Upload folder using huggingface_hub

attention_kindselective_n_heads4_seed1339/args.json

@@ -1 +1 @@
-{"hellaswag": true, "attention_kind": "selective", "log_dir": "wider_is_better_9/attention_kindselective_n_heads4_seed1339", "resume_checkpoint": null, "resume_optimizer": false, "add_a_head": false, "add_head_to_start": true, "new_head_init": "normal", "n_heads": 4, "protect_bos_token": true, "prevent_from_masking_myself": true, "max_steps": 17500, "warmup_steps": 200, "group": "wider_is_better_9", "use_wandb": true, "kill_self_after_run": false, "random_seed": 1339, "memory_penalty_epsilon": 0.1, "selection_head_linear_combo": "none", "selection_head_linear_combo_scale": 1.0, "protection_kind": "none", "leaky_relu_alpha": null, "leaky_relu_bias": null, "use_compile": true, "use_mini_model": false, "upload_to_hf": true, "seq_len": 256, "batch_size": 120, "total_batch_size": 30720, "protection_head_scaling_factor": 1.0, "protection_head_bias": 0.0, "n_sliced_masks": null, "n_latent_masks": null, "mask_layernorm": false, "residual_attention_masks": false, "compute_base_shapes": false, "base_shapes_savefile": null, "mup": true, "disable_selection": false, "mup_enable_coord_check_logging": false, "max_lr": 4e-05, "decay_lr": true, "readout_zero_init": false, "query_zero_init": false, "l1_loss": false, "debugpy": false, "key": "4e-5_30720_4_1339", "n_embd": 256}
+{"hellaswag": true, "attention_kind": "selective", "log_dir": "wider_is_better_9/attention_kindselective_n_heads4_seed1339", "resume_checkpoint": null, "resume_optimizer": false, "add_a_head": false, "add_head_to_start": true, "new_head_init": "normal", "n_heads": 4, "protect_bos_token": true, "prevent_from_masking_myself": true, "max_steps": 8750, "warmup_steps": 500, "group": "wider_is_better_9", "use_wandb": true, "kill_self_after_run": false, "random_seed": 1339, "memory_penalty_epsilon": 0.1, "selection_head_linear_combo": "none", "selection_head_linear_combo_scale": 1.0, "protection_kind": "none", "leaky_relu_alpha": null, "leaky_relu_bias": null, "use_compile": true, "use_mini_model": false, "upload_to_hf": true, "seq_len": 256, "batch_size": 120, "total_batch_size": 61440, "protection_head_scaling_factor": 1.0, "protection_head_bias": 0.0, "n_sliced_masks": null, "n_latent_masks": null, "mask_layernorm": false, "residual_attention_masks": false, "compute_base_shapes": false, "base_shapes_savefile": null, "mup": true, "disable_selection": false, "mup_enable_coord_check_logging": false, "max_lr": 5e-05, "decay_lr": true, "readout_zero_init": false, "query_zero_init": false, "l1_loss": false, "debugpy": false, "key": "5e-5_61440_4_1339", "n_embd": 256}

attention_kindselective_n_heads4_seed1339/dataloader_08749.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:953385078aa3787b69fc6857dfd48b0a2cd2f4d27c6f8892e01211aca53d07f5
+size 964

attention_kindselective_n_heads4_seed1339/log2.txt

@@ -1,1056 +1,414 @@
-max_steps: 17500
-max_steps: 17500
-0 val loss 11.2810
-0 val perplexity 79298.4922
-0 val loss 11.2810
-0 val perplexity 79298.4922
-0 train 11.295413 (lr=1.8750e-07) (hash(x)=150724848)
-0 train 11.295413 (lr=2.0000e-07) (hash(x)=150724848)
-100 val loss 9.8920
-100 val perplexity 19770.6621
-100 train 9.940850 (lr=1.8937e-05) (hash(x)=149217061)
-100 val loss 9.8892
-100 val perplexity 19716.9434
-100 train 9.938035 (lr=2.0200e-05) (hash(x)=149217061)
-200 val loss 8.7035
-200 val perplexity 6023.8257
-200 train 8.722006 (lr=3.7500e-05) (hash(x)=149910534)
-200 val loss 8.5860
-200 val perplexity 5356.3530
-200 train 8.602024 (lr=4.0000e-05) (hash(x)=149910534)
-300 val loss 7.9287
-300 val perplexity 2775.8965
-300 train 7.848921 (lr=3.7497e-05) (hash(x)=142185643)
-300 val loss 7.7731
-300 val perplexity 2375.8701
-300 train 7.680683 (lr=3.9997e-05) (hash(x)=142185643)
-400 val loss 7.6983
-400 val perplexity 2204.5137
-400 train 7.749911 (lr=3.7489e-05) (hash(x)=148123706)
-400 val loss 7.6116
-400 val perplexity 2021.4500
-400 train 7.655191 (lr=3.9988e-05) (hash(x)=148123706)
-500 val loss 7.6214
-500 val perplexity 2041.4889
-500 train 7.503228 (lr=3.7475e-05) (hash(x)=149952383)
-500 val loss 7.5217
-500 val perplexity 1847.7963
-500 train 7.393625 (lr=3.9973e-05) (hash(x)=149952383)
-600 val loss 7.5742
-600 val perplexity 1947.2408
-600 train 7.401229 (lr=3.7456e-05) (hash(x)=146678221)
-600 val loss 7.4720
-600 val perplexity 1758.0408
-600 train 7.277638 (lr=3.9953e-05) (hash(x)=146678221)
-700 val loss 7.5268
-700 val perplexity 1857.0763
-700 train 7.560060 (lr=3.7430e-05) (hash(x)=156180736)
-700 val loss 7.3901
-700 val perplexity 1619.8132
-700 train 7.425849 (lr=3.9926e-05) (hash(x)=156180736)
-800 val loss 7.4827
-800 val perplexity 1776.9448
-800 train 7.479188 (lr=3.7400e-05) (hash(x)=151700982)
-800 val loss 7.3359
-800 val perplexity 1534.3604
-800 train 7.325125 (lr=3.9893e-05) (hash(x)=151700982)
-900 val loss 7.4518
-900 val perplexity 1723.0034
-900 train 7.249286 (lr=3.7364e-05) (hash(x)=147288467)
-900 val loss 7.2879
-900 val perplexity 1462.5081
-900 train 7.073882 (lr=3.9855e-05) (hash(x)=147288467)
-1000 val loss 7.4266
-1000 val perplexity 1680.1207
-1000 train 7.467032 (lr=3.7322e-05) (hash(x)=156182087)
-1000 val loss 7.2358
-1000 val perplexity 1388.1892
-1000 train 7.278591 (lr=3.9810e-05) (hash(x)=156182087)
-1100 val loss 7.4096
-1100 val perplexity 1651.7021
-1100 train 7.287074 (lr=3.7275e-05) (hash(x)=147861550)
-1100 val loss 7.1888
-1100 val perplexity 1324.5430
-1100 train 7.058900 (lr=3.9760e-05) (hash(x)=147861550)
-1200 val loss 7.3875
-1200 val perplexity 1615.6384
-1200 train 7.260944 (lr=3.7223e-05) (hash(x)=149318660)
-1200 val loss 7.1559
-1200 val perplexity 1281.6619
-1200 train 7.013008 (lr=3.9704e-05) (hash(x)=149318660)
