 3.724024 (lr=1.9431e-03) (hash(x)=20419667)
278
+ 2300 val loss 3.7299
279
+ 2300 val perplexity 41.6747
280
+ 2300 train 3.704128 (lr=1.9318e-03) (hash(x)=16788415)
281
+ 2310 train 3.662086 (lr=1.9205e-03) (hash(x)=18566522)
282
+ 2320 train 3.720748 (lr=1.9091e-03) (hash(x)=21387107)
283
+ 2330 train 3.853412 (lr=1.8977e-03) (hash(x)=16815344)
284
+ 2340 train 3.820379 (lr=1.8863e-03) (hash(x)=19301309)
285
+ 2350 train 3.767606 (lr=1.8749e-03) (hash(x)=18411578)
286
+ 2360 train 3.940537 (lr=1.8635e-03) (hash(x)=17212565)
287
+ 2370 train 3.751610 (lr=1.8520e-03) (hash(x)=19083591)
288
+ 2380 train 3.667731 (lr=1.8406e-03) (hash(x)=19095330)
289
+ 2390 train 3.700210 (lr=1.8291e-03) (hash(x)=22137679)
290
+ 2400 val loss 3.7102
291
+ 2400 val perplexity 40.8628
292
+ 2400 train 3.689356 (lr=1.8176e-03) (hash(x)=19988438)
293
+ 2410 train 3.781353 (lr=1.8061e-03) (hash(x)=18845683)
294
+ 2420 train 3.714984 (lr=1.7946e-03) (hash(x)=18942135)
295
+ 2430 train 3.628436 (lr=1.7830e-03) (hash(x)=19844600)
296
+ 2440 train 3.599487 (lr=1.7715e-03) (hash(x)=20037517)
297
+ 2450 train 3.588943 (lr=1.7600e-03) (hash(x)=18071985)
298
+ 2460 train 3.620335 (lr=1.7484e-03) (hash(x)=19498889)
299
+ 2470 train 3.827003 (lr=1.7368e-03) (hash(x)=21020015)
300
+ 2480 train 3.830159 (lr=1.7253e-03) (hash(x)=24541400)
301
+ 2490 train 3.780132 (lr=1.7137e-03) (hash(x)=17117268)
302
+ 2500 val loss 3.6923
303
+ 2500 val perplexity 40.1380
304
+ 2500 train 3.716726 (lr=1.7021e-03) (hash(x)=20019265)
305
+ 2510 train 3.799250 (lr=1.6906e-03) (hash(x)=15424606)
306
+ 2520 train 3.767390 (lr=1.6790e-03) (hash(x)=16589718)
307
+ 2530 train 3.678441 (lr=1.6674e-03) (hash(x)=21793520)
308
+ 2540 train 3.639620 (lr=1.6558e-03) (hash(x)=26991001)
309
+ 2550 train 3.585176 (lr=1.6442e-03) (hash(x)=19542562)
310
+ 2560 train 3.693245 (lr=1.6326e-03) (hash(x)=16342890)
311
+ 2570 train 3.551289 (lr=1.6210e-03) (hash(x)=22032122)
312
+ 2580 train 3.548341 (lr=1.6094e-03) (hash(x)=20442648)
313
+ 2590 train 3.515868 (lr=1.5979e-03) (hash(x)=20229881)
314
+ 2600 val loss 3.6687
315
+ 2600 val perplexity 39.1997
316
+ 2600 train 3.605436 (lr=1.5863e-03) (hash(x)=20759842)
317
+ 2610 train 3.549573 (lr=1.5747e-03) (hash(x)=18590738)
318
+ 2620 train 3.732407 (lr=1.5632e-03) (hash(x)=20016008)
319
+ 2630 train 3.679354 (lr=1.5516e-03) (hash(x)=20966006)
320
+ 2640 train 3.760187 (lr=1.5400e-03) (hash(x)=20480389)
321
+ 2650 train 3.631406 (lr=1.5285e-03) (hash(x)=17625327)
322
+ 2660 train 3.591155 (lr=1.5170e-03) (hash(x)=17462746)
323
+ 2670 train 3.682171 (lr=1.5054e-03) (hash(x)=24540619)
324
+ 2680 train 3.604132 (lr=1.4939e-03) (hash(x)=18300466)
325
+ 2690 train 3.568769 (lr=1.4824e-03) (hash(x)=20785354)
326
+ 2700 val loss 3.6453
327
+ 2700 val perplexity 38.2956
328
+ 2700 train 3.671513 (lr=1.4709e-03) (hash(x)=19689301)
329
+ 2710 train 3.575251 (lr=1.4594e-03) (hash(x)=18041572)
