andrew-healey commited on
Commit
8cc585d
·
verified ·
1 Parent(s): 5cfcd36

Upload folder using huggingface_hub

Browse files
scaling_mha_nh_3/mha_nh_128_seed_1339/args.json ADDED
@@ -0,0 +1 @@
 
 
1
+ {"hellaswag": true, "attention_kind": "dense", "log_dir": "scaling_mha_nh_3/mha_nh_128_seed_1339", "resume_checkpoint": null, "resume_optimizer": false, "add_a_head": false, "add_head_to_start": true, "new_head_init": "normal", "n_heads": 128, "n_embd": 256, "head_dim": 32, "protect_bos_token": true, "prevent_from_masking_myself": true, "max_steps": 4375, "warmup_steps": 250, "group": "scaling_mha_nh_3", "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, "disable_selection_head_linear_combo_bias": false, "assert_latent_matches_no_head": false, "n_latent_masks": null, "init_latent_masks_to_identity": false, "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, "att_conv": false, "att_conv_init": "none", "att_conv_scale": 1.0, "att_conv_weight_decay": 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": 256, "batch_size": 32, "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, "mup_zero_init": true, "l1_loss": false, "debugpy": false, "key": "mha_nh_128", "latent_mask_precision": "bfloat16", "att_conv_precision": "bfloat16", "attn_precision": "bfloat16", "profile_kind": "none", "dense_attention_kind": "mha", "head_dim_value": 32, "sdpa_iter_size": null, "stabilize_attn_scores": false, "override_use_sdpa": false, "autocast_precision": "bfloat16", "simulate_micro_bs": null, "simulate_micro_bs_2": null, "ckpt_attn": false, "c_proj_scale_init": 1.0}
scaling_mha_nh_3/mha_nh_128_seed_1339/dataloader_04374.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:6031fd3e2855a036f7a5531cc24555aabd1115f9dd6618b8b2ca6f55279ef0b2
3
+ size 964
scaling_mha_nh_3/mha_nh_128_seed_1339/log2.txt ADDED
@@ -0,0 +1,529 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ max_steps: 4375
2
+ 0 val loss 11.3316
3
+ 0 val perplexity 83418.4688
4
+ 0 train 11.321663 (lr=4.1958e-06) (hash(x)=44992657)
5
+ 10 train 9.901344 (lr=4.6154e-05) (hash(x)=33468021)
6
+ 20 train 9.250380 (lr=8.8112e-05) (hash(x)=40941803)
7
+ 30 train 8.534311 (lr=1.3007e-04) (hash(x)=36715902)
8
+ 40 train 7.744551 (lr=1.7203e-04) (hash(x)=32710993)
9
+ 50 train 7.396730 (lr=2.1399e-04) (hash(x)=43839896)
10
+ 60 train 7.384048 (lr=2.5594e-04) (hash(x)=40675468)
11
+ 70 train 6.985838 (lr=2.9790e-04) (hash(x)=34592631)
12
+ 80 train 7.005982 (lr=3.3986e-04) (hash(x)=44444845)
13
+ 90 train 6.852059 (lr=3.8182e-04) (hash(x)=41965258)
14
+ 100 val loss 6.7693
15
+ 100 val perplexity 870.7180
16
+ 100 train 6.624976 (lr=4.2378e-04) (hash(x)=41284750)
17
+ 110 train 6.662151 (lr=4.6573e-04) (hash(x)=41118734)
18
+ 120 train 6.534238 (lr=5.0769e-04) (hash(x)=37537547)
19
+ 130 train 6.410255 (lr=5.4965e-04) (hash(x)=43625179)
20
+ 140 train 6.423154 (lr=5.9161e-04) (hash(x)=41940760)
21
+ 150 train 6.503355 (lr=6.3357e-04) (hash(x)=39210431)
22
+ 160 train 6.466762 (lr=6.7552e-04) (hash(x)=41128294)
23
+ 170 train 6.381975 (lr=7.1748e-04) (hash(x)=41590227)
24
+ 180 train 6.330620 (lr=7.5944e-04) (hash(x)=38084871)
25
+ 190 train 6.269332 (lr=8.0140e-04) (hash(x)=34534333)
26
+ 200 val loss 6.1846
27
+ 200 val perplexity 485.2410
28
+ 200 train 6.216866 (lr=8.4336e-04) (hash(x)=41614299)
29
+ 210 train 6.133046 (lr=8.8531e-04) (hash(x)=35551597)
30
+ 220 train 6.079537 (lr=9.2727e-04) (hash(x)=44615432)
31
+ 230 train 6.213718 (lr=9.6923e-04) (hash(x)=37078768)
32
+ 240 train 5.939479 (lr=1.0112e-03) (hash(x)=40644078)
33
+ 250 train 6.005541 (lr=1.0531e-03) (hash(x)=44636066)
34
+ 260 train 5.898379 (lr=1.0951e-03) (hash(x)=43247952)
35
+ 270 train 5.864364 (lr=1.1371e-03) (hash(x)=37989370)
36
+ 280 train 5.812238 (lr=1.1790e-03) (hash(x)=36229730)
37
+ 290 train 5.735842 (lr=1.2210e-03) (hash(x)=35032996)
38
+ 300 val loss 5.8494
39
+ 300 val perplexity 347.0144
40
+ 300 train 5.620330 (lr=1.2629e-03) (hash(x)=33868636)
