andrew-healey commited on
Commit
bd75838
·
verified ·
1 Parent(s): e58c6e9

Upload folder using huggingface_hub

Browse files
logs/64_head_dim_two_latent_masks/1_latent_mask_seed_1340/args.json ADDED
@@ -0,0 +1 @@
 
 
1
+ {"hellaswag": true, "attention_kind": "selective", "log_dir": "logs/64_head_dim_two_latent_masks/1_latent_mask_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_head_dim_two_latent_masks", "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": 1.0, "disable_selection_head_linear_combo_bias": false, "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": false, "use_mini_model": false, "upload_to_hf": true, "seq_len": 256, "batch_size": 16, "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": "1_latent_mask", "latent_mask_precision": "float32"}
logs/64_head_dim_two_latent_masks/1_latent_mask_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_head_dim_two_latent_masks/1_latent_mask_seed_1340/log2.txt ADDED
@@ -0,0 +1,548 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ max_steps: 4375
2
+ 0 val loss 11.0314
3
+ 0 val perplexity 61783.7656
4
+ 0 hella 0.2483
5
+ 0 train 11.035601 (lr=4.1958e-06) (hash(x)=19978355)
6
+ 10 train 9.671946 (lr=4.6154e-05) (hash(x)=24395949)
7
+ 20 train 9.182615 (lr=8.8112e-05) (hash(x)=20203571)
8
+ 30 train 8.475595 (lr=1.3007e-04) (hash(x)=22227211)
9
+ 40 train 7.818489 (lr=1.7203e-04) (hash(x)=16492366)
10
+ 50 train 7.594036 (lr=2.1399e-04) (hash(x)=17690059)
11
+ 60 train 7.522951 (lr=2.5594e-04) (hash(x)=19389177)
12
+ 70 train 7.521806 (lr=2.9790e-04) (hash(x)=16938017)
13
+ 80 train 7.339411 (lr=3.3986e-04) (hash(x)=18013370)
14
+ 90 train 7.103736 (lr=3.8182e-04) (hash(x)=22273301)
15
+ 100 val loss 6.9995
16
+ 100 val perplexity 1096.0629
17
+ 100 train 6.967471 (lr=4.2378e-04) (hash(x)=21867969)
18
+ 110 train 6.881469 (lr=4.6573e-04) (hash(x)=17559385)
19
+ 120 train 6.774395 (lr=5.0769e-04) (hash(x)=21415593)
20
+ 130 train 6.623225 (lr=5.4965e-04) (hash(x)=17493363)
21
+ 140 train 6.496814 (lr=5.9161e-04) (hash(x)=19060392)
22
+ 150 train 6.443040 (lr=6.3357e-04) (hash(x)=18669634)
23
+ 160 train 6.309308 (lr=6.7552e-04) (hash(x)=19128903)
24
+ 170 train 6.386444 (lr=7.1748e-04) (hash(x)=21173132)
25
+ 180 train 6.363103 (lr=7.5944e-04) (hash(x)=21482016)
26
+ 190 train 6.116847 (lr=8.0140e-04) (hash(x)=19274288)
27
+ 200 val loss 6.1025
28
+ 200 val perplexity 446.9810
29
+ 200 train 6.058624 (lr=8.4336e-04) (hash(x)=17153997)
30
+ 210 train 5.969953 (lr=8.8531e-04) (hash(x)=18659881)
31
+ 220 train 5.909371 (lr=9.2727e-04) (hash(x)=13041762)
32
+ 230 train 5.917233 (lr=9.6923e-04) (hash(x)=19089409)
33
+ 240 train 5.966206 (lr=1.0112e-03) (hash(x)=21588907)
34
+ 250 hella 0.2379
35
+ 250 train 5.923897 (lr=1.0531e-03) (hash(x)=22853384)
36
+ 260 train 5.849435 (lr=1.0951e-03) (hash(x)=21258918)
37
+ 270 train 5.706386 (lr=1.1371e-03) (hash(x)=18064675)
38
+ 280 train 5.680332 (lr=1.1790e-03) (hash(x)=17783654)
39
+ 290 train 5.615935 (lr=1.2210e-03) (hash(x)=16206303)
40
+ 300 val loss 5.6075
41
+ 300 val perplexity 272.4563
42
+ 300 train 5.621495 (lr=1.2629e-03) (hash(x)=23548851)
43
+ 310 train 5.468264 (lr=1.3049e-03) (hash(x)=17695462)
44
+ 320 train 5.542215 (lr=1.3469e-03) (hash(x)=16887671)
45
+ 330 train 5.451490 (lr=1.3888e-03) (hash(x)=19956923)
46
+ 340 train 5.452887 (lr=1.4308e-03) (hash(x)=16428364)
47
+ 350 train 5.358995 (lr=1.4727e-03) (hash(x)=20437634)
48
+ 360 train 5.310882 (lr=1.5147e-03) (hash(x)=19622578)
49
+ 370 train 5.333323 (lr=1.5566e-03) (hash(x)=21294724)
50
+ 380 train 5.306688 (lr=1.5986e-03) (hash(x)=18869667)
51
+ 390 train 5.292197 (lr=1.6406e-03) (hash(x)=22256474)
52
+ 400 val loss 5.2675
53
+ 400 val perplexity 193.9396
54
+ 400 train 5.340266 (lr=1.6825e-03) (hash(x)=20048534)
55
+ 410 train 5.260165 (lr=1.7245e-03) (hash(x)=23042068)
56
+ 420 train 5.254843 (lr=1.7664e-03) (hash(x)=19937603)
57