330
+ 2720 train 3.594952 (lr=1.4480e-03) (hash(x)=23202500)
331
+ 2730 train 3.563148 (lr=1.4365e-03) (hash(x)=25706624)
332
+ 2740 train 3.574028 (lr=1.4251e-03) (hash(x)=24172973)
333
+ 2750 train 3.609172 (lr=1.4137e-03) (hash(x)=22379491)
334
+ 2760 train 3.633919 (lr=1.4023e-03) (hash(x)=21701022)
335
+ 2770 train 3.682412 (lr=1.3909e-03) (hash(x)=21951124)
336
+ 2780 train 3.629105 (lr=1.3795e-03) (hash(x)=19525099)
337
+ 2790 train 3.649443 (lr=1.3682e-03) (hash(x)=18035060)
338
+ 2800 val loss 3.6222
339
+ 2800 val perplexity 37.4206
340
+ 2800 train 3.657320 (lr=1.3569e-03) (hash(x)=19475507)
341
+ 2810 train 3.619113 (lr=1.3456e-03) (hash(x)=20631788)
342
+ 2820 train 3.636583 (lr=1.3343e-03) (hash(x)=21229388)
343
+ 2830 train 3.621267 (lr=1.3230e-03) (hash(x)=17707440)
344
+ 2840 train 3.652616 (lr=1.3118e-03) (hash(x)=17822842)
345
+ 2850 train 3.596876 (lr=1.3006e-03) (hash(x)=20668384)
346
+ 2860 train 3.549347 (lr=1.2894e-03) (hash(x)=21880436)
347
+ 2870 train 3.567429 (lr=1.2783e-03) (hash(x)=22236229)
348
+ 2880 train 3.517517 (lr=1.2671e-03) (hash(x)=22480364)
349
+ 2890 train 3.460436 (lr=1.2560e-03) (hash(x)=20782341)
350
+ 2900 val loss 3.6175
351
+ 2900 val perplexity 37.2449
352
+ 2900 train 3.810467 (lr=1.2450e-03) (hash(x)=20786150)
353
+ 2910 train 3.590489 (lr=1.2339e-03) (hash(x)=20842137)
354
+ 2920 train 3.628046 (lr=1.2229e-03) (hash(x)=23939744)
355
+ 2930 train 3.710632 (lr=1.2119e-03) (hash(x)=20246627)
356
+ 2940 train 3.673259 (lr=1.2010e-03) (hash(x)=20386931)
357
+ 2950 train 3.606805 (lr=1.1901e-03) (hash(x)=19517612)
358
+ 2960 train 3.588694 (lr=1.1792e-03) (hash(x)=19643612)
359
+ 2970 train 3.514397 (lr=1.1684e-03) (hash(x)=17735841)
360
+ 2980 train 3.616668 (lr=1.1576e-03) (hash(x)=18179947)
361
+ 2990 train 3.517252 (lr=1.1468e-03) (hash(x)=19998132)
362
+ 3000 val loss 3.5881
363
+ 3000 val perplexity 36.1644
364
+ 3000 train 3.479401 (lr=1.1361e-03) (hash(x)=19400108)
365
+ 3010 train 3.459467 (lr=1.1254e-03) (hash(x)=19713177)
366
+ 3020 train 3.480014 (lr=1.1147e-03) (hash(x)=19871710)
367
+ 3030 train 3.442002 (lr=1.1041e-03) (hash(x)=16799044)
368
+ 3040 train 3.941165 (lr=1.0935e-03) (hash(x)=19431657)
369
+ 3050 train 3.708864 (lr=1.0830e-03) (hash(x)=22982891)
370
+ 3060 train 3.745234 (lr=1.0725e-03) (hash(x)=21562494)
371
+ 3070 train 3.625768 (lr=1.0620e-03) (hash(x)=33329781)
372
+ 3080 train 3.646884 (lr=1.0516e-03) (hash(x)=17296692)
373
+ 3090 train 3.659174 (lr=1.0412e-03) (hash(x)=19390072)
374
+ 3100 val loss 3.5723
375
+ 3100 val perplexity 35.5990
376
+ 3100 train 3.781618 (lr=1.0309e-03) (hash(x)=15455793)
377
+ 3110 train 3.643717 (lr=1.0207e-03) (hash(x)=20382524)
378
+ 3120 train 3.501483 (lr=1.0104e-03) (hash(x)=21248576)
379
+ 3130 train 3.541147 (lr=1.0002e-03) (hash(x)=20367412)
380
+ 3140 train 3.486457 (lr=9.9011e-04) (hash(x)=17826969)
381