-1300 val loss 7.3797
-1300 val perplexity 1603.1160
-1300 train 7.235526 (lr=3.7164e-05) (hash(x)=151631103)
-1300 val loss 7.1329
-1300 val perplexity 1252.5103
-1300 train 6.970014 (lr=3.9642e-05) (hash(x)=151631103)
-1400 val loss 7.3629
-1400 val perplexity 1576.4673
-1400 train 7.414872 (lr=3.7101e-05) (hash(x)=150482428)
-1400 val loss 7.0961
-1400 val perplexity 1207.2863
-1400 train 7.131195 (lr=3.9574e-05) (hash(x)=150482428)
-1500 val loss 7.3347
-1500 val perplexity 1532.5674
-1500 train 7.194934 (lr=3.7032e-05) (hash(x)=144458617)
-1500 val loss 7.0334
-1500 val perplexity 1133.9139
-1500 train 6.878065 (lr=3.9501e-05) (hash(x)=144458617)
-1600 val loss 7.3127
-1600 val perplexity 1499.2800
-1600 train 7.187172 (lr=3.6958e-05) (hash(x)=143268605)
-1600 val loss 6.9786
-1600 val perplexity 1073.4591
-1600 train 6.860932 (lr=3.9421e-05) (hash(x)=143268605)
-1700 val loss 7.2833
-1700 val perplexity 1455.7445
-1700 train 7.218442 (lr=3.6878e-05) (hash(x)=150817011)
-1700 val loss 6.9268
-1700 val perplexity 1019.2236
-1700 train 6.862553 (lr=3.9336e-05) (hash(x)=150817011)
-1800 val loss 7.2652
-1800 val perplexity 1429.6526
-1800 train 7.249866 (lr=3.6793e-05) (hash(x)=152322423)
-1800 val loss 6.8768
-1800 val perplexity 969.4940
-1800 train 6.858002 (lr=3.9246e-05) (hash(x)=152322423)
-1900 val loss 7.2479
-1900 val perplexity 1405.1208
-1900 train 7.074307 (lr=3.6702e-05) (hash(x)=143574126)
-1900 val loss 6.8296
-1900 val perplexity 924.8268
-1900 train 6.618206 (lr=3.9149e-05) (hash(x)=143574126)
-2000 val loss 7.2321
-2000 val perplexity 1383.1353
-2000 train 7.012033 (lr=3.6606e-05) (hash(x)=147904298)
-2000 val loss 6.7887
-2000 val perplexity 887.7144
-2000 train 6.533892 (lr=3.9047e-05) (hash(x)=147904298)
-2100 val loss 7.2056
-2100 val perplexity 1346.9266
-2100 train 6.963788 (lr=3.6505e-05) (hash(x)=144719400)
-2100 val loss 6.7451
-2100 val perplexity 849.8469
-2100 train 6.479913 (lr=3.8939e-05) (hash(x)=144719400)
-2200 val loss 7.1795
-2200 val perplexity 1312.2140
-2200 train 7.416286 (lr=3.6399e-05) (hash(x)=154343147)
-2200 val loss 6.6894
-2200 val perplexity 803.8664
-2200 train 6.935144 (lr=3.8826e-05) (hash(x)=154343147)
-2300 val loss 7.1358
-2300 val perplexity 1256.1078
-2300 train 7.187479 (lr=3.6288e-05) (hash(x)=154372635)
-2300 val loss 6.6321
-2300 val perplexity 759.0529
-2300 train 6.670909 (lr=3.8707e-05) (hash(x)=154372635)
-2400 val loss 7.1059
-2400 val perplexity 1219.1160
-2400 train 6.941741 (lr=3.6171e-05) (hash(x)=141843115)
-2400 val loss 6.5962
-2400 val perplexity 732.2883
-2400 train 6.420857 (lr=3.8583e-05) (hash(x)=141843115)
-2500 val loss 7.0782
-2500 val perplexity 1185.8218
-2500 train 6.975012 (lr=3.6049e-05) (hash(x)=149417679)
-2500 val loss 6.5524
-2500 val perplexity 700.9512
-2500 train 6.445707 (lr=3.8453e-05) (hash(x)=149417679)
-2600 val loss 7.0639
-2600 val perplexity 1168.9380
-2600 train 6.928061 (lr=3.5923e-05) (hash(x)=145279030)
-2600 val loss 6.5181
-2600 val perplexity 677.3092
-2600 train 6.392206 (lr=3.8317e-05) (hash(x)=145279030)
-2700 val loss 7.0370
-2700 val perplexity 1137.9823
-2700 train 6.938586 (lr=3.5791e-05) (hash(x)=146363911)
-2700 val loss 6.4870
-2700 val perplexity 656.5254
-2700 train 6.406978 (lr=3.8177e-05) (hash(x)=146363911)
-2800 val loss 7.0138
-2800 val perplexity 1111.8834
-2800 train 6.982116 (lr=3.5654e-05) (hash(x)=152507639)
-2800 val loss 6.4852
-2800 val perplexity 655.3987
-2800 train 6.433135 (lr=3.8031e-05) (hash(x)=152507639)
-2900 val loss 6.9887
-2900 val perplexity 1084.3630
-2900 train 6.780720 (lr=3.5512e-05) (hash(x)=139296191)
-2900 val loss 6.4438
-2900 val perplexity 628.7961
-2900 train 6.226076 (lr=3.7879e-05) (hash(x)=139296191)
-3000 val loss 6.9635
-3000 val perplexity 1057.3408
-3000 train 6.976216 (lr=3.5365e-05) (hash(x)=148473774)
-3000 val loss 6.4230
-3000 val perplexity 615.8414
-3000 train 6.430891 (lr=3.7723e-05) (hash(x)=148473774)
-3100 val loss 6.9340
-3100 val perplexity 1026.5642
-3100 train 7.038289 (lr=3.5214e-05) (hash(x)=154130527)
-3100 val loss 6.3778
-3100 val perplexity 588.6588
-3100 train 6.464539 (lr=3.7561e-05) (hash(x)=154130527)
-3200 val loss 6.8995
-3200 val perplexity 991.7739
-3200 train 6.837024 (lr=3.5057e-05) (hash(x)=151117002)
-3200 val loss 6.3583
-3200 val perplexity 577.2902
-3200 train 6.312740 (lr=3.7394e-05) (hash(x)=151117002)
-3300 val loss 6.8909
-3300 val perplexity 983.2383
-3300 train 7.017479 (lr=3.4896e-05) (hash(x)=151035383)
-3300 val loss 6.3573
-3300 val perplexity 576.6871
-3300 train 6.524789 (lr=3.7222e-05) (hash(x)=151035383)
-3400 val loss 6.8506
-3400 val perplexity 944.4594
-3400 train 6.938754 (lr=3.4730e-05) (hash(x)=150614749)
-3400 val loss 6.3145
-3400 val perplexity 552.5436
-3400 train 6.451954 (lr=3.7045e-05) (hash(x)=150614749)
-3500 val loss 6.8218
-3500 val perplexity 917.6775
-3500 train 6.866987 (lr=3.4560e-05) (hash(x)=151431391)
-3500 val loss 6.2900
-3500 val perplexity 539.1597
-3500 train 6.274175 (lr=3.6864e-05) (hash(x)=151431391)
-3600 val loss 6.7947
-3600 val perplexity 893.1364
-3600 train 6.717680 (lr=3.4384e-05) (hash(x)=151140797)
-3600 val loss 6.2696
-3600 val perplexity 528.2466
-3600 train 6.207747 (lr=3.6677e-05) (hash(x)=151140797)
-3700 val loss 6.7724
-3700 val perplexity 873.3726
-3700 train 6.998176 (lr=3.4205e-05) (hash(x)=154608340)
-3700 val loss 6.2443
-3700 val perplexity 515.0886
-3700 train 6.412389 (lr=3.6485e-05) (hash(x)=154608340)
-3800 val loss 6.7447
-3800 val perplexity 849.5219
-3800 train 6.773592 (lr=3.4021e-05) (hash(x)=169641309)
-3800 val loss 6.2278
-3800 val perplexity 506.6567
-3800 train 6.265309 (lr=3.6289e-05) (hash(x)=169641309)
-3900 val loss 6.7102