41
+ 310 train 5.577774 (lr=1.3049e-03) (hash(x)=39782705)
42
+ 320 train 5.643414 (lr=1.3469e-03) (hash(x)=43525765)
43
+ 330 train 5.729661 (lr=1.3888e-03) (hash(x)=38074926)
44
+ 340 train 5.753469 (lr=1.4308e-03) (hash(x)=50125702)
45
+ 350 train 5.752243 (lr=1.4727e-03) (hash(x)=35062222)
46
+ 360 train 5.677876 (lr=1.5147e-03) (hash(x)=39671001)
47
+ 370 train 5.671103 (lr=1.5566e-03) (hash(x)=35748364)
48
+ 380 train 5.482511 (lr=1.5986e-03) (hash(x)=39962666)
49
+ 390 train 5.457447 (lr=1.6406e-03) (hash(x)=39213360)
50
+ 400 val loss 5.4714
51
+ 400 val perplexity 237.7811
52
+ 400 train 5.459703 (lr=1.6825e-03) (hash(x)=37919584)
53
+ 410 train 5.382073 (lr=1.7245e-03) (hash(x)=42001527)
54
+ 420 train 5.383562 (lr=1.7664e-03) (hash(x)=33940005)
55
+ 430 train 5.418018 (lr=1.8084e-03) (hash(x)=41749217)
56
+ 440 train 5.295489 (lr=1.8503e-03) (hash(x)=45219739)
57
+ 450 train 5.279758 (lr=1.8923e-03) (hash(x)=38515090)
58
+ 460 train 5.208062 (lr=1.9343e-03) (hash(x)=37518607)
59
+ 470 train 5.024580 (lr=1.9762e-03) (hash(x)=46830716)
60
+ 480 train 5.122324 (lr=2.0182e-03) (hash(x)=49262598)
61
+ 490 train 4.948532 (lr=2.0601e-03) (hash(x)=41370559)
62
+ 500 val loss 5.2321
63
+ 500 val perplexity 187.1838
64
+ 500 train 5.019501 (lr=2.1021e-03) (hash(x)=35237459)
65
+ 510 train 5.144331 (lr=2.1441e-03) (hash(x)=35070579)
66
+ 520 train 5.194293 (lr=2.1860e-03) (hash(x)=37355110)
67
+ 530 train 5.308446 (lr=2.2280e-03) (hash(x)=39398044)
68
+ 540 train 5.210387 (lr=2.2699e-03) (hash(x)=39156676)
69
+ 550 train 5.174409 (lr=2.3119e-03) (hash(x)=38553909)
70
+ 560 train 5.213132 (lr=2.3538e-03) (hash(x)=36555371)
71
+ 570 train 5.157521 (lr=2.3958e-03) (hash(x)=42048991)
72
+ 580 train 5.100120 (lr=2.4378e-03) (hash(x)=37359399)
73
+ 590 train 5.099139 (lr=2.4797e-03) (hash(x)=39254663)
74
+ 600 val loss 5.0524
75
+ 600 val perplexity 156.4050
76
+ 600 train 5.080328 (lr=2.5217e-03) (hash(x)=38982539)
77
+ 610 train 5.014688 (lr=2.5636e-03) (hash(x)=35859145)
78
+ 620 train 4.986767 (lr=2.6056e-03) (hash(x)=35593542)
79
+ 630 train 4.901626 (lr=2.6476e-03) (hash(x)=39124418)
80
+ 640 train 4.898369 (lr=2.6895e-03) (hash(x)=39107374)
81
+ 650 train 4.860457 (lr=2.7315e-03) (hash(x)=34529151)
82
+ 660 train 4.759262 (lr=2.7734e-03) (hash(x)=39483931)
83
+ 670 train 4.776722 (lr=2.8154e-03) (hash(x)=43008010)
84
+ 680 train 4.673651 (lr=2.8573e-03) (hash(x)=33469328)
85
+ 690 train 4.706132 (lr=2.8993e-03) (hash(x)=35025253)
86
+ 700 val loss 4.9537
87
+ 700 val perplexity 141.7025
88
+ 700 train 4.684053 (lr=2.9413e-03) (hash(x)=43280188)
89
+ 710 train 4.974364 (lr=2.9832e-03) (hash(x)=40999263)
90
+ 720 train 4.929975 (lr=3.0000e-03) (hash(x)=38304742)
91
+ 730 train 4.947792 (lr=2.9999e-03) (hash(x)=34705188)
92
+ 740 train 4.909267 (lr=2.9997e-03) (hash(x)=40028964)
93
+ 750 train 4.886617 (lr=2.9994e-03) (hash(x)=39051591)
94
+ 760 train 4.889868 (lr=2.9990e-03) (hash(x)=41286198)
95
+ 770 train 4.832648 (lr=2.9985e-03) (hash(x)=44512967)
96
+ 780 train 4.841730 (lr=2.9979e-03) (hash(x)=45171627)
97
+ 790 train 4.810855 (lr=2.9972e-03) (hash(x)=46345673)
98
+ 800 val loss 4.7982
99
+ 800 val perplexity 121.2947
100
+ 800 train 4.753623 (lr=2.9964e-03) (hash(x)=35529545)
101
+ 810 train 4.687805 (lr=2.9955e-03) (hash(x)=42321793)
102
+ 820 train 4.684028 (lr=2.9945e-03) (hash(x)=49327569)
103
+ 830 train 4.767745 (lr=2.9934e-03) (hash(x)=41325057)
104
+ 840 train 4.699313 (lr=2.9922e-03) (hash(x)=37305614)
105
+ 850 train 4.637689 (lr=2.9909e-03) (hash(x)=38000800)
106
+ 860 train 4.759035 (lr=2.9896e-03) (hash(x)=44047697)
107
+ 870 train 4.773890 (lr=2.9881e-03) (hash(x)=41576206)
108
+ 880 train 4.729980 (lr=2.9865e-03) (hash(x)=44685458)
109
+ 890 train 4.698060 (lr=2.9848e-03) (hash(x)=40398597)
110
+ 900 val loss 4.6741
111
+ 900 val perplexity 107.1401
112
+ 900 train 4.640235 (lr=2.9830e-03) (hash(x)=39312673)
113