+ 3150 train 3.572067 (lr=9.8002e-04) (hash(x)=21631872)
382
+ 3160 train 3.656660 (lr=9.6999e-04) (hash(x)=19506122)
383
+ 3170 train 3.696061 (lr=9.6000e-04) (hash(x)=21748750)
384
+ 3180 train 3.629033 (lr=9.5007e-04) (hash(x)=20805736)
385
+ 3190 train 3.673952 (lr=9.4019e-04) (hash(x)=19850067)
386
+ 3200 val loss 3.5512
387
+ 3200 val perplexity 34.8542
388
+ 3200 train 3.595512 (lr=9.3036e-04) (hash(x)=21173706)
389
+ 3210 train 3.602419 (lr=9.2058e-04) (hash(x)=21256910)
390
+ 3220 train 3.651968 (lr=9.1085e-04) (hash(x)=20765073)
391
+ 3230 train 3.617232 (lr=9.0118e-04) (hash(x)=17737503)
392
+ 3240 train 3.603663 (lr=8.9157e-04) (hash(x)=19598782)
393
+ 3250 train 3.573792 (lr=8.8201e-04) (hash(x)=24566640)
394
+ 3260 train 3.593848 (lr=8.7251e-04) (hash(x)=22002791)
395
+ 3270 train 3.484314 (lr=8.6307e-04) (hash(x)=20295322)
396
+ 3280 train 3.452937 (lr=8.5368e-04) (hash(x)=26247277)
397
+ 3290 train 3.483474 (lr=8.4435e-04) (hash(x)=20140665)
398
+ 3300 val loss 3.5402
399
+ 3300 val perplexity 34.4731
400
+ 3300 train 3.539965 (lr=8.3508e-04) (hash(x)=23553774)
401
+ 3310 train 3.650389 (lr=8.2588e-04) (hash(x)=17529719)
402
+ 3320 train 3.526107 (lr=8.1673e-04) (hash(x)=18795190)
403
+ 3330 train 3.626060 (lr=8.0764e-04) (hash(x)=20419570)
404
+ 3340 train 3.554729 (lr=7.9862e-04) (hash(x)=19596575)
405
+ 3350 train 3.567971 (lr=7.8966e-04) (hash(x)=17475331)
406
+ 3360 train 3.443928 (lr=7.8076e-04) (hash(x)=18856371)
407
+ 3370 train 3.619570 (lr=7.7192e-04) (hash(x)=19732920)
408
+ 3380 train 3.552999 (lr=7.6315e-04) (hash(x)=19151956)
409
+ 3390 train 3.523599 (lr=7.5445e-04) (hash(x)=22601715)
410
+ 3400 val loss 3.5239
411
+ 3400 val perplexity 33.9168
412
+ 3400 train 3.447089 (lr=7.4581e-04) (hash(x)=20446586)
413
+ 3410 train 3.608190 (lr=7.3724e-04) (hash(x)=27373387)
414
+ 3420 train 3.457381 (lr=7.2874e-04) (hash(x)=16894087)
415
+ 3430 train 3.477704 (lr=7.2030e-04) (hash(x)=19147767)
416
+ 3440 train 3.713491 (lr=7.1193e-04) (hash(x)=19394634)
417
+ 3450 train 3.565720 (lr=7.0363e-04) (hash(x)=20804137)
418
+ 3460 train 3.662098 (lr=6.9541e-04) (hash(x)=25730788)
419
+ 3470 train 3.525631 (lr=6.8725e-04) (hash(x)=18471384)
420
+ 3480 train 3.484756 (lr=6.7916e-04) (hash(x)=19884759)
421
+ 3490 train 3.529820 (lr=6.7114e-04) (hash(x)=21527738)
422
+ 3500 val loss 3.5072
423
+ 3500 val perplexity 33.3554
424
+ 3500 train 3.618383 (lr=6.6320e-04) (hash(x)=21108301)
425
+ 3510 train 3.524401 (lr=6.5533e-04) (hash(x)=19438613)
426
+ 3520 train 3.519203 (lr=6.4753e-04) (hash(x)=20391609)
427
+ 3530 train 3.418611 (lr=6.3981e-04) (hash(x)=21474987)
428
+ 3540 train 3.468692 (lr=6.3216e-04) (hash(x)=20544629)
429
+ 3550 train 3.525279 (lr=6.2458e-04) (hash(x)=18365326)
430
+ 3560 train 3.481968 (lr=6.1708e-04) (hash(x)=18443516)
431
+ 3570 train 3.525001 (lr=6.0966e-04) (hash(x)=18235257)
432
+ 3580 train 3.544608 (lr=6.0231e-04) (hash(x)=20958034)