-3900 val perplexity 820.7698
-3900 train 6.665777 (lr=3.3832e-05) (hash(x)=150592479)
-3900 val loss 6.2083
-3900 val perplexity 496.8782
-3900 train 6.183831 (lr=3.6088e-05) (hash(x)=150592479)
-4000 val loss 6.6884
-4000 val perplexity 803.0477
-4000 train 6.668771 (lr=3.3639e-05) (hash(x)=155495317)
-4000 val loss 6.1923
-4000 val perplexity 488.9843
-4000 train 6.184135 (lr=3.5882e-05) (hash(x)=155495317)
-4100 val loss 6.6582
-4100 val perplexity 779.1608
-4100 train 6.861329 (lr=3.3442e-05) (hash(x)=170323970)
-4100 val loss 6.1827
-4100 val perplexity 484.3083
-4100 train 6.371322 (lr=3.5671e-05) (hash(x)=170323970)
-4200 val loss 6.6526
-4200 val perplexity 774.7971
-4200 train 6.458316 (lr=3.3240e-05) (hash(x)=140854360)
-4200 val loss 6.1780
-4200 val perplexity 482.0302
-4200 train 6.021700 (lr=3.5457e-05) (hash(x)=140854360)
-4300 val loss 6.6006
-4300 val perplexity 735.5699
-4300 train 6.599814 (lr=3.3035e-05) (hash(x)=148926644)
-4300 val loss 6.1474
-4300 val perplexity 467.4936
-4300 train 6.165276 (lr=3.5237e-05) (hash(x)=148926644)
-4400 val loss 6.5684
-4400 val perplexity 712.2296
-4400 train 6.605387 (lr=3.2825e-05) (hash(x)=149997871)
-4400 val loss 6.1302
-4400 val perplexity 459.5324
-4400 train 6.172115 (lr=3.5014e-05) (hash(x)=149997871)
-4500 val loss 6.5417
-4500 val perplexity 693.4786
-4500 train 6.497378 (lr=3.2611e-05) (hash(x)=149219289)
-4500 val loss 6.1173
-4500 val perplexity 453.6170
-4500 train 6.054480 (lr=3.4786e-05) (hash(x)=149219289)
-4600 val loss 6.5138
-4600 val perplexity 674.4045
-4600 train 6.310131 (lr=3.2394e-05) (hash(x)=139836197)
-4600 val loss 6.0995
-4600 val perplexity 445.6441
-4600 train 5.907494 (lr=3.4553e-05) (hash(x)=139836197)
-4700 val loss 6.4972
-4700 val perplexity 663.3148
-4700 train 6.451049 (lr=3.2172e-05) (hash(x)=150722883)
-4700 val loss 6.0911
-4700 val perplexity 441.9263
-4700 train 6.046262 (lr=3.4317e-05) (hash(x)=150722883)
-4800 val loss 6.4677
-4800 val perplexity 644.0198
-4800 train 6.666210 (lr=3.1947e-05) (hash(x)=166864648)
-4800 val loss 6.0826
-4800 val perplexity 438.1559
-4800 train 6.288001 (lr=3.4077e-05) (hash(x)=166864648)
-4900 val loss 6.4483
-4900 val perplexity 631.6287
-4900 train 6.556377 (lr=3.1718e-05) (hash(x)=157746743)
-4900 val loss 6.0648
-4900 val perplexity 430.4201
-4900 train 6.170938 (lr=3.3832e-05) (hash(x)=157746743)
-5000 val loss 6.4205
-5000 val perplexity 614.2855
-5000 train 6.371796 (lr=3.1485e-05) (hash(x)=153032257)
-5000 val loss 6.0495
-5000 val perplexity 423.8893
-5000 train 5.990033 (lr=3.3584e-05) (hash(x)=153032257)
-5100 val loss 6.4025
-5100 val perplexity 603.3539
-5100 train 6.482961 (lr=3.1249e-05) (hash(x)=151862371)
-5100 val loss 6.0405
-5100 val perplexity 420.1125
-5100 train 6.099496 (lr=3.3332e-05) (hash(x)=151862371)
-5200 val loss 6.3816
-5200 val perplexity 590.8616
-5200 train 6.701185 (lr=3.1009e-05) (hash(x)=168862995)
-5200 val loss 6.0248
-5200 val perplexity 413.5569
-5200 train 6.326960 (lr=3.3076e-05) (hash(x)=168862995)
-5300 val loss 6.3669
-5300 val perplexity 582.2534
-5300 train 6.339127 (lr=3.0766e-05) (hash(x)=148243567)
-5300 val loss 6.0171
-5300 val perplexity 410.3859
-5300 train 5.976335 (lr=3.2817e-05) (hash(x)=148243567)
-5400 val loss 6.3465
-5400 val perplexity 570.4892
-5400 train 6.265824 (lr=3.0519e-05) (hash(x)=145381742)
-5400 val loss 6.0088
-5400 val perplexity 407.0121
-5400 train 5.949846 (lr=3.2554e-05) (hash(x)=145381742)
-5500 val loss 6.3343
-5500 val perplexity 563.5786
-5500 train 6.492792 (lr=3.0269e-05) (hash(x)=155602174)
-5500 val loss 5.9979
-5500 val perplexity 402.5925
-5500 train 6.146212 (lr=3.2287e-05) (hash(x)=155602174)
-5600 val loss 6.3026
-5600 val perplexity 545.9857
-5600 train 6.393070 (lr=3.0016e-05) (hash(x)=160637672)
-5600 val loss 5.9769
-5600 val perplexity 394.2003
-5600 train 6.044743 (lr=3.2017e-05) (hash(x)=160637672)
-5700 val loss 6.2884
-5700 val perplexity 538.2720
-5700 train 6.213086 (lr=2.9760e-05) (hash(x)=152101131)
-5700 val loss 5.9691
-5700 val perplexity 391.1516
-5700 train 5.917782 (lr=3.1744e-05) (hash(x)=152101131)
-5800 val loss 6.2701
-5800 val perplexity 528.5106
-5800 train 6.232066 (lr=2.9501e-05) (hash(x)=152327089)
-5800 val loss 5.9564
-5800 val perplexity 386.2349
-5800 train 5.918398 (lr=3.1468e-05) (hash(x)=152327089)
-5900 val loss 6.2712
-5900 val perplexity 529.1312
-5900 train 6.214243 (lr=2.9239e-05) (hash(x)=151094543)
-5900 val loss 5.9514
-5900 val perplexity 384.3059
-5900 train 5.916299 (lr=3.1188e-05) (hash(x)=151094543)
-6000 val loss 6.2522
-6000 val perplexity 519.1394
-6000 train 6.358566 (lr=2.8974e-05) (hash(x)=160233315)
-6000 val loss 5.9423
-6000 val perplexity 380.8282
-6000 train 6.028115 (lr=3.0905e-05) (hash(x)=160233315)
-6100 val loss 6.2356
-6100 val perplexity 510.5935
-6100 train 6.335859 (lr=2.8706e-05) (hash(x)=158295337)
-6100 val loss 5.9316
-6100 val perplexity 376.7630
-6100 train 6.034167 (lr=3.0620e-05) (hash(x)=158295337)
-6200 val loss 6.2173
-6200 val perplexity 501.3618
-6200 train 6.177991 (lr=2.8436e-05) (hash(x)=151269817)
-6200 val loss 5.9158
-6200 val perplexity 370.8564
-6200 train 5.867013 (lr=3.0332e-05) (hash(x)=151269817)
-6300 val loss 6.1963
-6300 val perplexity 490.9506
-6300 train 6.339870 (lr=2.8163e-05) (hash(x)=156912756)
-6300 val loss 5.9049
-6300 val perplexity 366.8421
-6300 train 6.066171 (lr=3.0040e-05) (hash(x)=156912756)
-6400 val loss 6.1899
-6400 val perplexity 487.8059
-6400 train 6.127525 (lr=2.7888e-05) (hash(x)=125779637)
-6400 val loss 5.8971
-6400 val perplexity 363.9658
-6400 train 5.816467 (lr=2.9747e-05) (hash(x)=125779637)
-6500 val loss 6.1775
-6500 val perplexity 481.8019
-6500 train 6.068864 (lr=2.7610e-05) (hash(x)=148157661)