+ 910 train 4.641690 (lr=2.9811e-03) (hash(x)=38881875)
114
+ 920 train 4.688841 (lr=2.9792e-03) (hash(x)=39140166)
115
+ 930 train 4.570423 (lr=2.9771e-03) (hash(x)=39075781)
116
+ 940 train 4.591663 (lr=2.9749e-03) (hash(x)=46887071)
117
+ 950 train 4.572663 (lr=2.9726e-03) (hash(x)=39740603)
118
+ 960 train 4.643112 (lr=2.9703e-03) (hash(x)=39590268)
119
+ 970 train 4.582217 (lr=2.9678e-03) (hash(x)=43580052)
120
+ 980 train 4.503237 (lr=2.9652e-03) (hash(x)=40226735)
121
+ 990 train 4.616406 (lr=2.9626e-03) (hash(x)=40723896)
122
+ 1000 val loss 4.5614
123
+ 1000 val perplexity 95.7204
124
+ 1000 train 4.569756 (lr=2.9598e-03) (hash(x)=39408050)
125
+ 1010 train 4.737310 (lr=2.9570e-03) (hash(x)=32247950)
126
+ 1020 train 4.649333 (lr=2.9540e-03) (hash(x)=39049938)
127
+ 1030 train 4.606517 (lr=2.9510e-03) (hash(x)=37453736)
128
+ 1040 train 4.514922 (lr=2.9478e-03) (hash(x)=36649666)
129
+ 1050 train 4.585509 (lr=2.9446e-03) (hash(x)=41404097)
130
+ 1060 train 4.572637 (lr=2.9412e-03) (hash(x)=40701599)
131
+ 1070 train 4.580634 (lr=2.9378e-03) (hash(x)=41787393)
132
+ 1080 train 4.472532 (lr=2.9343e-03) (hash(x)=42325465)
133
+ 1090 train 4.499130 (lr=2.9307e-03) (hash(x)=45018925)
134
+ 1100 val loss 4.4886
135
+ 1100 val perplexity 88.9986
136
+ 1100 train 4.526674 (lr=2.9270e-03) (hash(x)=53751982)
137
+ 1110 train 4.457052 (lr=2.9231e-03) (hash(x)=43118376)
138
+ 1120 train 4.473028 (lr=2.9192e-03) (hash(x)=41489942)
139
+ 1130 train 4.495576 (lr=2.9152e-03) (hash(x)=48020487)
140
+ 1140 train 4.365980 (lr=2.9112e-03) (hash(x)=43266072)
141
+ 1150 train 4.570970 (lr=2.9070e-03) (hash(x)=41790132)
142
+ 1160 train 4.521842 (lr=2.9027e-03) (hash(x)=33739090)
143
+ 1170 train 4.497469 (lr=2.8983e-03) (hash(x)=43885030)
144
+ 1180 train 4.443029 (lr=2.8939e-03) (hash(x)=39046714)
145
+ 1190 train 4.475221 (lr=2.8893e-03) (hash(x)=44387023)
146
+ 1200 val loss 4.4066
147
+ 1200 val perplexity 81.9908
148
+ 1200 train 4.332144 (lr=2.8847e-03) (hash(x)=36667141)
149
+ 1210 train 4.409903 (lr=2.8800e-03) (hash(x)=45487179)
150
+ 1220 train 4.380666 (lr=2.8751e-03) (hash(x)=43456413)
151
+ 1230 train 4.314510 (lr=2.8702e-03) (hash(x)=42473632)
152
+ 1240 train 4.386058 (lr=2.8652e-03) (hash(x)=42966315)
153
+ 1250 train 4.233016 (lr=2.8601e-03) (hash(x)=42930524)
154
+ 1260 train 4.382577 (lr=2.8550e-03) (hash(x)=43467676)
155
+ 1270 train 4.309188 (lr=2.8497e-03) (hash(x)=40872580)
156
+ 1280 train 4.376933 (lr=2.8443e-03) (hash(x)=40619157)
157
+ 1290 train 4.396895 (lr=2.8389e-03) (hash(x)=44406165)
158
+ 1300 val loss 4.3588
159
+ 1300 val perplexity 78.1656
160
+ 1300 train 4.353383 (lr=2.8333e-03) (hash(x)=40808029)
161
+ 1310 train 4.541702 (lr=2.8277e-03) (hash(x)=46125736)
162
+ 1320 train 4.286896 (lr=2.8220e-03) (hash(x)=41698487)
163
+ 1330 train 4.328863 (lr=2.8162e-03) (hash(x)=39302878)
164
+ 1340 train 4.437850 (lr=2.8103e-03) (hash(x)=41826369)
165
+ 1350 train 4.256717 (lr=2.8044e-03) (hash(x)=37061413)
166
+ 1360 train 4.377034 (lr=2.7983e-03) (hash(x)=40770974)
167
+ 1370 train 4.300165 (lr=2.7922e-03) (hash(x)=38738718)
168
+ 1380 train 4.285026 (lr=2.7860e-03) (hash(x)=34961558)
169
+ 1390 train 4.338543 (lr=2.7797e-03) (hash(x)=38559342)
170
+ 1400 val loss 4.3363
171
+ 1400 val perplexity 76.4280
172
+ 1400 train 4.286959 (lr=2.7733e-03) (hash(x)=38840020)
173
+ 1410 train 4.443886 (lr=2.7668e-03) (hash(x)=44448100)
174
+ 1420 train 4.514891 (lr=2.7603e-03) (hash(x)=39199838)
175
+ 1430 train 4.298913 (lr=2.7536e-03) (hash(x)=41669873)
176
+ 1440 train 4.431464 (lr=2.7469e-03) (hash(x)=53253836)
177
+ 1450 train 4.349299 (lr=2.7401e-03) (hash(x)=41320631)
178
+ 1460 train 4.266146 (lr=2.7333e-03) (hash(x)=48462803)
179
+ 1470 train 4.306919 (lr=2.7263e-03) (hash(x)=42356864)
180
+ 1480 train 4.351571 (lr=2.7193e-03) (hash(x)=41118068)
181
+ 1490 train 4.263971 (lr=2.7121e-03) (hash(x)=37242854)
182
+ 1500 val loss 4.2848
183
+ 1500 val perplexity 72.5900
184