+ 430 train 5.233081 (lr=1.8084e-03) (hash(x)=22150243)
58
+ 440 train 5.228151 (lr=1.8503e-03) (hash(x)=23468794)
59
+ 450 train 5.147303 (lr=1.8923e-03) (hash(x)=18432856)
60
+ 460 train 5.246174 (lr=1.9343e-03) (hash(x)=21494081)
61
+ 470 train 5.108386 (lr=1.9762e-03) (hash(x)=18494966)
62
+ 480 train 5.022178 (lr=2.0182e-03) (hash(x)=20099891)
63
+ 490 train 5.113085 (lr=2.0601e-03) (hash(x)=21778663)
64
+ 500 val loss 5.0512
65
+ 500 val perplexity 156.2052
66
+ 500 hella 0.2441
67
+ 500 train 4.978273 (lr=2.1021e-03) (hash(x)=21164498)
68
+ 510 train 4.968841 (lr=2.1441e-03) (hash(x)=20617181)
69
+ 520 train 5.043192 (lr=2.1860e-03) (hash(x)=24124812)
70
+ 530 train 4.993603 (lr=2.2280e-03) (hash(x)=18270550)
71
+ 540 train 5.001164 (lr=2.2699e-03) (hash(x)=17643805)
72
+ 550 train 4.865990 (lr=2.3119e-03) (hash(x)=19633450)
73
+ 560 train 4.909146 (lr=2.3538e-03) (hash(x)=18215455)
74
+ 570 train 4.997000 (lr=2.3958e-03) (hash(x)=23240966)
75
+ 580 train 4.869910 (lr=2.4378e-03) (hash(x)=20144929)
76
+ 590 train 4.892120 (lr=2.4797e-03) (hash(x)=16296841)
77
+ 600 val loss 4.8018
78
+ 600 val perplexity 121.7302
79
+ 600 train 4.836444 (lr=2.5217e-03) (hash(x)=19836338)
80
+ 610 train 5.171877 (lr=2.5636e-03) (hash(x)=19928421)
81
+ 620 train 4.714991 (lr=2.6056e-03) (hash(x)=17727783)
82
+ 630 train 4.793374 (lr=2.6476e-03) (hash(x)=20940923)
83
+ 640 train 4.733313 (lr=2.6895e-03) (hash(x)=29374503)
84
+ 650 train 4.739740 (lr=2.7315e-03) (hash(x)=18000892)
85
+ 660 train 4.760598 (lr=2.7734e-03) (hash(x)=20632048)
86
+ 670 train 4.755336 (lr=2.8154e-03) (hash(x)=16874750)
87
+ 680 train 4.849604 (lr=2.8573e-03) (hash(x)=16996745)
88
+ 690 train 4.666236 (lr=2.8993e-03) (hash(x)=17242267)
89
+ 700 val loss 4.6437
90
+ 700 val perplexity 103.9230
91
+ 700 train 4.545478 (lr=2.9413e-03) (hash(x)=4532237)
92
+ 710 train 4.799165 (lr=2.9832e-03) (hash(x)=18455888)
93
+ 720 train 4.604057 (lr=3.0000e-03) (hash(x)=22389104)
94
+ 730 train 4.603886 (lr=2.9999e-03) (hash(x)=19048232)
95
+ 740 train 4.553998 (lr=2.9997e-03) (hash(x)=21823621)
96
+ 750 hella 0.2491
97
+ 750 train 4.577466 (lr=2.9994e-03) (hash(x)=19834992)
98
+ 760 train 4.482804 (lr=2.9990e-03) (hash(x)=20804217)
99
+ 770 train 4.379577 (lr=2.9985e-03) (hash(x)=21676803)
100
+ 780 train 4.334875 (lr=2.9979e-03) (hash(x)=20436543)
101
+ 790 train 4.386044 (lr=2.9972e-03) (hash(x)=24049409)
102
+ 800 val loss 4.5491
103
+ 800 val perplexity 94.5445
104
+ 800 train 4.295110 (lr=2.9964e-03) (hash(x)=21101828)
105
+ 810 train 4.477986 (lr=2.9955e-03) (hash(x)=21158051)
106
+ 820 train 4.435212 (lr=2.9945e-03) (hash(x)=17068830)
107
+ 830 train 4.486201 (lr=2.9934e-03) (hash(x)=20708816)
108
+ 840 train 4.529250 (lr=2.9922e-03) (hash(x)=23031486)
109
+ 850 train 4.555763 (lr=2.9909e-03) (hash(x)=17808332)
110
+ 860 train 4.422565 (lr=2.9896e-03) (hash(x)=19935633)
111
+ 870 train 4.467145 (lr=2.9881e-03) (hash(x)=21546448)
112
+ 880 train 4.469337 (lr=2.9865e-03) (hash(x)=18928060)
113
+ 890 train 4.366578 (lr=2.9848e-03) (hash(x)=19620142)
114
+ 900 val loss 4.3952
115
+ 900 val perplexity 81.0575
116
+ 900 train 4.390852 (lr=2.9830e-03) (hash(x)=15281589)
117
+ 910 train 4.331115 (lr=2.9811e-03) (hash(x)=19313761)
118
+ 920 train 4.347125 (lr=2.9792e-03) (hash(x)=21534098)
119
+ 930 train 4.355960 (lr=2.9771e-03) (hash(x)=22007834)
120
+ 940 train 4.397508 (lr=2.9749e-03) (hash(x)=19528278)
121
+ 950 train 4.359200 (lr=2.9726e-03) (hash(x)=18826968)
122
+ 960 train 4.360198 (lr=2.9703e-03) (hash(x)=24147820)
123
+ 970 train 4.170729 (lr=2.9678e-03) (hash(x)=22646220)
124
+ 980 train 4.169567 (lr=2.9652e-03) (hash(x)=21026041)
125
+ 990 train 4.170161 (lr=2.9626e-03) (hash(x)=17067127)
126
+ 1000 val loss 4.3497
127
+ 1000 val perplexity 77.4522
128
+ 1000 hella 0.2526
129
+ 1000 train 4.512960 (lr=2.9598e-03) (hash(x)=18557650)
130
+ 1010 train 4.399590 (lr=2.9570e-03) (hash(x)=20276417)