-6500 val loss 5.8918
-6500 val perplexity 362.0443
-6500 train 5.793604 (lr=2.9450e-05) (hash(x)=148157661)
-6600 val loss 6.1667
-6600 val perplexity 476.6075
-6600 train 6.197901 (lr=2.7330e-05) (hash(x)=152820049)
-6600 val loss 5.8860
-6600 val perplexity 359.9784
-6600 train 5.888762 (lr=2.9152e-05) (hash(x)=152820049)
-6700 val loss 6.1553
-6700 val perplexity 471.1977
-6700 train 6.099213 (lr=2.7047e-05) (hash(x)=155818717)
-6700 val loss 5.8753
-6700 val perplexity 356.1459
-6700 train 5.810640 (lr=2.8851e-05) (hash(x)=155818717)
-6800 val loss 6.1344
-6800 val perplexity 461.4436
-6800 train 6.172260 (lr=2.6763e-05) (hash(x)=158833881)
-6800 val loss 5.8571
-6800 val perplexity 349.6964
-6800 train 5.852377 (lr=2.8547e-05) (hash(x)=158833881)
-6900 val loss 6.1326
-6900 val perplexity 460.6177
-6900 train 6.066556 (lr=2.6477e-05) (hash(x)=164984909)
-6900 val loss 5.8510
-6900 val perplexity 347.5731
-6900 train 5.733030 (lr=2.8242e-05) (hash(x)=164984909)
-7000 val loss 6.1168
-7000 val perplexity 453.4012
-7000 train 6.016676 (lr=2.6188e-05) (hash(x)=151172493)
-7000 val loss 5.8428
-7000 val perplexity 344.7496
-7000 train 5.743183 (lr=2.7934e-05) (hash(x)=151172493)
-7100 val loss 6.1157
-7100 val perplexity 452.9291
-7100 train 6.155096 (lr=2.5898e-05) (hash(x)=154023633)
-7100 val loss 5.8395
-7100 val perplexity 343.6040
-7100 train 5.904043 (lr=2.7624e-05) (hash(x)=154023633)
-7200 val loss 6.0983
-7200 val perplexity 445.1168
-7200 train 5.955569 (lr=2.5606e-05) (hash(x)=139351250)
-7200 val loss 5.8291
-7200 val perplexity 340.0587
-7200 train 5.676873 (lr=2.7313e-05) (hash(x)=139351250)
-7300 val loss 6.0909
-7300 val perplexity 441.8268
-7300 train 6.039002 (lr=2.5312e-05) (hash(x)=158719458)
-7300 val loss 5.8239
-7300 val perplexity 338.2815
-7300 train 5.789658 (lr=2.7000e-05) (hash(x)=158719458)
-7400 val loss 6.0725
-7400 val perplexity 433.7473
-7400 train 6.140794 (lr=2.5017e-05) (hash(x)=162250810)
-7400 val loss 5.8076
-7400 val perplexity 332.8254
-7400 train 5.877717 (lr=2.6685e-05) (hash(x)=162250810)
-7500 val loss 6.0607
-7500 val perplexity 428.6643
-7500 train 5.937808 (lr=2.4721e-05) (hash(x)=142827557)
-7500 val loss 5.7964
-7500 val perplexity 329.1051
-7500 train 5.685374 (lr=2.6369e-05) (hash(x)=142827557)
-7600 val loss 6.0534
-7600 val perplexity 425.5662
-7600 train 5.953993 (lr=2.4423e-05) (hash(x)=147830801)
-7600 val loss 5.7925
-7600 val perplexity 327.8361
-7600 train 5.709580 (lr=2.6051e-05) (hash(x)=147830801)
-7700 val loss 6.0466
-7700 val perplexity 422.6890
-7700 train 6.135414 (lr=2.4124e-05) (hash(x)=165233314)
-7700 val loss 5.7853
-7700 val perplexity 325.4878
-7700 train 5.887728 (lr=2.5732e-05) (hash(x)=165233314)
-7800 val loss 6.0379
-7800 val perplexity 418.9930
-7800 train 6.028669 (lr=2.3823e-05) (hash(x)=160654717)
-7800 val loss 5.7776
-7800 val perplexity 322.9735
-7800 train 5.765421 (lr=2.5411e-05) (hash(x)=160654717)
-7900 val loss 6.0338
-7900 val perplexity 417.2910
-7900 train 5.882531 (lr=2.3522e-05) (hash(x)=140712586)
-7900 val loss 5.7761
-7900 val perplexity 322.4857
-7900 train 5.653329 (lr=2.5090e-05) (hash(x)=140712586)
-8000 val loss 6.0164
-8000 val perplexity 410.0987
-8000 train 6.180608 (lr=2.3219e-05) (hash(x)=150309061)
-8000 val loss 5.7610
-8000 val perplexity 317.6568
-8000 train 5.930829 (lr=2.4767e-05) (hash(x)=150309061)
-8100 val loss 6.0151
-8100 val perplexity 409.5820
-8100 train 6.010431 (lr=2.2916e-05) (hash(x)=158805466)
-8100 val loss 5.7605
-8100 val perplexity 317.4916
-8100 train 5.754591 (lr=2.4444e-05) (hash(x)=158805466)
-8200 val loss 5.9996
-8200 val perplexity 403.2486
-8200 train 5.881342 (lr=2.2612e-05) (hash(x)=142814078)
-8200 val loss 5.7481
-8200 val perplexity 313.6090
-8200 train 5.645060 (lr=2.4120e-05) (hash(x)=142814078)
-8300 val loss 5.9939
-8300 val perplexity 400.9659
-8300 train 5.953493 (lr=2.2308e-05) (hash(x)=148340417)
-8300 val loss 5.7425
-8300 val perplexity 311.8304
-8300 train 5.710927 (lr=2.3795e-05) (hash(x)=148340417)
-8400 val loss 5.9912
-8400 val perplexity 399.8812
-8400 train 5.927708 (lr=2.2002e-05) (hash(x)=157132639)
-8400 val loss 5.7362
-8400 val perplexity 309.8906
-8400 train 5.664308 (lr=2.3469e-05) (hash(x)=157132639)
-8500 val loss 5.9836
-8500 val perplexity 396.8719
-8500 train 5.911011 (lr=2.1697e-05) (hash(x)=152994240)
-8500 val loss 5.7317
-8500 val perplexity 308.4876
-8500 train 5.661629 (lr=2.3143e-05) (hash(x)=152994240)
-8600 val loss 5.9692
-8600 val perplexity 391.2051
-8600 train 5.927732 (lr=2.1391e-05) (hash(x)=152210154)
-8600 val loss 5.7182
-8600 val perplexity 304.3698
-8600 train 5.687449 (lr=2.2817e-05) (hash(x)=152210154)
-8700 val loss 5.9625
-8700 val perplexity 388.5811
-8700 train 6.040589 (lr=2.1085e-05) (hash(x)=149631247)
-8700 val loss 5.7136
-8700 val perplexity 302.9607
-8700 train 5.818038 (lr=2.2490e-05) (hash(x)=149631247)
-8800 val loss 5.9546
-8800 val perplexity 385.5366
-8800 train 5.811369 (lr=2.0778e-05) (hash(x)=145617784)
-8800 val loss 5.7087
-8800 val perplexity 301.4867
-8800 train 5.592268 (lr=2.2163e-05) (hash(x)=145617784)
-8900 val loss 5.9494
-8900 val perplexity 383.5336
-8900 train 6.043489 (lr=2.0472e-05) (hash(x)=155984970)
-8900 val loss 5.7001
-8900 val perplexity 298.8987
-8900 train 5.810061 (lr=2.1837e-05) (hash(x)=155984970)
-9000 val loss 5.9442
-9000 val perplexity 381.5475
-9000 train 5.677837 (lr=2.0165e-05) (hash(x)=143433013)
-9000 val loss 5.6995
-9000 val perplexity 298.7101
-9000 train 5.443996 (lr=2.1510e-05) (hash(x)=143433013)
-9100 val loss 5.9373
-9100 val perplexity 378.9221
-9100 train 5.827633 (lr=1.9859e-05) (hash(x)=152507533)
-9100 val loss 5.6936