+ 1500 train 4.309509 (lr=2.7050e-03) (hash(x)=40571031)
185
+ 1510 train 4.265206 (lr=2.6977e-03) (hash(x)=35384608)
186
+ 1520 train 4.292012 (lr=2.6903e-03) (hash(x)=35341056)
187
+ 1530 train 4.183613 (lr=2.6829e-03) (hash(x)=36471779)
188
+ 1540 train 4.219311 (lr=2.6754e-03) (hash(x)=40356153)
189
+ 1550 train 4.255385 (lr=2.6678e-03) (hash(x)=40133377)
190
+ 1560 train 4.189948 (lr=2.6602e-03) (hash(x)=33722160)
191
+ 1570 train 4.336834 (lr=2.6525e-03) (hash(x)=39555295)
192
+ 1580 train 4.243438 (lr=2.6447e-03) (hash(x)=35950666)
193
+ 1590 train 4.385462 (lr=2.6368e-03) (hash(x)=36528535)
194
+ 1600 val loss 4.2643
195
+ 1600 val perplexity 71.1133
196
+ 1600 train 4.282264 (lr=2.6289e-03) (hash(x)=42613017)
197
+ 1610 train 4.422867 (lr=2.6208e-03) (hash(x)=42845965)
198
+ 1620 train 4.296845 (lr=2.6128e-03) (hash(x)=36137074)
199
+ 1630 train 4.263924 (lr=2.6046e-03) (hash(x)=37560569)
200
+ 1640 train 4.356207 (lr=2.5964e-03) (hash(x)=43680614)
201
+ 1650 train 4.219458 (lr=2.5881e-03) (hash(x)=37281218)
202
+ 1660 train 4.248601 (lr=2.5797e-03) (hash(x)=36165222)
203
+ 1670 train 4.196683 (lr=2.5713e-03) (hash(x)=50222513)
204
+ 1680 train 4.087750 (lr=2.5628e-03) (hash(x)=35239173)
205
+ 1690 train 4.140374 (lr=2.5542e-03) (hash(x)=40287349)
206
+ 1700 val loss 4.2613
207
+ 1700 val perplexity 70.9049
208
+ 1700 train 4.330542 (lr=2.5455e-03) (hash(x)=37444780)
209
+ 1710 train 4.314125 (lr=2.5368e-03) (hash(x)=38962723)
210
+ 1720 train 4.310055 (lr=2.5281e-03) (hash(x)=33882305)
211
+ 1730 train 4.289991 (lr=2.5192e-03) (hash(x)=36501540)
212
+ 1740 train 4.293424 (lr=2.5103e-03) (hash(x)=38723266)
213
+ 1750 train 4.223124 (lr=2.5014e-03) (hash(x)=38772789)
214
+ 1760 train 4.295739 (lr=2.4924e-03) (hash(x)=37707898)
215
+ 1770 train 4.157452 (lr=2.4833e-03) (hash(x)=30529327)
216
+ 1780 train 4.249148 (lr=2.4741e-03) (hash(x)=37885464)
217
+ 1790 train 4.258193 (lr=2.4649e-03) (hash(x)=39013967)
218
+ 1800 val loss 4.2172
219
+ 1800 val perplexity 67.8439
220
+ 1800 train 4.293533 (lr=2.4556e-03) (hash(x)=34906955)
221
+ 1810 train 4.170959 (lr=2.4463e-03) (hash(x)=40790682)
222
+ 1820 train 4.037841 (lr=2.4369e-03) (hash(x)=39912223)
223
+ 1830 train 4.161463 (lr=2.4275e-03) (hash(x)=39304486)
224
+ 1840 train 4.134646 (lr=2.4180e-03) (hash(x)=37247975)
225
+ 1850 train 4.121092 (lr=2.4084e-03) (hash(x)=41340614)
226
+ 1860 train 4.279132 (lr=2.3988e-03) (hash(x)=39924627)
227
+ 1870 train 4.262952 (lr=2.3891e-03) (hash(x)=40235993)
228
+ 1880 train 4.160750 (lr=2.3794e-03) (hash(x)=37693476)
229
+ 1890 train 4.263515 (lr=2.3696e-03) (hash(x)=40449864)
230
+ 1900 val loss 4.1850
231
+ 1900 val perplexity 65.6933
232
+ 1900 train 4.176617 (lr=2.3598e-03) (hash(x)=37907749)
233
+ 1910 train 4.142320 (lr=2.3499e-03) (hash(x)=34547901)
234
+ 1920 train 4.179153 (lr=2.3400e-03) (hash(x)=37771788)
235
+ 1930 train 4.047329 (lr=2.3300e-03) (hash(x)=36049454)
236
+ 1940 train 4.146369 (lr=2.3200e-03) (hash(x)=29564336)
237
+ 1950 train 4.136189 (lr=2.3099e-03) (hash(x)=41689281)
238
+ 1960 train 4.049983 (lr=2.2998e-03) (hash(x)=41976699)
239
+ 1970 train 4.244591 (lr=2.2896e-03) (hash(x)=41538621)
240
+ 1980 train 4.131462 (lr=2.2793e-03) (hash(x)=36490902)
241
+ 1990 train 4.159683 (lr=2.2691e-03) (hash(x)=42739315)
242
+ 2000 val loss 4.1650
243
+ 2000 val perplexity 64.3931
244
+ 2000 train 4.181746 (lr=2.2588e-03) (hash(x)=38600074)
245
+ 2010 train 4.196359 (lr=2.2484e-03) (hash(x)=34555488)
246
+ 2020 train 4.211807 (lr=2.2380e-03) (hash(x)=37227345)
247
+ 2030 train 4.160433 (lr=2.2275e-03) (hash(x)=38172386)
248
+ 2040 train 4.267165 (lr=2.2170e-03) (hash(x)=39340786)
249
+ 2050 train 4.112098 (lr=2.2065e-03) (hash(x)=37237298)
250
+ 2060 train 4.351136 (lr=2.1959e-03) (hash(x)=35894782)
251
+ 2070 train 4.169959 (lr=2.1853e-03) (hash(x)=44161861)
252
+ 2080 train 4.188072 (lr=2.1746e-03) (hash(x)=38062558)
253
+ 2090 train 4.038575 (lr=2.1639e-03) (hash(x)=38608301)