131
+ 1020 train 4.438538 (lr=2.9540e-03) (hash(x)=25973944)
132
+ 1030 train 4.434834 (lr=2.9510e-03) (hash(x)=21253298)
133
+ 1040 train 4.285232 (lr=2.9478e-03) (hash(x)=24702365)
134
+ 1050 train 4.365864 (lr=2.9446e-03) (hash(x)=19691634)
135
+ 1060 train 4.276920 (lr=2.9412e-03) (hash(x)=16659615)
136
+ 1070 train 4.244260 (lr=2.9378e-03) (hash(x)=18095563)
137
+ 1080 train 4.297185 (lr=2.9343e-03) (hash(x)=18443383)
138
+ 1090 train 4.285233 (lr=2.9307e-03) (hash(x)=18330349)
139
+ 1100 val loss 4.2635
140
+ 1100 val perplexity 71.0549
141
+ 1100 train 4.235328 (lr=2.9270e-03) (hash(x)=19031432)
142
+ 1110 train 4.423933 (lr=2.9231e-03) (hash(x)=27724663)
143
+ 1120 train 4.194479 (lr=2.9192e-03) (hash(x)=20063068)
144
+ 1130 train 4.126320 (lr=2.9152e-03) (hash(x)=18456567)
145
+ 1140 train 4.143937 (lr=2.9112e-03) (hash(x)=21274378)
146
+ 1150 train 4.431369 (lr=2.9070e-03) (hash(x)=20166104)
147
+ 1160 train 3.970883 (lr=2.9027e-03) (hash(x)=19016072)
148
+ 1170 train 4.013882 (lr=2.8983e-03) (hash(x)=20983879)
149
+ 1180 train 3.998129 (lr=2.8939e-03) (hash(x)=22773283)
150
+ 1190 train 4.232429 (lr=2.8893e-03) (hash(x)=19469721)
151
+ 1200 val loss 4.2222
152
+ 1200 val perplexity 68.1828
153
+ 1200 train 4.256883 (lr=2.8847e-03) (hash(x)=21239510)
154
+ 1210 train 4.174796 (lr=2.8800e-03) (hash(x)=21515094)
155
+ 1220 train 4.154643 (lr=2.8751e-03) (hash(x)=16618399)
156
+ 1230 train 4.275473 (lr=2.8702e-03) (hash(x)=19402906)
157
+ 1240 train 4.325212 (lr=2.8652e-03) (hash(x)=21224237)
158
+ 1250 hella 0.2564
159
+ 1250 train 4.160768 (lr=2.8601e-03) (hash(x)=16261179)
160
+ 1260 train 4.249974 (lr=2.8550e-03) (hash(x)=19045671)
161
+ 1270 train 4.170993 (lr=2.8497e-03) (hash(x)=13880880)
162
+ 1280 train 4.299049 (lr=2.8443e-03) (hash(x)=22126650)
163
+ 1290 train 4.089890 (lr=2.8389e-03) (hash(x)=22005339)
164
+ 1300 val loss 4.1462
165
+ 1300 val perplexity 63.1949
166
+ 1300 train 4.146296 (lr=2.8333e-03) (hash(x)=22293195)
167
+ 1310 train 4.200065 (lr=2.8277e-03) (hash(x)=19908750)
168
+ 1320 train 4.039923 (lr=2.8220e-03) (hash(x)=24199231)
169
+ 1330 train 4.033680 (lr=2.8162e-03) (hash(x)=20450695)
170
+ 1340 train 3.981503 (lr=2.8103e-03) (hash(x)=17483865)
171
+ 1350 train 3.984835 (lr=2.8044e-03) (hash(x)=20351158)
172
+ 1360 train 3.969366 (lr=2.7983e-03) (hash(x)=23798487)
173
+ 1370 train 3.922527 (lr=2.7922e-03) (hash(x)=20090485)
174
+ 1380 train 4.083005 (lr=2.7860e-03) (hash(x)=24781076)
175
+ 1390 train 4.145727 (lr=2.7797e-03) (hash(x)=21234884)
176
+ 1400 val loss 4.0988
177
+ 1400 val perplexity 60.2686
178
+ 1400 train 4.062721 (lr=2.7733e-03) (hash(x)=22226319)
179
+ 1410 train 4.336025 (lr=2.7668e-03) (hash(x)=25309506)
180
+ 1420 train 4.193780 (lr=2.7603e-03) (hash(x)=22351162)
181
+ 1430 train 4.036261 (lr=2.7536e-03) (hash(x)=20164493)
182
+ 1440 train 4.099714 (lr=2.7469e-03) (hash(x)=16571343)
183
+ 1450 train 4.188511 (lr=2.7401e-03) (hash(x)=17493318)
184
+ 1460 train 4.171222 (lr=2.7333e-03) (hash(x)=19324896)
185
+ 1470 train 4.082887 (lr=2.7263e-03) (hash(x)=19804956)
186
+ 1480 train 4.120047 (lr=2.7193e-03) (hash(x)=19628407)
187
+ 1490 train 4.059060 (lr=2.7121e-03) (hash(x)=19770129)
188
+ 1500 val loss 4.0766
189
+ 1500 val perplexity 58.9452
190
+ 1500 hella 0.2546
191
+ 1500 train 4.036770 (lr=2.7050e-03) (hash(x)=16945773)
192
+ 1510 train 4.097395 (lr=2.6977e-03) (hash(x)=18780367)
193
+ 1520 train 3.892389 (lr=2.6903e-03) (hash(x)=19997660)
194
+ 1530 train 4.159437 (lr=2.6829e-03) (hash(x)=17063727)
195
+ 1540 train 4.077147 (lr=2.6754e-03) (hash(x)=20300139)
196
+ 1550 train 3.970974 (lr=2.6678e-03) (hash(x)=21541648)
197
+ 1560 train 4.039214 (lr=2.6602e-03) (hash(x)=25169762)
198
+ 1570 train 4.009277 (lr=2.6525e-03) (hash(x)=18020155)
199
+ 1580 train 3.985390 (lr=2.6447e-03) (hash(x)=21439415)
200
+ 1590 train 3.975525 (lr=2.6368e-03) (hash(x)=20345952)