-9100 val perplexity 296.9637
-9100 train 5.598932 (lr=2.1183e-05) (hash(x)=152507533)
-9200 val loss 5.9295
-9200 val perplexity 375.9717
-9200 train 6.067265 (lr=1.9553e-05) (hash(x)=162400650)
-9200 val loss 5.6854
-9200 val perplexity 294.5436
-9200 train 5.809351 (lr=2.0857e-05) (hash(x)=162400650)
-9300 val loss 5.9234
-9300 val perplexity 373.6941
-9300 train 5.969053 (lr=1.9248e-05) (hash(x)=148419068)
-9300 val loss 5.6822
-9300 val perplexity 293.5955
-9300 train 5.727042 (lr=2.0531e-05) (hash(x)=148419068)
-9400 val loss 5.9160
-9400 val perplexity 370.9360
-9400 train 5.881715 (lr=1.8942e-05) (hash(x)=145025169)
-9400 val loss 5.6712
-9400 val perplexity 290.3748
-9400 train 5.642601 (lr=2.0205e-05) (hash(x)=145025169)
-9500 val loss 5.9054
-9500 val perplexity 367.0152
-9500 train 5.877307 (lr=1.8638e-05) (hash(x)=145104235)
-9500 val loss 5.6654
-9500 val perplexity 288.7047
-9500 train 5.641552 (lr=1.9880e-05) (hash(x)=145104235)
-9600 val loss 5.9024
-9600 val perplexity 365.9068
-9600 train 5.836945 (lr=1.8334e-05) (hash(x)=150499797)
-9600 val loss 5.6634
-9600 val perplexity 288.1191
-9600 train 5.599575 (lr=1.9556e-05) (hash(x)=150499797)
-9700 val loss 5.8969
-9700 val perplexity 363.9092
-9700 train 5.695178 (lr=1.8031e-05) (hash(x)=139690022)
-9700 val loss 5.6585
-9700 val perplexity 286.7106
-9700 train 5.472332 (lr=1.9233e-05) (hash(x)=139690022)
-9800 val loss 5.8880
-9800 val perplexity 360.6846
-9800 train 5.884799 (lr=1.7728e-05) (hash(x)=145645279)
-9800 val loss 5.6541
-9800 val perplexity 285.4582
-9800 train 5.667852 (lr=1.8910e-05) (hash(x)=145645279)
-9900 val loss 5.8780
-9900 val perplexity 357.1094
-9900 train 5.797610 (lr=1.7427e-05) (hash(x)=150103432)
-9900 val loss 5.6451
-9900 val perplexity 282.8907
-9900 train 5.576068 (lr=1.8589e-05) (hash(x)=150103432)
-10000 val loss 5.8753
-10000 val perplexity 356.1355
-10000 train 6.028806 (lr=1.7126e-05) (hash(x)=162309473)
-10000 val loss 5.6387
-10000 val perplexity 281.0946
-10000 train 5.791598 (lr=1.8268e-05) (hash(x)=162309473)
-10100 val loss 5.8695
-10100 val perplexity 354.0836
-10100 train 5.753899 (lr=1.6827e-05) (hash(x)=146586772)
-10100 val loss 5.6350
-10100 val perplexity 280.0534
-10100 train 5.533384 (lr=1.7949e-05) (hash(x)=146586772)
-10200 val loss 5.8623
-10200 val perplexity 351.5154
-10200 train 5.765655 (lr=1.6529e-05) (hash(x)=149256313)
-10200 val loss 5.6319
-10200 val perplexity 279.1962
-10200 train 5.533719 (lr=1.7631e-05) (hash(x)=149256313)
-10300 val loss 5.8580
-10300 val perplexity 350.0123
-10300 train 5.819844 (lr=1.6233e-05) (hash(x)=151455083)
-10300 val loss 5.6270
-10300 val perplexity 277.8158
-10300 train 5.576078 (lr=1.7315e-05) (hash(x)=151455083)
-10400 val loss 5.8561
-10400 val perplexity 349.3732
-10400 train 5.640420 (lr=1.5938e-05) (hash(x)=150210444)
-10400 val loss 5.6245
-10400 val perplexity 277.1461
-10400 train 5.374650 (lr=1.7000e-05) (hash(x)=150210444)
-10500 val loss 5.8610
-10500 val perplexity 351.0689
-10500 train 5.633465 (lr=1.5644e-05) (hash(x)=152144428)
-10500 val loss 5.6257
-10500 val perplexity 277.4727
-10500 train 5.393909 (lr=1.6687e-05) (hash(x)=152144428)
-10600 val loss 5.8443
-10600 val perplexity 345.2592
-10600 train 5.903284 (lr=1.5352e-05) (hash(x)=168628935)
-10600 val loss 5.6154
-10600 val perplexity 274.6116
-10600 train 5.629866 (lr=1.6376e-05) (hash(x)=168628935)
-10700 val loss 5.8357
-10700 val perplexity 342.3191
-10700 train 5.653769 (lr=1.5062e-05) (hash(x)=164667785)
-10700 val loss 5.6089
-10700 val perplexity 272.8553
-10700 train 5.377524 (lr=1.6066e-05) (hash(x)=164667785)
-10800 val loss 5.8284
-10800 val perplexity 339.8024
-10800 train 5.784979 (lr=1.4773e-05) (hash(x)=143621491)
-10800 val loss 5.6017
-10800 val perplexity 270.8960
-10800 train 5.585045 (lr=1.5758e-05) (hash(x)=143621491)
-10900 val loss 5.8282
-10900 val perplexity 339.7538
-10900 train 5.716103 (lr=1.4487e-05) (hash(x)=143914826)
-10900 val loss 5.5983
-10900 val perplexity 269.9630
-10900 train 5.495556 (lr=1.5453e-05) (hash(x)=143914826)
-11000 val loss 5.8235
-11000 val perplexity 338.1537
-11000 train 5.596159 (lr=1.4203e-05) (hash(x)=137663435)
-11000 val loss 5.5962
-11000 val perplexity 269.4052
-11000 train 5.388721 (lr=1.5149e-05) (hash(x)=137663435)
-11100 val loss 5.8186
-11100 val perplexity 336.5119
-11100 train 5.695492 (lr=1.3920e-05) (hash(x)=149054007)
-11100 val loss 5.5939
-11100 val perplexity 268.7805
-11100 train 5.476272 (lr=1.4848e-05) (hash(x)=149054007)
-11200 val loss 5.8212
-11200 val perplexity 337.3617
-11200 train 5.827100 (lr=1.3640e-05) (hash(x)=153473924)
-11200 val loss 5.5936
-11200 val perplexity 268.7033
-11200 train 5.615970 (lr=1.4550e-05) (hash(x)=153473924)
-11300 val loss 5.8178
-11300 val perplexity 336.2411
-11300 train 5.608507 (lr=1.3362e-05) (hash(x)=158563275)
-11300 val loss 5.5922
-11300 val perplexity 268.3133
-11300 train 5.359133 (lr=1.4253e-05) (hash(x)=158563275)
-11400 val loss 5.8136
-11400 val perplexity 334.8136
-11400 train 5.946541 (lr=1.3087e-05) (hash(x)=161016667)
-11400 val loss 5.5860
-11400 val perplexity 266.6766
-11400 train 5.716009 (lr=1.3960e-05) (hash(x)=161016667)
-11500 val loss 5.8071
-11500 val perplexity 332.6627
-11500 train 5.853693 (lr=1.2814e-05) (hash(x)=157344822)
-11500 val loss 5.5861
-11500 val perplexity 266.6995
-11500 train 5.636026 (lr=1.3668e-05) (hash(x)=157344822)
-11600 val loss 5.7953
-11600 val perplexity 328.7356
-11600 train 5.679436 (lr=1.2544e-05) (hash(x)=139977713)
-11600 val loss 5.5719
-11600 val perplexity 262.9367
-11600 train 5.476897 (lr=1.3380e-05) (hash(x)=139977713)
-11700 val loss 5.7915
-11700 val perplexity 327.4971
-11700 train 5.782972 (lr=1.2276e-05) (hash(x)=156511738)