254
+ 2100 val loss 4.1365
255
+ 2100 val perplexity 62.5856
256
+ 2100 train 4.065739 (lr=2.1532e-03) (hash(x)=47611537)
257
+ 2110 train 4.064577 (lr=2.1424e-03) (hash(x)=38076651)
258
+ 2120 train 4.071224 (lr=2.1316e-03) (hash(x)=39870893)
259
+ 2130 train 4.225191 (lr=2.1208e-03) (hash(x)=48123321)
260
+ 2140 train 4.256063 (lr=2.1099e-03) (hash(x)=44124977)
261
+ 2150 train 4.152084 (lr=2.0990e-03) (hash(x)=42181374)
262
+ 2160 train 4.175650 (lr=2.0881e-03) (hash(x)=41837878)
263
+ 2170 train 4.199287 (lr=2.0771e-03) (hash(x)=42161375)
264
+ 2180 train 4.173043 (lr=2.0661e-03) (hash(x)=41792986)
265
+ 2190 train 4.135368 (lr=2.0550e-03) (hash(x)=41872416)
266
+ 2200 val loss 4.1126
267
+ 2200 val perplexity 61.1027
268
+ 2200 train 4.103052 (lr=2.0440e-03) (hash(x)=37395985)
269
+ 2210 train 4.154723 (lr=2.0329e-03) (hash(x)=40616096)
270
+ 2220 train 4.061225 (lr=2.0217e-03) (hash(x)=40118423)
271
+ 2230 train 3.973494 (lr=2.0106e-03) (hash(x)=39641586)
272
+ 2240 train 4.063206 (lr=1.9994e-03) (hash(x)=34720116)
273
+ 2250 train 4.004764 (lr=1.9882e-03) (hash(x)=43641508)
274
+ 2260 train 4.019839 (lr=1.9770e-03) (hash(x)=31179786)
275
+ 2270 train 4.000600 (lr=1.9657e-03) (hash(x)=42721932)
276
+ 2280 train 4.321383 (lr=1.9544e-03) (hash(x)=38474505)
277
+ 2290 train 4.148006 (lr=1.9431e-03) (hash(x)=41398545)
278
+ 2300 val loss 4.0907
279
+ 2300 val perplexity 59.7818
280
+ 2300 train 4.081727 (lr=1.9318e-03) (hash(x)=44131094)
281
+ 2310 train 4.154692 (lr=1.9205e-03) (hash(x)=41667948)
282
+ 2320 train 4.238863 (lr=1.9091e-03) (hash(x)=36339232)
283
+ 2330 train 4.112579 (lr=1.8977e-03) (hash(x)=41630644)
284
+ 2340 train 4.195613 (lr=1.8863e-03) (hash(x)=41278955)
285
+ 2350 train 4.072313 (lr=1.8749e-03) (hash(x)=39491976)
286
+ 2360 train 4.140317 (lr=1.8635e-03) (hash(x)=41771238)
287
+ 2370 train 4.229527 (lr=1.8520e-03) (hash(x)=45374570)
288
+ 2380 train 4.107696 (lr=1.8406e-03) (hash(x)=41419267)
289
+ 2390 train 4.142347 (lr=1.8291e-03) (hash(x)=38869169)
290
+ 2400 val loss 4.0807
291
+ 2400 val perplexity 59.1877
292
+ 2400 train 4.043715 (lr=1.8176e-03) (hash(x)=38619293)
293
+ 2410 train 3.863710 (lr=1.8061e-03) (hash(x)=41600240)
294
+ 2420 train 3.993495 (lr=1.7946e-03) (hash(x)=40891045)
295
+ 2430 train 3.989757 (lr=1.7830e-03) (hash(x)=41023249)
296
+ 2440 train 3.957728 (lr=1.7715e-03) (hash(x)=33813452)
297
+ 2450 train 3.904814 (lr=1.7600e-03) (hash(x)=38464119)
298
+ 2460 train 3.802065 (lr=1.7484e-03) (hash(x)=40699982)
299
+ 2470 train 3.812940 (lr=1.7368e-03) (hash(x)=38254854)
300
+ 2480 train 4.099656 (lr=1.7253e-03) (hash(x)=42736069)
301
+ 2490 train 4.113184 (lr=1.7137e-03) (hash(x)=41928525)
302
+ 2500 val loss 4.0548
303
+ 2500 val perplexity 57.6757
304
+ 2500 train 4.067765 (lr=1.7021e-03) (hash(x)=45864011)
305
+ 2510 train 4.080892 (lr=1.6906e-03) (hash(x)=41380960)
306
+ 2520 train 4.079009 (lr=1.6790e-03) (hash(x)=38366545)
307
+ 2530 train 4.077400 (lr=1.6674e-03) (hash(x)=41045176)
308
+ 2540 train 4.108951 (lr=1.6558e-03) (hash(x)=43435705)
309
+ 2550 train 4.058837 (lr=1.6442e-03) (hash(x)=40693090)
310
+ 2560 train 3.978027 (lr=1.6326e-03) (hash(x)=40502478)
311
+ 2570 train 4.004478 (lr=1.6210e-03) (hash(x)=38610920)
312
+ 2580 train 3.938304 (lr=1.6094e-03) (hash(x)=38333499)
313
+ 2590 train 4.007317 (lr=1.5979e-03) (hash(x)=41982736)
314
+ 2600 val loss 4.0348
315
+ 2600 val perplexity 56.5322
316
+ 2600 train 3.986619 (lr=1.5863e-03) (hash(x)=37724702)
317
+ 2610 train 3.924408 (lr=1.5747e-03) (hash(x)=39564630)
318
+ 2620 train 3.874835 (lr=1.5632e-03) (hash(x)=38465803)
319
+ 2630 train 3.789392 (lr=1.5516e-03) (hash(x)=36955007)
320
+ 2640 train 3.824954 (lr=1.5400e-03) (hash(x)=32994568)
321
+ 2650 train 4.010929 (lr=1.5285e-03) (hash(x)=41165765)
322
+ 2660 train 4.028086 (lr=1.5170e-03) (hash(x)=42934000)
323
+ 2670 train 4.216432 (lr=1.5054e-03) (hash(x)=42726316)
324