201
+ 1600 val loss 4.0339
202
+ 1600 val perplexity 56.4831
203
+ 1600 train 3.892110 (lr=2.6289e-03) (hash(x)=19677331)
204
+ 1610 train 4.011547 (lr=2.6208e-03) (hash(x)=23142852)
205
+ 1620 train 4.012848 (lr=2.6128e-03) (hash(x)=18712875)
206
+ 1630 train 3.871821 (lr=2.6046e-03) (hash(x)=17940151)
207
+ 1640 train 3.847362 (lr=2.5964e-03) (hash(x)=18928868)
208
+ 1650 train 4.016849 (lr=2.5881e-03) (hash(x)=21562811)
209
+ 1660 train 4.088722 (lr=2.5797e-03) (hash(x)=19917546)
210
+ 1670 train 4.107987 (lr=2.5713e-03) (hash(x)=19862190)
211
+ 1680 train 4.079481 (lr=2.5628e-03) (hash(x)=19510122)
212
+ 1690 train 3.955552 (lr=2.5542e-03) (hash(x)=18819854)
213
+ 1700 val loss 4.0008
214
+ 1700 val perplexity 54.6405
215
+ 1700 train 3.939885 (lr=2.5455e-03) (hash(x)=20098069)
216
+ 1710 train 3.946827 (lr=2.5368e-03) (hash(x)=19180845)
217
+ 1720 train 3.981010 (lr=2.5281e-03) (hash(x)=22610840)
218
+ 1730 train 3.995194 (lr=2.5192e-03) (hash(x)=17804171)
219
+ 1740 train 3.953915 (lr=2.5103e-03) (hash(x)=18275747)
220
+ 1750 hella 0.2569
221
+ 1750 train 3.883433 (lr=2.5014e-03) (hash(x)=20884937)
222
+ 1760 train 3.894905 (lr=2.4924e-03) (hash(x)=18783179)
223
+ 1770 train 3.947295 (lr=2.4833e-03) (hash(x)=20225710)
224
+ 1780 train 3.877046 (lr=2.4741e-03) (hash(x)=19910233)
225
+ 1790 train 4.011127 (lr=2.4649e-03) (hash(x)=19722753)
226
+ 1800 val loss 3.9818
227
+ 1800 val perplexity 53.6158
228
+ 1800 train 3.979106 (lr=2.4556e-03) (hash(x)=21871844)
229
+ 1810 train 3.997065 (lr=2.4463e-03) (hash(x)=19497156)
230
+ 1820 train 4.034789 (lr=2.4369e-03) (hash(x)=18788042)
231
+ 1830 train 3.969046 (lr=2.4275e-03) (hash(x)=19961057)
232
+ 1840 train 3.914623 (lr=2.4180e-03) (hash(x)=17863548)
233
+ 1850 train 4.066847 (lr=2.4084e-03) (hash(x)=17751881)
234
+ 1860 train 3.905634 (lr=2.3988e-03) (hash(x)=19219734)
235
+ 1870 train 3.878024 (lr=2.3891e-03) (hash(x)=19455814)
236
+ 1880 train 3.928726 (lr=2.3794e-03) (hash(x)=20580620)
237
+ 1890 train 3.909274 (lr=2.3696e-03) (hash(x)=17077823)
238
+ 1900 val loss 3.9683
239
+ 1900 val perplexity 52.8964
240
+ 1900 train 3.926562 (lr=2.3598e-03) (hash(x)=17322645)
241
+ 1910 train 3.838995 (lr=2.3499e-03) (hash(x)=20740743)
242
+ 1920 train 4.030829 (lr=2.3400e-03) (hash(x)=20599034)
243
+ 1930 train 3.982111 (lr=2.3300e-03) (hash(x)=21358394)
244
+ 1940 train 4.033349 (lr=2.3200e-03) (hash(x)=19899662)
245
+ 1950 train 3.967762 (lr=2.3099e-03) (hash(x)=15623618)
246
+ 1960 train 3.848743 (lr=2.2998e-03) (hash(x)=19094692)
247
+ 1970 train 3.922780 (lr=2.2896e-03) (hash(x)=21521915)
248
+ 1980 train 3.870940 (lr=2.2793e-03) (hash(x)=18396717)
249
+ 1990 train 3.954028 (lr=2.2691e-03) (hash(x)=20176898)
250
+ 2000 val loss 3.9274
251
+ 2000 val perplexity 50.7759
252
+ 2000 hella 0.2626
253
+ 2000 train 3.873100 (lr=2.2588e-03) (hash(x)=17939830)
254
+ 2010 train 3.921104 (lr=2.2484e-03) (hash(x)=18313312)
255
+ 2020 train 3.822767 (lr=2.2380e-03) (hash(x)=22132416)
256
+ 2030 train 3.864434 (lr=2.2275e-03) (hash(x)=18403620)
257
+ 2040 train 3.775062 (lr=2.2170e-03) (hash(x)=19852232)
258
+ 2050 train 3.771405 (lr=2.2065e-03) (hash(x)=21010646)
259
+ 2060 train 3.964508 (lr=2.1959e-03) (hash(x)=21119229)
260
+ 2070 train 4.036797 (lr=2.1853e-03) (hash(x)=22447456)
261
+ 2080 train 3.917071 (lr=2.1746e-03) (hash(x)=19656298)
262
+ 2090 train 3.973149 (lr=2.1639e-03) (hash(x)=17300349)
263
+ 2100 val loss 3.9127
264
+ 2100 val perplexity 50.0359
265
+ 2100 train 3.861718 (lr=2.1532e-03) (hash(x)=19489545)
266
+ 2110 train 3.790474 (lr=2.1424e-03) (hash(x)=18592164)
267
+ 2120 train 3.874874 (lr=2.1316e-03) (hash(x)=18607056)
268
+ 2130 train 3.878049 (lr=2.1208e-03) (hash(x)=18652918)
269
+ 2140 train 3.911024 (lr=2.1099e-03) (hash(x)=21040103)
270
+ 2150 train 3.793377 (lr=2.0990e-03) (hash(x)=16399764)
271
+ 2160 train 3.828374 (lr=2.0881e-03) (hash(x)=17535477)