-11700 val loss 5.5674
-11700 val perplexity 261.7418
-11700 train 5.542856 (lr=1.3095e-05) (hash(x)=156511738)
-11800 val loss 5.7884
-11800 val perplexity 326.4805
-11800 train 6.230186 (lr=1.2011e-05) (hash(x)=185372558)
-11800 val loss 5.5674
-11800 val perplexity 261.7533
-11800 train 5.948181 (lr=1.2812e-05) (hash(x)=185372558)
-11900 val loss 5.7832
-11900 val perplexity 324.7906
-11900 train 5.621591 (lr=1.1749e-05) (hash(x)=138085470)
-11900 val loss 5.5627
-11900 val perplexity 260.5332
-11900 train 5.431148 (lr=1.2532e-05) (hash(x)=138085470)
-12000 val loss 5.7834
-12000 val perplexity 324.8578
-12000 train 5.635805 (lr=1.1490e-05) (hash(x)=142928235)
-12000 val loss 5.5626
-12000 val perplexity 260.5064
-12000 train 5.406928 (lr=1.2256e-05) (hash(x)=142928235)
-12100 val loss 5.7828
-12100 val perplexity 324.6824
-12100 train 5.632250 (lr=1.1234e-05) (hash(x)=151960396)
-12100 val loss 5.5622
-12100 val perplexity 260.4002
-12100 train 5.407411 (lr=1.1983e-05) (hash(x)=151960396)
-12200 val loss 5.7718
-12200 val perplexity 321.1224
-12200 train 5.828484 (lr=1.0981e-05) (hash(x)=156794523)
-12200 val loss 5.5532
-12200 val perplexity 258.0563
-12200 train 5.590004 (lr=1.1713e-05) (hash(x)=156794523)
-12300 val loss 5.7668
-12300 val perplexity 319.5281
-12300 train 5.579047 (lr=1.0731e-05) (hash(x)=138082358)
-12300 val loss 5.5463
-12300 val perplexity 256.2923
-12300 train 5.355970 (lr=1.1446e-05) (hash(x)=138082358)
-12400 val loss 5.7622
-12400 val perplexity 318.0525
-12400 train 5.666190 (lr=1.0484e-05) (hash(x)=155255899)
-12400 val loss 5.5444
-12400 val perplexity 255.8121
-12400 train 5.450579 (lr=1.1183e-05) (hash(x)=155255899)
-12500 val loss 5.7579
-12500 val perplexity 316.6952
-12500 train 5.723560 (lr=1.0241e-05) (hash(x)=150490528)
-12500 val loss 5.5400
-12500 val perplexity 254.6694
-12500 train 5.493838 (lr=1.0924e-05) (hash(x)=150490528)
-12600 val loss 5.7566
-12600 val perplexity 316.2633
-12600 train 5.901272 (lr=1.0001e-05) (hash(x)=154831027)
-12600 val loss 5.5394
-12600 val perplexity 254.5151
-12600 train 5.718703 (lr=1.0668e-05) (hash(x)=154831027)
-12700 val loss 5.7545
-12700 val perplexity 315.5970
-12700 train 5.704981 (lr=9.7649e-06) (hash(x)=147341052)
-12700 val loss 5.5385
-12700 val perplexity 254.3070
-12700 train 5.490286 (lr=1.0416e-05) (hash(x)=147341052)
-12800 val loss 5.7534
-12800 val perplexity 315.2712
-12800 train 5.785439 (lr=9.5322e-06) (hash(x)=159333795)
-12800 val loss 5.5359
-12800 val perplexity 253.6327
-12800 train 5.563532 (lr=1.0168e-05) (hash(x)=159333795)
-12900 val loss 5.7550
-12900 val perplexity 315.7724
-12900 train 5.598507 (lr=9.3031e-06) (hash(x)=143896867)
-12900 val loss 5.5382
-12900 val perplexity 254.2111
-12900 train 5.371953 (lr=9.9233e-06) (hash(x)=143896867)
-13000 val loss 5.7436
-13000 val perplexity 312.1739
-13000 train 5.751026 (lr=9.0777e-06) (hash(x)=147161537)
-13000 val loss 5.5275
-13000 val perplexity 251.5187
-13000 train 5.549591 (lr=9.6829e-06) (hash(x)=147161537)
-13100 val loss 5.7422
-13100 val perplexity 311.7380
-13100 train 5.870570 (lr=8.8562e-06) (hash(x)=147081918)
-13100 val loss 5.5249
-13100 val perplexity 250.8569
-13100 train 5.663218 (lr=9.4466e-06) (hash(x)=147081918)
-13200 val loss 5.7385
-13200 val perplexity 310.5829
-13200 train 5.503904 (lr=8.6385e-06) (hash(x)=145079287)
-13200 val loss 5.5241
-13200 val perplexity 250.6500
-13200 train 5.294759 (lr=9.2144e-06) (hash(x)=145079287)
-13300 val loss 5.7375
-13300 val perplexity 310.2796
-13300 train 5.542725 (lr=8.4248e-06) (hash(x)=142515041)
-13300 val loss 5.5217
-13300 val perplexity 250.0569
-13300 train 5.322534 (lr=8.9865e-06) (hash(x)=142515041)
-13400 val loss 5.7343
-13400 val perplexity 309.3112
-13400 train 5.524352 (lr=8.2151e-06) (hash(x)=144081985)
-13400 val loss 5.5212
-13400 val perplexity 249.9334
-13400 train 5.311021 (lr=8.7628e-06) (hash(x)=144081985)
-13500 val loss 5.7319
-13500 val perplexity 308.5533
-13500 train 5.817440 (lr=8.0095e-06) (hash(x)=149957292)
-13500 val loss 5.5183
-13500 val perplexity 249.2086
-13500 train 5.610835 (lr=8.5435e-06) (hash(x)=149957292)
-13600 val loss 5.7283
-13600 val perplexity 307.4504
-13600 train 5.761745 (lr=7.8081e-06) (hash(x)=157123033)
-13600 val loss 5.5120
-13600 val perplexity 247.6491
-13600 train 5.547890 (lr=8.3286e-06) (hash(x)=157123033)
-13700 val loss 5.7234
-13700 val perplexity 305.9554
-13700 train 5.703632 (lr=7.6109e-06) (hash(x)=148277247)
-13700 val loss 5.5094
-13700 val perplexity 247.0001
-13700 train 5.489895 (lr=8.1183e-06) (hash(x)=148277247)
-13800 val loss 5.7274
-13800 val perplexity 307.1811
-13800 train 5.850340 (lr=7.4180e-06) (hash(x)=142769907)
-13800 val loss 5.5152
-13800 val perplexity 248.4417
-13800 train 5.671909 (lr=7.9125e-06) (hash(x)=142769907)
-13900 val loss 5.7188
-13900 val perplexity 304.5252
-13900 train 5.651237 (lr=7.2294e-06) (hash(x)=144932004)
-13900 val loss 5.5049
-13900 val perplexity 245.9041
-13900 train 5.479329 (lr=7.7114e-06) (hash(x)=144932004)
-14000 val loss 5.7157
-14000 val perplexity 303.5841
-14000 train 5.726858 (lr=7.0452e-06) (hash(x)=149532265)
-14000 val loss 5.5043
-14000 val perplexity 245.7397
-14000 train 5.515040 (lr=7.5149e-06) (hash(x)=149532265)
-14100 val loss 5.7150
-14100 val perplexity 303.3718
-14100 train 5.543785 (lr=6.8656e-06) (hash(x)=149301374)
-14100 val loss 5.5016
-14100 val perplexity 245.0862
-14100 train 5.308578 (lr=7.3233e-06) (hash(x)=149301374)
-14200 val loss 5.7140
-14200 val perplexity 303.0953
-14200 train 5.461953 (lr=6.6904e-06) (hash(x)=142871075)
-14200 val loss 5.5020
-14200 val perplexity 245.1914
-14200 train 5.235135 (lr=7.1365e-06) (hash(x)=142871075)
-14300 val loss 5.7094