+ 2680 train 4.111498 (lr=1.4939e-03) (hash(x)=35880318)
325
+ 2690 train 3.985913 (lr=1.4824e-03) (hash(x)=41747329)
326
+ 2700 val loss 4.0259
327
+ 2700 val perplexity 56.0334
328
+ 2700 train 4.113819 (lr=1.4709e-03) (hash(x)=40259630)
329
+ 2710 train 4.240404 (lr=1.4594e-03) (hash(x)=39895798)
330
+ 2720 train 3.967150 (lr=1.4480e-03) (hash(x)=36146683)
331
+ 2730 train 3.940982 (lr=1.4365e-03) (hash(x)=36181984)
332
+ 2740 train 3.923096 (lr=1.4251e-03) (hash(x)=43700349)
333
+ 2750 train 3.993842 (lr=1.4137e-03) (hash(x)=41610597)
334
+ 2760 train 3.977494 (lr=1.4023e-03) (hash(x)=31183639)
335
+ 2770 train 3.974437 (lr=1.3909e-03) (hash(x)=37722489)
336
+ 2780 train 3.917837 (lr=1.3795e-03) (hash(x)=47290688)
337
+ 2790 train 3.955759 (lr=1.3682e-03) (hash(x)=41205574)
338
+ 2800 val loss 4.0070
339
+ 2800 val perplexity 54.9825
340
+ 2800 train 3.802703 (lr=1.3569e-03) (hash(x)=40257962)
341
+ 2810 train 3.779823 (lr=1.3456e-03) (hash(x)=39529014)
342
+ 2820 train 3.802684 (lr=1.3343e-03) (hash(x)=42244749)
343
+ 2830 train 3.851337 (lr=1.3230e-03) (hash(x)=32820090)
344
+ 2840 train 3.810652 (lr=1.3118e-03) (hash(x)=40315769)
345
+ 2850 train 4.054941 (lr=1.3006e-03) (hash(x)=41524462)
346
+ 2860 train 4.049762 (lr=1.2894e-03) (hash(x)=38365734)
347
+ 2870 train 4.177267 (lr=1.2783e-03) (hash(x)=37682602)
348
+ 2880 train 4.029356 (lr=1.2671e-03) (hash(x)=39162991)
349
+ 2890 train 3.885901 (lr=1.2560e-03) (hash(x)=33316384)
350
+ 2900 val loss 3.9817
351
+ 2900 val perplexity 53.6065
352
+ 2900 train 3.937141 (lr=1.2450e-03) (hash(x)=37271132)
353
+ 2910 train 4.124488 (lr=1.2339e-03) (hash(x)=35586242)
354
+ 2920 train 4.011837 (lr=1.2229e-03) (hash(x)=33320586)
355
+ 2930 train 4.029016 (lr=1.2119e-03) (hash(x)=43531361)
356
+ 2940 train 3.914333 (lr=1.2010e-03) (hash(x)=37368286)
357
+ 2950 train 3.954691 (lr=1.1901e-03) (hash(x)=40363394)
358
+ 2960 train 3.915760 (lr=1.1792e-03) (hash(x)=42253792)
359
+ 2970 train 3.855254 (lr=1.1684e-03) (hash(x)=38072598)
360
+ 2980 train 3.932886 (lr=1.1576e-03) (hash(x)=41470557)
361
+ 2990 train 3.854511 (lr=1.1468e-03) (hash(x)=42600033)
362
+ 3000 val loss 3.9812
363
+ 3000 val perplexity 53.5807
364
+ 3000 train 3.810907 (lr=1.1361e-03) (hash(x)=46890983)
365
+ 3010 train 3.676366 (lr=1.1254e-03) (hash(x)=37986759)
366
+ 3020 train 3.803369 (lr=1.1147e-03) (hash(x)=36270703)
367
+ 3030 train 4.070835 (lr=1.1041e-03) (hash(x)=38228599)
368
+ 3040 train 3.919982 (lr=1.0935e-03) (hash(x)=44344296)
369
+ 3050 train 4.005668 (lr=1.0830e-03) (hash(x)=39962297)
370
+ 3060 train 3.985434 (lr=1.0725e-03) (hash(x)=39817394)
371
+ 3070 train 3.979778 (lr=1.0620e-03) (hash(x)=41763868)
372
+ 3080 train 4.061150 (lr=1.0516e-03) (hash(x)=42343051)
373
+ 3090 train 4.066368 (lr=1.0412e-03) (hash(x)=39204893)
374
+ 3100 val loss 3.9553
375
+ 3100 val perplexity 52.2093
376
+ 3100 train 3.964817 (lr=1.0309e-03) (hash(x)=39353599)
377
+ 3110 train 4.033218 (lr=1.0207e-03) (hash(x)=36118073)
378
+ 3120 train 3.736624 (lr=1.0104e-03) (hash(x)=44070271)
379
+ 3130 train 3.761448 (lr=1.0002e-03) (hash(x)=43368151)
380
+ 3140 train 3.837734 (lr=9.9011e-04) (hash(x)=40612203)
381
+ 3150 train 3.975158 (lr=9.8002e-04) (hash(x)=39598447)
382
+ 3160 train 4.026445 (lr=9.6999e-04) (hash(x)=43897682)
383
+ 3170 train 4.009558 (lr=9.6000e-04) (hash(x)=39029243)
384
+ 3180 train 4.014506 (lr=9.5007e-04) (hash(x)=38586684)
385
+ 3190 train 3.969123 (lr=9.4019e-04) (hash(x)=39299115)
386
+ 3200 val loss 3.9365
387
+ 3200 val perplexity 51.2376
388
+ 3200 train 3.999774 (lr=9.3036e-04) (hash(x)=41064949)
389
+ 3210 train 3.984034 (lr=9.2058e-04) (hash(x)=41752822)
390
+ 3220 train 3.990107 (lr=9.1085e-04) (hash(x)=37484805)
391
+ 3230 train 3.953846 (lr=9.0118e-04) (hash(x)=40514919)
392
+ 3240 train 4.005499 (lr=8.9157e-04) (hash(x)=42440343)
393
+ 3250 train 3.906151 (lr=8.8201e-04) (hash(x)=32464136)
394
+ 3260 train 3.921327 (lr=8.7251e-04) (hash(x)=47270558)