272
+ 2170 train 3.749187 (lr=2.0771e-03) (hash(x)=19890415)
273
+ 2180 train 3.886533 (lr=2.0661e-03) (hash(x)=16652534)
274
+ 2190 train 3.809416 (lr=2.0550e-03) (hash(x)=22131443)
275
+ 2200 val loss 3.8914
276
+ 2200 val perplexity 48.9786
277
+ 2200 train 3.928072 (lr=2.0440e-03) (hash(x)=19148364)
278
+ 2210 train 4.026750 (lr=2.0329e-03) (hash(x)=15715527)
279
+ 2220 train 3.858968 (lr=2.0217e-03) (hash(x)=23414699)
280
+ 2230 train 3.914368 (lr=2.0106e-03) (hash(x)=19689870)
281
+ 2240 train 3.939881 (lr=1.9994e-03) (hash(x)=22678814)
282
+ 2250 hella 0.2589
283
+ 2250 train 3.799642 (lr=1.9882e-03) (hash(x)=18976933)
284
+ 2260 train 3.893148 (lr=1.9770e-03) (hash(x)=20397863)
285
+ 2270 train 3.805707 (lr=1.9657e-03) (hash(x)=19308365)
286
+ 2280 train 3.826259 (lr=1.9544e-03) (hash(x)=18078666)
287
+ 2290 train 3.855768 (lr=1.9431e-03) (hash(x)=20419667)
288
+ 2300 val loss 3.8591
289
+ 2300 val perplexity 47.4241
290
+ 2300 train 3.825915 (lr=1.9318e-03) (hash(x)=16788415)
291
+ 2310 train 3.764396 (lr=1.9205e-03) (hash(x)=18566522)
292
+ 2320 train 3.843877 (lr=1.9091e-03) (hash(x)=21387107)
293
+ 2330 train 3.975788 (lr=1.8977e-03) (hash(x)=16815344)
294
+ 2340 train 3.954148 (lr=1.8863e-03) (hash(x)=19301309)
295
+ 2350 train 3.887287 (lr=1.8749e-03) (hash(x)=18411578)
296
+ 2360 train 4.061715 (lr=1.8635e-03) (hash(x)=17212565)
297
+ 2370 train 3.866417 (lr=1.8520e-03) (hash(x)=19083591)
298
+ 2380 train 3.775690 (lr=1.8406e-03) (hash(x)=19095330)
299
+ 2390 train 3.842422 (lr=1.8291e-03) (hash(x)=22137679)
300
+ 2400 val loss 3.8401
301
+ 2400 val perplexity 46.5307
302
+ 2400 train 3.822761 (lr=1.8176e-03) (hash(x)=19988438)
303
+ 2410 train 3.908319 (lr=1.8061e-03) (hash(x)=18845683)
304
+ 2420 train 3.842633 (lr=1.7946e-03) (hash(x)=18942135)
305
+ 2430 train 3.744929 (lr=1.7830e-03) (hash(x)=19844600)
306
+ 2440 train 3.723804 (lr=1.7715e-03) (hash(x)=20037517)
307
+ 2450 train 3.701990 (lr=1.7600e-03) (hash(x)=18071985)
308
+ 2460 train 3.729635 (lr=1.7484e-03) (hash(x)=19498889)
309
+ 2470 train 3.975096 (lr=1.7368e-03) (hash(x)=21020015)
310
+ 2480 train 3.999651 (lr=1.7253e-03) (hash(x)=24541400)
311
+ 2490 train 3.890455 (lr=1.7137e-03) (hash(x)=17117268)
312
+ 2500 val loss 3.8184
313
+ 2500 val perplexity 45.5294
314
+ 2500 hella 0.2605
315
+ 2500 train 3.828212 (lr=1.7021e-03) (hash(x)=20019265)
316
+ 2510 train 3.922086 (lr=1.6906e-03) (hash(x)=15424606)
317
+ 2520 train 3.879414 (lr=1.6790e-03) (hash(x)=16589718)
318
+ 2530 train 3.785809 (lr=1.6674e-03) (hash(x)=21793520)
319
+ 2540 train 3.758087 (lr=1.6558e-03) (hash(x)=26991001)
320
+ 2550 train 3.703202 (lr=1.6442e-03) (hash(x)=19542562)
321
+ 2560 train 3.808804 (lr=1.6326e-03) (hash(x)=16342890)
322
+ 2570 train 3.678167 (lr=1.6210e-03) (hash(x)=22032122)
323
+ 2580 train 3.673751 (lr=1.6094e-03) (hash(x)=20442648)
324
+ 2590 train 3.632080 (lr=1.5979e-03) (hash(x)=20229881)
325
+ 2600 val loss 3.8010
326
+ 2600 val perplexity 44.7443
327
+ 2600 train 3.716985 (lr=1.5863e-03) (hash(x)=20759842)
328
+ 2610 train 3.662948 (lr=1.5747e-03) (hash(x)=18590738)
329
+ 2620 train 3.874747 (lr=1.5632e-03) (hash(x)=20016008)
330
+ 2630 train 3.821032 (lr=1.5516e-03) (hash(x)=20966006)
331
+ 2640 train 3.915936 (lr=1.5400e-03) (hash(x)=20480389)
332
+ 2650 train 3.749489 (lr=1.5285e-03) (hash(x)=17625327)
333
+ 2660 train 3.726404 (lr=1.5170e-03) (hash(x)=17462746)
334
+ 2670 train 3.809113 (lr=1.5054e-03) (hash(x)=24540619)
335
+ 2680 train 3.738880 (lr=1.4939e-03) (hash(x)=18300466)
336
+ 2690 train 3.672699 (lr=1.4824e-03) (hash(x)=20785354)
337
+ 2700 val loss 3.7790
338
+ 2700 val perplexity 43.7715
339
+ 2700 train 3.782812 (lr=1.4709e-03) (hash(x)=19689301)
340
+ 2710 train 3.687641 (lr=1.4594e-03) (hash(x)=18041572)
341
+ 2720 train 3.719274 (lr=1.4480e-03) (hash(x)=23202500)
342
+ 2730 train 3.686029 (lr=1.4365e-03) (hash(x)=25706624)