433
+ 3590 train 3.527061 (lr=5.9504e-04) (hash(x)=20598224)
434
+ 3600 val loss 3.4946
435
+ 3600 val perplexity 32.9374
436
+ 3600 train 3.591177 (lr=5.8785e-04) (hash(x)=18613771)
437
+ 3610 train 3.679473 (lr=5.8074e-04) (hash(x)=19248168)
438
+ 3620 train 3.575057 (lr=5.7370e-04) (hash(x)=20935084)
439
+ 3630 train 3.679751 (lr=5.6675e-04) (hash(x)=20280969)
440
+ 3640 train 3.629335 (lr=5.5987e-04) (hash(x)=20735321)
441
+ 3650 train 3.515123 (lr=5.5308e-04) (hash(x)=20760704)
442
+ 3660 train 3.486288 (lr=5.4636e-04) (hash(x)=19656305)
443
+ 3670 train 3.486052 (lr=5.3973e-04) (hash(x)=19035247)
444
+ 3680 train 3.542181 (lr=5.3318e-04) (hash(x)=17935451)
445
+ 3690 train 3.502067 (lr=5.2671e-04) (hash(x)=22685971)
446
+ 3700 val loss 3.4879
447
+ 3700 val perplexity 32.7176
448
+ 3700 train 3.498343 (lr=5.2033e-04) (hash(x)=18466220)
449
+ 3710 train 3.359440 (lr=5.1402e-04) (hash(x)=19986945)
450
+ 3720 train 3.433963 (lr=5.0780e-04) (hash(x)=18017736)
451
+ 3730 train 3.479486 (lr=5.0167e-04) (hash(x)=19998802)
452
+ 3740 train 3.574218 (lr=4.9562e-04) (hash(x)=20145468)
453
+ 3750 train 3.485925 (lr=4.8965e-04) (hash(x)=16764686)
454
+ 3760 train 3.508090 (lr=4.8377e-04) (hash(x)=18833432)
455
+ 3770 train 3.435400 (lr=4.7798e-04) (hash(x)=16701754)
456
+ 3780 train 3.550174 (lr=4.7227e-04) (hash(x)=19608729)
457
+ 3790 train 3.584569 (lr=4.6665e-04) (hash(x)=19966203)
458
+ 3800 val loss 3.4676
459
+ 3800 val perplexity 32.0586
460
+ 3800 train 3.481163 (lr=4.6112e-04) (hash(x)=21276630)
461
+ 3810 train 3.589881 (lr=4.5567e-04) (hash(x)=19385436)
462
+ 3820 train 3.462910 (lr=4.5031e-04) (hash(x)=20894001)
463
+ 3830 train 3.430014 (lr=4.4504e-04) (hash(x)=22827106)
464
+ 3840 train 3.490487 (lr=4.3986e-04) (hash(x)=19263023)
465
+ 3850 train 3.474616 (lr=4.3477e-04) (hash(x)=19916874)
466
+ 3860 train 3.308574 (lr=4.2977e-04) (hash(x)=20568352)
467
+ 3870 train 3.264428 (lr=4.2486e-04) (hash(x)=24120742)
468
+ 3880 train 3.313142 (lr=4.2004e-04) (hash(x)=16266481)
469
+ 3890 train 3.388299 (lr=4.1530e-04) (hash(x)=19310691)
470
+ 3900 val loss 3.4718
471
+ 3900 val perplexity 32.1935
472
+ 3900 train 3.248407 (lr=4.1066e-04) (hash(x)=17735115)
473
+ 3910 train 3.457561 (lr=4.0611e-04) (hash(x)=18175918)
474
+ 3920 train 3.450331 (lr=4.0166e-04) (hash(x)=21662789)
475
+ 3930 train 3.485642 (lr=3.9729e-04) (hash(x)=21034272)
476
+ 3940 train 3.484607 (lr=3.9302e-04) (hash(x)=19724980)
477
+ 3950 train 3.467769 (lr=3.8884e-04) (hash(x)=20066944)
478
+ 3960 train 3.573454 (lr=3.8475e-04) (hash(x)=22065490)
479
+ 3970 train 3.488898 (lr=3.8076e-04) (hash(x)=20065365)
480
+ 3980 train 3.561086 (lr=3.7685e-04) (hash(x)=21419563)
481
+ 3990 train 3.547145 (lr=3.7305e-04) (hash(x)=21180852)
482
+ 4000 val loss 3.4512
483
+ 4000 val perplexity 31.5378
484
+ 4000 train 3.522856 (lr=3.6933e-04) (hash(x)=18548544)
485