395
+ 3270 train 3.867382 (lr=8.6307e-04) (hash(x)=39307546)
396
+ 3280 train 3.903457 (lr=8.5368e-04) (hash(x)=40410299)
397
+ 3290 train 3.898590 (lr=8.4435e-04) (hash(x)=40541497)
398
+ 3300 val loss 3.9278
399
+ 3300 val perplexity 50.7967
400
+ 3300 train 3.795523 (lr=8.3508e-04) (hash(x)=38089139)
401
+ 3310 train 3.827644 (lr=8.2588e-04) (hash(x)=42006576)
402
+ 3320 train 3.922816 (lr=8.1673e-04) (hash(x)=40422556)
403
+ 3330 train 3.709562 (lr=8.0764e-04) (hash(x)=39021643)
404
+ 3340 train 3.974061 (lr=7.9862e-04) (hash(x)=40434305)
405
+ 3350 train 4.004039 (lr=7.8966e-04) (hash(x)=40833559)
406
+ 3360 train 4.003766 (lr=7.8076e-04) (hash(x)=40045231)
407
+ 3370 train 3.965215 (lr=7.7192e-04) (hash(x)=34668317)
408
+ 3380 train 3.954990 (lr=7.6315e-04) (hash(x)=42738568)
409
+ 3390 train 4.057551 (lr=7.5445e-04) (hash(x)=38980166)
410
+ 3400 val loss 3.9058
411
+ 3400 val perplexity 49.6876
412
+ 3400 train 3.965125 (lr=7.4581e-04) (hash(x)=43878176)
413
+ 3410 train 4.011271 (lr=7.3724e-04) (hash(x)=38201991)
414
+ 3420 train 3.983220 (lr=7.2874e-04) (hash(x)=34905889)
415
+ 3430 train 3.983684 (lr=7.2030e-04) (hash(x)=41680405)
416
+ 3440 train 3.947958 (lr=7.1193e-04) (hash(x)=39948184)
417
+ 3450 train 3.996583 (lr=7.0363e-04) (hash(x)=33623728)
418
+ 3460 train 3.866271 (lr=6.9541e-04) (hash(x)=41027647)
419
+ 3470 train 3.897913 (lr=6.8725e-04) (hash(x)=42779482)
420
+ 3480 train 3.919784 (lr=6.7916e-04) (hash(x)=41579288)
421
+ 3490 train 3.773850 (lr=6.7114e-04) (hash(x)=37174846)
422
+ 3500 val loss 3.9005
423
+ 3500 val perplexity 49.4247
424
+ 3500 train 3.643483 (lr=6.6320e-04) (hash(x)=37082090)
425
+ 3510 train 3.846830 (lr=6.5533e-04) (hash(x)=43131340)
426
+ 3520 train 3.761972 (lr=6.4753e-04) (hash(x)=42457250)
427
+ 3530 train 3.992289 (lr=6.3981e-04) (hash(x)=39504492)
428
+ 3540 train 3.942380 (lr=6.3216e-04) (hash(x)=36225997)
429
+ 3550 train 4.063926 (lr=6.2458e-04) (hash(x)=40912087)
430
+ 3560 train 3.906060 (lr=6.1708e-04) (hash(x)=41545925)
431
+ 3570 train 3.994450 (lr=6.0966e-04) (hash(x)=32553193)
432
+ 3580 train 3.835348 (lr=6.0231e-04) (hash(x)=38799856)
433
+ 3590 train 3.882989 (lr=5.9504e-04) (hash(x)=40772075)
434
+ 3600 val loss 3.8806
435
+ 3600 val perplexity 48.4537
436
+ 3600 train 3.927463 (lr=5.8785e-04) (hash(x)=39299903)
437
+ 3610 train 3.969415 (lr=5.8074e-04) (hash(x)=39503247)
438
+ 3620 train 3.865871 (lr=5.7370e-04) (hash(x)=38730668)
439
+ 3630 train 3.885098 (lr=5.6675e-04) (hash(x)=40503276)
440
+ 3640 train 3.894083 (lr=5.5987e-04) (hash(x)=36504573)
441
+ 3650 train 3.914043 (lr=5.5308e-04) (hash(x)=44903075)
442
+ 3660 train 3.821544 (lr=5.4636e-04) (hash(x)=33800148)
443
+ 3670 train 3.731025 (lr=5.3973e-04) (hash(x)=39404541)
444
+ 3680 train 3.679420 (lr=5.3318e-04) (hash(x)=44777432)
445
+ 3690 train 3.771711 (lr=5.2671e-04) (hash(x)=45649913)
446
+ 3700 val loss 3.8819
447
+ 3700 val perplexity 48.5157
448
+ 3700 train 3.899338 (lr=5.2033e-04) (hash(x)=33812537)
449
+ 3710 train 4.048628 (lr=5.1402e-04) (hash(x)=36161222)
450
+ 3720 train 3.924482 (lr=5.0780e-04) (hash(x)=43690616)
451
+ 3730 train 3.929818 (lr=5.0167e-04) (hash(x)=35462313)
452
+ 3740 train 4.224585 (lr=4.9562e-04) (hash(x)=44674254)
453
+ 3750 train 3.907825 (lr=4.8965e-04) (hash(x)=44276297)
454
+ 3760 train 3.953198 (lr=4.8377e-04) (hash(x)=38492178)
455
+ 3770 train 3.960186 (lr=4.7798e-04) (hash(x)=44347531)
456
+ 3780 train 3.870839 (lr=4.7227e-04) (hash(x)=33792059)
457
+ 3790 train 3.967735 (lr=4.6665e-04) (hash(x)=38549318)
458
+ 3800 val loss 3.8632
459
+ 3800 val perplexity 47.6160
460
+ 3800 train 3.848335 (lr=4.6112e-04) (hash(x)=41437448)
461
+ 3810 train 3.748403 (lr=4.5567e-04) (hash(x)=36656768)
462
+ 3820 train 3.905023 (lr=4.5031e-04) (hash(x)=46387893)
463
+ 3830 train 3.907978 (lr=4.4504e-04) (hash(x)=44312129)
464
+ 3840 train 3.826357 (lr=4.3986e-04) (hash(x)=40341896)
465
+ 3850 train 3.879124 (lr=4.3477e-04) (hash(x)=43589160)