343
+ 2740 train 3.691536 (lr=1.4251e-03) (hash(x)=24172973)
344
+ 2750 hella 0.2643
345
+ 2750 train 3.728926 (lr=1.4137e-03) (hash(x)=22379491)
346
+ 2760 train 3.779565 (lr=1.4023e-03) (hash(x)=21701022)
347
+ 2770 train 3.826560 (lr=1.3909e-03) (hash(x)=21951124)
348
+ 2780 train 3.757556 (lr=1.3795e-03) (hash(x)=19525099)
349
+ 2790 train 3.802535 (lr=1.3682e-03) (hash(x)=18035060)
350
+ 2800 val loss 3.7565
351
+ 2800 val perplexity 42.7985
352
+ 2800 train 3.789171 (lr=1.3569e-03) (hash(x)=19475507)
353
+ 2810 train 3.737863 (lr=1.3456e-03) (hash(x)=20631788)
354
+ 2820 train 3.748224 (lr=1.3343e-03) (hash(x)=21229388)
355
+ 2830 train 3.756998 (lr=1.3230e-03) (hash(x)=17707440)
356
+ 2840 train 3.769183 (lr=1.3118e-03) (hash(x)=17822842)
357
+ 2850 train 3.727228 (lr=1.3006e-03) (hash(x)=20668384)
358
+ 2860 train 3.687619 (lr=1.2894e-03) (hash(x)=21880436)
359
+ 2870 train 3.698210 (lr=1.2783e-03) (hash(x)=22236229)
360
+ 2880 train 3.628624 (lr=1.2671e-03) (hash(x)=22480364)
361
+ 2890 train 3.589179 (lr=1.2560e-03) (hash(x)=20782341)
362
+ 2900 val loss 3.7521
363
+ 2900 val perplexity 42.6105
364
+ 2900 train 3.941689 (lr=1.2450e-03) (hash(x)=20786150)
365
+ 2910 train 3.728560 (lr=1.2339e-03) (hash(x)=20842137)
366
+ 2920 train 3.750448 (lr=1.2229e-03) (hash(x)=23939744)
367
+ 2930 train 3.823665 (lr=1.2119e-03) (hash(x)=20246627)
368
+ 2940 train 3.798024 (lr=1.2010e-03) (hash(x)=20386931)
369
+ 2950 train 3.740004 (lr=1.1901e-03) (hash(x)=19517612)
370
+ 2960 train 3.715865 (lr=1.1792e-03) (hash(x)=19643612)
371
+ 2970 train 3.638864 (lr=1.1684e-03) (hash(x)=17735841)
372
+ 2980 train 3.723222 (lr=1.1576e-03) (hash(x)=18179947)
373
+ 2990 train 3.641973 (lr=1.1468e-03) (hash(x)=19998132)
374
+ 3000 val loss 3.7229
375
+ 3000 val perplexity 41.3826
376
+ 3000 hella 0.2665
377
+ 3000 train 3.594529 (lr=1.1361e-03) (hash(x)=19400108)
378
+ 3010 train 3.581515 (lr=1.1254e-03) (hash(x)=19713177)
379
+ 3020 train 3.608329 (lr=1.1147e-03) (hash(x)=19871710)
380
+ 3030 train 3.570170 (lr=1.1041e-03) (hash(x)=16799044)
381
+ 3040 train 4.101714 (lr=1.0935e-03) (hash(x)=19431657)
382
+ 3050 train 3.844886 (lr=1.0830e-03) (hash(x)=22982891)
383
+ 3060 train 3.869033 (lr=1.0725e-03) (hash(x)=21562494)
384
+ 3070 train 3.748679 (lr=1.0620e-03) (hash(x)=33329781)
385
+ 3080 train 3.762794 (lr=1.0516e-03) (hash(x)=17296692)
386
+ 3090 train 3.789459 (lr=1.0412e-03) (hash(x)=19390072)
387
+ 3100 val loss 3.7087
388
+ 3100 val perplexity 40.8022
389
+ 3100 train 3.909796 (lr=1.0309e-03) (hash(x)=15455793)
390
+ 3110 train 3.757603 (lr=1.0207e-03) (hash(x)=20382524)
391
+ 3120 train 3.620229 (lr=1.0104e-03) (hash(x)=21248576)
392
+ 3130 train 3.656837 (lr=1.0002e-03) (hash(x)=20367412)
393
+ 3140 train 3.616878 (lr=9.9011e-04) (hash(x)=17826969)
394
+ 3150 train 3.711803 (lr=9.8002e-04) (hash(x)=21631872)
395
+ 3160 train 3.800831 (lr=9.6999e-04) (hash(x)=19506122)
396
+ 3170 train 3.832550 (lr=9.6000e-04) (hash(x)=21748750)
397
+ 3180 train 3.766370 (lr=9.5007e-04) (hash(x)=20805736)
398
+ 3190 train 3.798119 (lr=9.4019e-04) (hash(x)=19850067)
399
+ 3200 val loss 3.6888
400
+ 3200 val perplexity 39.9956
401
+ 3200 train 3.724610 (lr=9.3036e-04) (hash(x)=21173706)
402
+ 3210 train 3.724799 (lr=9.2058e-04) (hash(x)=21256910)
403
+ 3220 train 3.817132 (lr=9.1085e-04) (hash(x)=20765073)
404
+ 3230 train 3.740230 (lr=9.0118e-04) (hash(x)=17737503)
405
+ 3240 train 3.736017 (lr=8.9157e-04) (hash(x)=19598782)
406
+ 3250 hella 0.2619
407
+ 3250 train 3.708842 (lr=8.8201e-04) (hash(x)=24566640)
408
+ 3260 train 3.739036 (lr=8.7251e-04) (hash(x)=22002791)
409
+ 3270 train 3.607440 (lr=8.6307e-04) (hash(x)=20295322)
410
+ 3280 train 3.569705 (lr=8.5368e-04) (hash(x)=26247277)
411
+ 3290 train 3.611667 (lr=8.4435e-04) (hash(x)=20140665)
412
+ 3300 val loss 3.6771
413
+ 3300 val perplexity 39.5297
414