-14300 val perplexity 301.6765
-14300 train 5.865572 (lr=6.5199e-06) (hash(x)=166744960)
-14300 val loss 5.4964
-14300 val perplexity 243.8035
-14300 train 5.635840 (lr=6.9546e-06) (hash(x)=166744960)
-14400 val loss 5.7051
-14400 val perplexity 300.4096
-14400 train 5.674760 (lr=6.3540e-06) (hash(x)=145950920)
-14400 val loss 5.4931
-14400 val perplexity 243.0033
-14400 train 5.448152 (lr=6.7776e-06) (hash(x)=145950920)
-14500 val loss 5.7030
-14500 val perplexity 299.7748
-14500 train 5.690929 (lr=6.1928e-06) (hash(x)=141089155)
-14500 val loss 5.4908
-14500 val perplexity 242.4427
-14500 train 5.497694 (lr=6.6057e-06) (hash(x)=141089155)
-14600 val loss 5.7002
-14600 val perplexity 298.9186
-14600 train 5.852167 (lr=6.0364e-06) (hash(x)=174516275)
-14600 val loss 5.4892
-14600 val perplexity 242.0709
-14600 train 5.642091 (lr=6.4389e-06) (hash(x)=174516275)
-14700 val loss 5.6985
-14700 val perplexity 298.4080
-14700 train 5.502239 (lr=5.8848e-06) (hash(x)=137896065)
-14700 val loss 5.4878
-14700 val perplexity 241.7357
-14700 train 5.310824 (lr=6.2771e-06) (hash(x)=137896065)
-14800 val loss 5.6982
-14800 val perplexity 298.3280
-14800 train 5.728738 (lr=5.7381e-06) (hash(x)=154396556)
-14800 val loss 5.4868
-14800 val perplexity 241.4946
-14800 train 5.521058 (lr=6.1206e-06) (hash(x)=154396556)
-14900 val loss 5.6981
-14900 val perplexity 298.3020
-14900 train 5.544933 (lr=5.5962e-06) (hash(x)=149322394)
-14900 val loss 5.4872
-14900 val perplexity 241.5863
-14900 train 5.323401 (lr=5.9693e-06) (hash(x)=149322394)
-15000 val loss 5.6981
-15000 val perplexity 298.3115
-15000 train 5.610722 (lr=5.4593e-06) (hash(x)=147410513)
-15000 val loss 5.4872
-15000 val perplexity 241.5842
-15000 train 5.404994 (lr=5.8233e-06) (hash(x)=147410513)
-15100 val loss 5.6947
-15100 val perplexity 297.2931
-15100 train 5.604572 (lr=5.3275e-06) (hash(x)=145245634)
-15100 val loss 5.4839
-15100 val perplexity 240.7724
-15100 train 5.395303 (lr=5.6826e-06) (hash(x)=145245634)
-15200 val loss 5.6903
-15200 val perplexity 295.9903
-15200 train 5.774277 (lr=5.2006e-06) (hash(x)=148341108)
-15200 val loss 5.4795
-15200 val perplexity 239.7191
-15200 train 5.560233 (lr=5.5473e-06) (hash(x)=148341108)
-15300 val loss 5.6883
-15300 val perplexity 295.3842
-15300 train 5.497147 (lr=5.0789e-06) (hash(x)=140162437)
-15300 val loss 5.4766
-15300 val perplexity 239.0419
-15300 train 5.292723 (lr=5.4175e-06) (hash(x)=140162437)
-15400 val loss 5.6865
-15400 val perplexity 294.8504
-15400 train 5.889404 (lr=4.9622e-06) (hash(x)=142166081)
-15400 val loss 5.4757
-15400 val perplexity 238.8262
-15400 train 5.706107 (lr=5.2931e-06) (hash(x)=142166081)
-15500 val loss 5.6855
-15500 val perplexity 294.5707
-15500 train 5.590559 (lr=4.8508e-06) (hash(x)=141042196)
-15500 val loss 5.4749
-15500 val perplexity 238.6198
-15500 train 5.418190 (lr=5.1742e-06) (hash(x)=141042196)
-15600 val loss 5.6839
-15600 val perplexity 294.1076
-15600 train 5.804760 (lr=4.7445e-06) (hash(x)=148672701)
-15600 val loss 5.4752
-15600 val perplexity 238.6886
-15600 train 5.604007 (lr=5.0608e-06) (hash(x)=148672701)
-15700 val loss 5.6840
-15700 val perplexity 294.1362
-15700 train 5.586970 (lr=4.6435e-06) (hash(x)=149672385)
-15700 val loss 5.4744
-15700 val perplexity 238.5025
-15700 train 5.381526 (lr=4.9531e-06) (hash(x)=149672385)
-15800 val loss 5.6840
-15800 val perplexity 294.1187
-15800 train 5.705602 (lr=4.5478e-06) (hash(x)=152379562)
-15800 val loss 5.4739
-15800 val perplexity 238.3925
-15800 train 5.498224 (lr=4.8509e-06) (hash(x)=152379562)
-15900 val loss 5.6794
-15900 val perplexity 292.7766
-15900 train 5.719800 (lr=4.4573e-06) (hash(x)=154674265)
-15900 val loss 5.4694
-15900 val perplexity 237.3285
-15900 train 5.510931 (lr=4.7545e-06) (hash(x)=154674265)
-16000 val loss 5.6778
-16000 val perplexity 292.3029
-16000 train 5.719006 (lr=4.3722e-06) (hash(x)=150368078)
-16000 val loss 5.4672
-16000 val perplexity 236.8024
-16000 train 5.519518 (lr=4.6637e-06) (hash(x)=150368078)
-16100 val loss 5.6746
-16100 val perplexity 291.3651
-16100 train 5.784335 (lr=4.2924e-06) (hash(x)=160390561)
-16100 val loss 5.4654
-16100 val perplexity 236.3686
-16100 train 5.580384 (lr=4.5786e-06) (hash(x)=160390561)
-16200 val loss 5.6741
-16200 val perplexity 291.2354
-16200 train 6.027804 (lr=4.2180e-06) (hash(x)=155218378)
-16200 val loss 5.4652
-16200 val perplexity 236.3124
-16200 train 5.828511 (lr=4.4993e-06) (hash(x)=155218378)
-16300 val loss 5.6723
-16300 val perplexity 290.6990
-16300 train 5.863964 (lr=4.1491e-06) (hash(x)=160891077)
-16300 val loss 5.4638
-16300 val perplexity 235.9971
-16300 train 5.650701 (lr=4.4257e-06) (hash(x)=160891077)
-16400 val loss 5.6722
-16400 val perplexity 290.6846
-16400 train 5.577696 (lr=4.0856e-06) (hash(x)=142789253)
-16400 val loss 5.4630
-16400 val perplexity 235.7978
-16400 train 5.364003 (lr=4.3579e-06) (hash(x)=142789253)
-16500 val loss 5.6721
-16500 val perplexity 290.6511
-16500 train 5.416285 (lr=4.0275e-06) (hash(x)=141497841)
-16500 val loss 5.4626
-16500 val perplexity 235.7060
-16500 train 5.210999 (lr=4.2960e-06) (hash(x)=141497841)
-16600 val loss 5.6708
-16600 val perplexity 290.2732
-16600 train 5.635384 (lr=3.9749e-06) (hash(x)=157947018)
-16600 val loss 5.4621
-16600 val perplexity 235.5873
-16600 train 5.423563 (lr=4.2399e-06) (hash(x)=157947018)
-16700 val loss 5.6669
-16700 val perplexity 289.1373
-16700 train 5.618238 (lr=3.9278e-06) (hash(x)=146039309)
-16700 val loss 5.4583
-16700 val perplexity 234.6897
-16700 train 5.432078 (lr=4.1896e-06) (hash(x)=146039309)
-16800 val loss 5.6668
-16800 val perplexity 289.1097
-16800 train 5.634438 (lr=3.8862e-06) (hash(x)=156381445)
-16800 val loss 5.4569