+ 4010 train 3.518970 (lr=3.6572e-04) (hash(x)=20609113)
486
+ 4020 train 3.406380 (lr=3.6219e-04) (hash(x)=16977034)
487
+ 4030 train 3.438291 (lr=3.5876e-04) (hash(x)=19248645)
488
+ 4040 train 3.141098 (lr=3.5543e-04) (hash(x)=16881719)
489
+ 4050 train 3.213471 (lr=3.5219e-04) (hash(x)=18497713)
490
+ 4060 train 3.156633 (lr=3.4905e-04) (hash(x)=20036068)
491
+ 4070 train 3.132728 (lr=3.4600e-04) (hash(x)=21636163)
492
+ 4080 train 3.336069 (lr=3.4305e-04) (hash(x)=23072887)
493
+ 4090 train 3.451056 (lr=3.4019e-04) (hash(x)=27646813)
494
+ 4100 val loss 3.4488
495
+ 4100 val perplexity 31.4639
496
+ 4100 train 3.590968 (lr=3.3744e-04) (hash(x)=20052100)
497
+ 4110 train 3.493044 (lr=3.3477e-04) (hash(x)=21623032)
498
+ 4120 train 3.517746 (lr=3.3221e-04) (hash(x)=17755383)
499
+ 4130 train 3.515046 (lr=3.2974e-04) (hash(x)=20577475)
500
+ 4140 train 3.518400 (lr=3.2737e-04) (hash(x)=17308418)
501
+ 4150 train 3.563785 (lr=3.2510e-04) (hash(x)=20786408)
502
+ 4160 train 3.494988 (lr=3.2292e-04) (hash(x)=20006893)
503
+ 4170 train 3.459144 (lr=3.2085e-04) (hash(x)=20283286)
504
+ 4180 train 3.458465 (lr=3.1887e-04) (hash(x)=17799852)
505
+ 4190 train 3.385463 (lr=3.1699e-04) (hash(x)=20291468)
506
+ 4200 val loss 3.4391
507
+ 4200 val perplexity 31.1594
508
+ 4200 train 3.438119 (lr=3.1520e-04) (hash(x)=16871586)
509
+ 4210 train 3.507307 (lr=3.1352e-04) (hash(x)=21608030)
510
+ 4220 train 3.229509 (lr=3.1193e-04) (hash(x)=22256819)
511
+ 4230 train 3.299121 (lr=3.1044e-04) (hash(x)=20929071)
512
+ 4240 train 3.315285 (lr=3.0905e-04) (hash(x)=22408188)
513
+ 4250 train 3.223896 (lr=3.0776e-04) (hash(x)=19862925)
514
+ 4260 train 3.164011 (lr=3.0657e-04) (hash(x)=17765583)
515
+ 4270 train 3.435705 (lr=3.0548e-04) (hash(x)=21260687)
516
+ 4280 train 3.384398 (lr=3.0449e-04) (hash(x)=20333361)
517
+ 4290 train 3.438148 (lr=3.0359e-04) (hash(x)=21242172)
518
+ 4300 val loss 3.4327
519
+ 4300 val perplexity 30.9591
520
+ 4300 train 3.397286 (lr=3.0280e-04) (hash(x)=19835810)
521
+ 4310 train 3.427557 (lr=3.0210e-04) (hash(x)=32580183)
522
+ 4320 train 3.454242 (lr=3.0150e-04) (hash(x)=17944217)
523
+ 4330 train 3.449493 (lr=3.0101e-04) (hash(x)=20116568)
524
+ 4340 train 3.447457 (lr=3.0061e-04) (hash(x)=22423070)
525
+ 4350 train 3.505254 (lr=3.0031e-04) (hash(x)=15493166)
526
+ 4360 train 3.414657 (lr=3.0011e-04) (hash(x)=20130986)
527
+ 4370 train 3.497459 (lr=3.0001e-04) (hash(x)=20037228)
528
+ 4374 val loss 3.4310
529
+ 4374 val perplexity 30.9072
logs/64_latent_masks_degenerate/degen_compile_seed_1340/model_04374.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:3164e4067f767971417b5eb6f5ff6444e15a6010ff44786eaa59e18910a5999c
3
+ size 555244890
logs/64_latent_masks_degenerate/degen_compile_seed_1340/optimizer_04374.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:e30b0da01e05db73e5204220e40ce3e2ece0c981cab85d646fe82ac9d3d6bb0f
3
+ size 1009837598