+ 3300 train 3.663679 (lr=8.3508e-04) (hash(x)=23553774)
415
+ 3310 train 3.782167 (lr=8.2588e-04) (hash(x)=17529719)
416
+ 3320 train 3.660716 (lr=8.1673e-04) (hash(x)=18795190)
417
+ 3330 train 3.799308 (lr=8.0764e-04) (hash(x)=20419570)
418
+ 3340 train 3.686274 (lr=7.9862e-04) (hash(x)=19596575)
419
+ 3350 train 3.698280 (lr=7.8966e-04) (hash(x)=17475331)
420
+ 3360 train 3.572740 (lr=7.8076e-04) (hash(x)=18856371)
421
+ 3370 train 3.750377 (lr=7.7192e-04) (hash(x)=19732920)
422
+ 3380 train 3.686846 (lr=7.6315e-04) (hash(x)=19151956)
423
+ 3390 train 3.649399 (lr=7.5445e-04) (hash(x)=22601715)
424
+ 3400 val loss 3.6608
425
+ 3400 val perplexity 38.8909
426
+ 3400 train 3.564934 (lr=7.4581e-04) (hash(x)=20446586)
427
+ 3410 train 3.754932 (lr=7.3724e-04) (hash(x)=27373387)
428
+ 3420 train 3.574583 (lr=7.2874e-04) (hash(x)=16894087)
429
+ 3430 train 3.612021 (lr=7.2030e-04) (hash(x)=19147767)
430
+ 3440 train 3.852565 (lr=7.1193e-04) (hash(x)=19394634)
431
+ 3450 train 3.732405 (lr=7.0363e-04) (hash(x)=20804137)
432
+ 3460 train 3.801222 (lr=6.9541e-04) (hash(x)=25730788)
433
+ 3470 train 3.667721 (lr=6.8725e-04) (hash(x)=18471384)
434
+ 3480 train 3.611242 (lr=6.7916e-04) (hash(x)=19884759)
435
+ 3490 train 3.655028 (lr=6.7114e-04) (hash(x)=21527738)
436
+ 3500 val loss 3.6449
437
+ 3500 val perplexity 38.2784
438
+ 3500 hella 0.2622
439
+ 3500 train 3.733375 (lr=6.6320e-04) (hash(x)=21108301)
440
+ 3510 train 3.661279 (lr=6.5533e-04) (hash(x)=19438613)
441
+ 3520 train 3.639672 (lr=6.4753e-04) (hash(x)=20391609)
442
+ 3530 train 3.545232 (lr=6.3981e-04) (hash(x)=21474987)
443
+ 3540 train 3.607800 (lr=6.3216e-04) (hash(x)=20544629)
444
+ 3550 train 3.655994 (lr=6.2458e-04) (hash(x)=18365326)
445
+ 3560 train 3.624661 (lr=6.1708e-04) (hash(x)=18443516)
446
+ 3570 train 3.642581 (lr=6.0966e-04) (hash(x)=18235257)
447
+ 3580 train 3.677064 (lr=6.0231e-04) (hash(x)=20958034)
448
+ 3590 train 3.662529 (lr=5.9504e-04) (hash(x)=20598224)
449
+ 3600 val loss 3.6348
450
+ 3600 val perplexity 37.8930
451
+ 3600 train 3.729249 (lr=5.8785e-04) (hash(x)=18613771)
452
+ 3610 train 3.836769 (lr=5.8074e-04) (hash(x)=19248168)
453
+ 3620 train 3.733755 (lr=5.7370e-04) (hash(x)=20935084)
454
+ 3630 train 3.802009 (lr=5.6675e-04) (hash(x)=20280969)
455
+ 3640 train 3.777437 (lr=5.5987e-04) (hash(x)=20735321)
456
+ 3650 train 3.646416 (lr=5.5308e-04) (hash(x)=20760704)
457
+ 3660 train 3.630726 (lr=5.4636e-04) (hash(x)=19656305)
458
+ 3670 train 3.637967 (lr=5.3973e-04) (hash(x)=19035247)
459
+ 3680 train 3.664675 (lr=5.3318e-04) (hash(x)=17935451)
460
+ 3690 train 3.600853 (lr=5.2671e-04) (hash(x)=22685971)
461
+ 3700 val loss 3.6253
462
+ 3700 val perplexity 37.5362
463
+ 3700 train 3.630539 (lr=5.2033e-04) (hash(x)=18466220)
464
+ 3710 train 3.484691 (lr=5.1402e-04) (hash(x)=19986945)
465
+ 3720 train 3.547804 (lr=5.0780e-04) (hash(x)=18017736)
466
+ 3730 train 3.610846 (lr=5.0167e-04) (hash(x)=19998802)
467
+ 3740 train 3.704241 (lr=4.9562e-04) (hash(x)=20145468)
468
+ 3750 hella 0.2658
469
+ 3750 train 3.622490 (lr=4.8965e-04) (hash(x)=16764686)
470
+ 3760 train 3.647695 (lr=4.8377e-04) (hash(x)=18833432)
471
+ 3770 train 3.561972 (lr=4.7798e-04) (hash(x)=16701754)
472
+ 3780 train 3.670729 (lr=4.7227e-04) (hash(x)=19608729)
473
+ 3790 train 3.694996 (lr=4.6665e-04) (hash(x)=19966203)
474
+ 3800 val loss 3.6039
475
+ 3800 val perplexity 36.7421
476
+ 3800 train 3.608176 (lr=4.6112e-04) (hash(x)=21276630)
477
+ 3810 train 3.719189 (lr=4.5567e-04) (hash(x)=19385436)
478
+ 3820 train 3.582239 (lr=4.5031e-04) (hash(x)=20894001)
479
+ 3830 train 3.563914 (lr=4.4504e-04) (hash(x)=22827106)
480
+ 3840 train 3.610210 (lr=4.3986e-04) (hash(x)=19263023)
481
+ 3850 train 3.594754 (lr=4.3477e-04) (hash(x)=19916874)
482
+ 3860 train 3.452785 (lr=4.2977e-04) (hash(x)=20568352)
483
+ 3870 train 3.389007 (lr=4.2486e-04) (hash(x)=24120742)
484