466
+ 3860 train 3.858800 (lr=4.2977e-04) (hash(x)=40388945)
467
+ 3870 train 3.760086 (lr=4.2486e-04) (hash(x)=33336679)
468
+ 3880 train 3.855033 (lr=4.2004e-04) (hash(x)=37643311)
469
+ 3890 train 3.877756 (lr=4.1530e-04) (hash(x)=38272927)
470
+ 3900 val loss 3.8536
471
+ 3900 val perplexity 47.1646
472
+ 3900 train 3.822834 (lr=4.1066e-04) (hash(x)=41925748)
473
+ 3910 train 3.914198 (lr=4.0611e-04) (hash(x)=38377253)
474
+ 3920 train 3.917940 (lr=4.0166e-04) (hash(x)=41616611)
475
+ 3930 train 3.841764 (lr=3.9729e-04) (hash(x)=32301827)
476
+ 3940 train 3.896273 (lr=3.9302e-04) (hash(x)=47697363)
477
+ 3950 train 3.851194 (lr=3.8884e-04) (hash(x)=37867767)
478
+ 3960 train 3.847948 (lr=3.8475e-04) (hash(x)=37187295)
479
+ 3970 train 3.899066 (lr=3.8076e-04) (hash(x)=41952752)
480
+ 3980 train 3.872138 (lr=3.7685e-04) (hash(x)=38358660)
481
+ 3990 train 3.741080 (lr=3.7305e-04) (hash(x)=40207878)
482
+ 4000 val loss 3.8475
483
+ 4000 val perplexity 46.8744
484
+ 4000 train 3.829473 (lr=3.6933e-04) (hash(x)=39134015)
485
+ 4010 train 3.898206 (lr=3.6572e-04) (hash(x)=38313135)
486
+ 4020 train 3.825788 (lr=3.6219e-04) (hash(x)=40710513)
487
+ 4030 train 3.961014 (lr=3.5876e-04) (hash(x)=35530165)
488
+ 4040 train 3.877270 (lr=3.5543e-04) (hash(x)=39494066)
489
+ 4050 train 3.870382 (lr=3.5219e-04) (hash(x)=40861777)
490
+ 4060 train 3.850515 (lr=3.4905e-04) (hash(x)=36151939)
491
+ 4070 train 3.812214 (lr=3.4600e-04) (hash(x)=39008241)
492
+ 4080 train 3.903254 (lr=3.4305e-04) (hash(x)=40395746)
493
+ 4090 train 3.939052 (lr=3.4019e-04) (hash(x)=38855480)
494
+ 4100 val loss 3.8343
495
+ 4100 val perplexity 46.2614
496
+ 4100 train 3.874414 (lr=3.3744e-04) (hash(x)=45013254)
497
+ 4110 train 3.881991 (lr=3.3477e-04) (hash(x)=39934818)
498
+ 4120 train 3.881070 (lr=3.3221e-04) (hash(x)=33417705)
499
+ 4130 train 3.816348 (lr=3.2974e-04) (hash(x)=41267978)
500
+ 4140 train 3.812355 (lr=3.2737e-04) (hash(x)=38961609)
501
+ 4150 train 3.814623 (lr=3.2510e-04) (hash(x)=42173878)
502
+ 4160 train 3.779139 (lr=3.2292e-04) (hash(x)=36867405)
503
+ 4170 train 3.757079 (lr=3.2085e-04) (hash(x)=38991954)
504
+ 4180 train 3.823215 (lr=3.1887e-04) (hash(x)=39032592)
505
+ 4190 train 3.797640 (lr=3.1699e-04) (hash(x)=61329402)
506
+ 4200 val loss 3.8343
507
+ 4200 val perplexity 46.2614
508
+ 4200 train 3.813282 (lr=3.1520e-04) (hash(x)=40455200)
509
+ 4210 train 3.840207 (lr=3.1352e-04) (hash(x)=40913417)
510
+ 4220 train 3.872471 (lr=3.1193e-04) (hash(x)=38809704)
511
+ 4230 train 3.856660 (lr=3.1044e-04) (hash(x)=39560204)
512
+ 4240 train 3.855508 (lr=3.0905e-04) (hash(x)=41496549)
513
+ 4250 train 3.897245 (lr=3.0776e-04) (hash(x)=33994768)
514
+ 4260 train 3.881306 (lr=3.0657e-04) (hash(x)=39658687)
515
+ 4270 train 3.763676 (lr=3.0548e-04) (hash(x)=42583064)
516
+ 4280 train 3.826487 (lr=3.0449e-04) (hash(x)=40635811)
517
+ 4290 train 3.723177 (lr=3.0359e-04) (hash(x)=35313841)
518
+ 4300 val loss 3.8299
519
+ 4300 val perplexity 46.0561
520
+ 4300 train 3.902964 (lr=3.0280e-04) (hash(x)=39332245)
521
+ 4310 train 3.798486 (lr=3.0210e-04) (hash(x)=43732999)
522
+ 4320 train 3.806530 (lr=3.0150e-04) (hash(x)=40831835)
523
+ 4330 train 3.940930 (lr=3.0101e-04) (hash(x)=41813368)
524
+ 4340 train 3.845439 (lr=3.0061e-04) (hash(x)=36952586)
525
+ 4350 train 3.869406 (lr=3.0031e-04) (hash(x)=42470745)
526
+ 4360 train 3.797637 (lr=3.0011e-04) (hash(x)=39581495)
527
+ 4370 train 3.991636 (lr=3.0001e-04) (hash(x)=35914538)
528
+ 4374 val loss 3.8225
529
+ 4374 val perplexity 45.7191
scaling_mha_nh_3/mha_nh_128_seed_1339/model_04374.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:a2e9867a428ef35ee9ab8393bde7d5aeb07ac03a3da39d3710a8e5547825ca39
3
+ size 282143938
scaling_mha_nh_3/mha_nh_128_seed_1339/optimizer_04374.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:a3aa95b7121034409812d83b70b5308e692d52f6edc0957e94986d8302dd5ffb
3
+ size 557999750