+ 3880 train 3.460385 (lr=4.2004e-04) (hash(x)=16266481)
485
+ 3890 train 3.506950 (lr=4.1530e-04) (hash(x)=19310691)
486
+ 3900 val loss 3.6113
487
+ 3900 val perplexity 37.0156
488
+ 3900 train 3.372241 (lr=4.1066e-04) (hash(x)=17735115)
489
+ 3910 train 3.593260 (lr=4.0611e-04) (hash(x)=18175918)
490
+ 3920 train 3.581697 (lr=4.0166e-04) (hash(x)=21662789)
491
+ 3930 train 3.617537 (lr=3.9729e-04) (hash(x)=21034272)
492
+ 3940 train 3.614618 (lr=3.9302e-04) (hash(x)=19724980)
493
+ 3950 train 3.589032 (lr=3.8884e-04) (hash(x)=20066944)
494
+ 3960 train 3.715968 (lr=3.8475e-04) (hash(x)=22065490)
495
+ 3970 train 3.616991 (lr=3.8076e-04) (hash(x)=20065365)
496
+ 3980 train 3.683762 (lr=3.7685e-04) (hash(x)=21419563)
497
+ 3990 train 3.673355 (lr=3.7305e-04) (hash(x)=21180852)
498
+ 4000 val loss 3.5890
499
+ 4000 val perplexity 36.1961
500
+ 4000 hella 0.2653
501
+ 4000 train 3.654587 (lr=3.6933e-04) (hash(x)=18548544)
502
+ 4010 train 3.644896 (lr=3.6572e-04) (hash(x)=20609113)
503
+ 4020 train 3.533629 (lr=3.6219e-04) (hash(x)=16977034)
504
+ 4030 train 3.561779 (lr=3.5876e-04) (hash(x)=19248645)
505
+ 4040 train 3.264720 (lr=3.5543e-04) (hash(x)=16881719)
506
+ 4050 train 3.347124 (lr=3.5219e-04) (hash(x)=18497713)
507
+ 4060 train 3.287878 (lr=3.4905e-04) (hash(x)=20036068)
508
+ 4070 train 3.262286 (lr=3.4600e-04) (hash(x)=21636163)
509
+ 4080 train 3.459686 (lr=3.4305e-04) (hash(x)=23072887)
510
+ 4090 train 3.588258 (lr=3.4019e-04) (hash(x)=27646813)
511
+ 4100 val loss 3.5849
512
+ 4100 val perplexity 36.0512
513
+ 4100 train 3.713919 (lr=3.3744e-04) (hash(x)=20052100)
514
+ 4110 train 3.630593 (lr=3.3477e-04) (hash(x)=21623032)
515
+ 4120 train 3.634803 (lr=3.3221e-04) (hash(x)=17755383)
516
+ 4130 train 3.648375 (lr=3.2974e-04) (hash(x)=20577475)
517
+ 4140 train 3.630732 (lr=3.2737e-04) (hash(x)=17308418)
518
+ 4150 train 3.685702 (lr=3.2510e-04) (hash(x)=20786408)
519
+ 4160 train 3.622211 (lr=3.2292e-04) (hash(x)=20006893)
520
+ 4170 train 3.588493 (lr=3.2085e-04) (hash(x)=20283286)
521
+ 4180 train 3.597498 (lr=3.1887e-04) (hash(x)=17799852)
522
+ 4190 train 3.516423 (lr=3.1699e-04) (hash(x)=20291468)
523
+ 4200 val loss 3.5768
524
+ 4200 val perplexity 35.7587
525
+ 4200 train 3.586923 (lr=3.1520e-04) (hash(x)=16871586)
526
+ 4210 train 3.633571 (lr=3.1352e-04) (hash(x)=21608030)
527
+ 4220 train 3.370864 (lr=3.1193e-04) (hash(x)=22256819)
528
+ 4230 train 3.442004 (lr=3.1044e-04) (hash(x)=20929071)
529
+ 4240 train 3.435757 (lr=3.0905e-04) (hash(x)=22408188)
530
+ 4250 hella 0.2659
531
+ 4250 train 3.358335 (lr=3.0776e-04) (hash(x)=19862925)
532
+ 4260 train 3.292884 (lr=3.0657e-04) (hash(x)=17765583)
533
+ 4270 train 3.604023 (lr=3.0548e-04) (hash(x)=21260687)
534
+ 4280 train 3.528854 (lr=3.0449e-04) (hash(x)=20333361)
535
+ 4290 train 3.577451 (lr=3.0359e-04) (hash(x)=21242172)
536
+ 4300 val loss 3.5691
537
+ 4300 val perplexity 35.4833
538
+ 4300 train 3.533235 (lr=3.0280e-04) (hash(x)=19835810)
539
+ 4310 train 3.582749 (lr=3.0210e-04) (hash(x)=32580183)
540
+ 4320 train 3.576273 (lr=3.0150e-04) (hash(x)=17944217)
541
+ 4330 train 3.587335 (lr=3.0101e-04) (hash(x)=20116568)
542
+ 4340 train 3.597612 (lr=3.0061e-04) (hash(x)=22423070)
543
+ 4350 train 3.634744 (lr=3.0031e-04) (hash(x)=15493166)
544
+ 4360 train 3.540227 (lr=3.0011e-04) (hash(x)=20130986)
545
+ 4370 train 3.624957 (lr=3.0001e-04) (hash(x)=20037228)
546
+ 4374 val loss 3.5686
547
+ 4374 val perplexity 35.4674
548
+ 4374 hella 0.2652
logs/64_head_dim_two_latent_masks/1_latent_mask_seed_1340/model_04374.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:503340d116825d1fd7116a2037cfc5dd265cae55c6816aaa0187e092eca9b46d
3
+ size 505703666
logs/64_head_dim_two_latent_masks/1_latent_mask_seed_1340/optimizer_04374.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:536faab5e0770cb7ffca84ac3559a15bd69cfac3d73958c7f967e17a8b599223
3
+ size 1005129270