-16800 val perplexity 234.3673
-16800 train 5.423639 (lr=4.1452e-06) (hash(x)=156381445)
-16900 val loss 5.6643
-16900 val perplexity 288.3940
-16900 train 5.624663 (lr=3.8501e-06) (hash(x)=146537354)
-16900 val loss 5.4553
-16900 val perplexity 233.9896
-16900 train 5.410178 (lr=4.1067e-06) (hash(x)=146537354)
-17000 val loss 5.6638
-17000 val perplexity 288.2406
-17000 train 5.582037 (lr=3.8195e-06) (hash(x)=135939444)
-17000 val loss 5.4545
-17000 val perplexity 233.8000
-17000 train 5.407039 (lr=4.0741e-06) (hash(x)=135939444)
-17100 val loss 5.6635
-17100 val perplexity 288.1500
-17100 train 5.736537 (lr=3.7945e-06) (hash(x)=147819652)
-17100 val loss 5.4544
-17100 val perplexity 233.7901
-17100 train 5.548837 (lr=4.0475e-06) (hash(x)=147819652)
-17200 val loss 5.6636
-17200 val perplexity 288.1846
-17200 train 5.764005 (lr=3.7750e-06) (hash(x)=155307909)
-17200 val loss 5.4544
-17200 val perplexity 233.7908
-17200 train 5.525397 (lr=4.0267e-06) (hash(x)=155307909)
-17300 val loss 5.6608
-17300 val perplexity 287.3757
-17300 train 5.615438 (lr=3.7611e-06) (hash(x)=151879548)
-17300 val loss 5.4510
-17300 val perplexity 233.0012
-17300 train 5.396317 (lr=4.0119e-06) (hash(x)=151879548)
-17400 val loss 5.6585
-17400 val perplexity 286.7057
-17400 train 5.578130 (lr=3.7528e-06) (hash(x)=143641792)
-17400 val loss 5.4501
-17400 val perplexity 232.7875
-17400 train 5.374997 (lr=4.0030e-06) (hash(x)=143641792)
-17499 val loss 5.6564
-17499 val perplexity 286.1226
-17499 val loss 5.4482
-17499 val perplexity 232.3446
+max_steps: 8750
+1900 val loss 6.4870
+1900 val perplexity 656.5473
+1900 train 6.404645 (lr=3.7502e-05) (hash(x)=141769419)
+0 val loss 11.2808
+0 val perplexity 79287.6797
+2000 val loss 6.4397
+2000 val perplexity 626.2369
+2000 train 6.376182 (lr=3.7143e-05) (hash(x)=151963443)
+0 train 11.289297 (lr=1.0000e-07) (hash(x)=150724848)
+2100 val loss 6.4155
+2100 val perplexity 611.2753
+2100 train 6.466409 (lr=3.6761e-05) (hash(x)=162947470)
+100 val loss 10.0094
+100 val perplexity 22235.4102
+100 train 10.023370 (lr=1.0100e-05) (hash(x)=149910534)
+2200 val loss 6.3645
+2200 val perplexity 580.8438
+2200 train 6.606410 (lr=3.6358e-05) (hash(x)=154954810)
+200 val loss 9.2733
+200 val perplexity 10649.9316
+200 train 9.301048 (lr=2.0100e-05) (hash(x)=148123706)
+2300 val loss 6.3335
+2300 val perplexity 563.1455
+2300 train 6.322995 (lr=3.5935e-05) (hash(x)=151878111)
+300 val loss 8.1598
+300 val perplexity 3497.6423
+300 train 8.051448 (lr=3.0100e-05) (hash(x)=146678221)
+2400 val loss 6.3098
+2400 val perplexity 549.9563
+2400 train 6.204768 (lr=3.5491e-05) (hash(x)=158661057)
+400 val loss 7.6335
+400 val perplexity 2066.3218
+400 train 7.528817 (lr=4.0100e-05) (hash(x)=151700982)
+2500 val loss 6.2701
+2500 val perplexity 528.5074
+2500 train 6.296215 (lr=3.5027e-05) (hash(x)=150925584)
+500 val loss 7.4756
+500 val perplexity 1764.5261
+500 train 7.565072 (lr=5.0000e-05) (hash(x)=156182087)
+2600 val loss 6.2500
+2600 val perplexity 518.0140
+2600 train 6.152835 (lr=3.4545e-05) (hash(x)=144515755)
+600 val loss 7.3815
+600 val perplexity 1605.9897
+600 train 7.367438 (lr=4.9984e-05) (hash(x)=149318660)
+2700 val loss 6.2234
+2700 val perplexity 504.4197
+2700 train 6.215466 (lr=3.4044e-05) (hash(x)=153109144)
+700 val loss 7.3039
+700 val perplexity 1486.1003
+700 train 7.343727 (lr=4.9935e-05) (hash(x)=150482428)
+2800 val loss 6.1948
+2800 val perplexity 490.2085
+2800 train 6.096365 (lr=3.3526e-05) (hash(x)=151152897)
+800 val loss 7.1792
+800 val perplexity 1311.8173
+800 train 7.056084 (lr=4.9853e-05) (hash(x)=143268605)
+2900 val loss 6.1718
+2900 val perplexity 479.0310
+2900 train 6.109528 (lr=3.2992e-05) (hash(x)=145800210)
+900 val loss 7.0813
+900 val perplexity 1189.4648
+900 train 7.047098 (lr=4.9739e-05) (hash(x)=152322423)
+3000 val loss 6.1559
+3000 val perplexity 471.4690
+3000 train 5.950811 (lr=3.2441e-05) (hash(x)=141997485)
+1000 val loss 7.0176
+1000 val perplexity 1116.0995
+1000 train 6.884840 (lr=4.9593e-05) (hash(x)=147904298)
+3100 val loss 6.1280
+3100 val perplexity 458.5026
+3100 train 6.036367 (lr=3.1875e-05) (hash(x)=154049740)
+1100 val loss 6.9427
+1100 val perplexity 1035.5698
+1100 train 7.138206 (lr=4.9415e-05) (hash(x)=154343147)
+3200 val loss 6.1142
+3200 val perplexity 452.2266
+3200 train 6.052361 (lr=3.1295e-05) (hash(x)=150471842)
+1200 val loss 6.8632
+1200 val perplexity 956.4412
+1200 train 6.851167 (lr=4.9205e-05) (hash(x)=141843115)
+3300 val loss 6.0974
+3300 val perplexity 444.6876
+3300 train 6.039181 (lr=3.0702e-05) (hash(x)=149048126)
+1300 val loss 6.8038
+1300 val perplexity 901.3065
+1300 train 6.687198 (lr=4.8964e-05) (hash(x)=145279030)
+3400 val loss 6.0719
+3400 val perplexity 433.5064
+3400 train 6.197778 (lr=3.0095e-05) (hash(x)=161261339)
+1400 val loss 6.7539
+1400 val perplexity 857.3593
+1400 train 6.705647 (lr=4.8691e-05) (hash(x)=152507639)
+3500 val loss 6.0590
+3500 val perplexity 427.9403
+3500 train 6.026854 (lr=2.9477e-05) (hash(x)=157495564)
+1500 val loss 6.6972
+1500 val perplexity 810.1442
+1500 train 6.731950 (lr=4.8388e-05) (hash(x)=148473774)
+3600 val loss 6.0460
+3600 val perplexity 422.4147
+3600 train 5.944861 (lr=2.8849e-05) (hash(x)=144352932)
+1600 val loss 6.6193
+1600 val perplexity 749.4503
+1600 train 6.657555 (lr=4.8055e-05) (hash(x)=151117002)
+3700 val loss 6.0294
+3700 val perplexity 415.4617
+3700 train 6.011248 (lr=2.8210e-05) (hash(x)=149389012)
+1700 val loss 6.5588
+1700 val perplexity 705.4470
+1700 train 6.489615 (lr=4.7691e-05) (hash(x)=138011335)
+3800 val loss 6.0066
+3800 val perplexity 406.1071
+3800 train 5.947698 (lr=2.7562e-05) (hash(x)=146607620)
+1800 val loss 6.5096
+1800 val perplexity 671.5671
+1800 train 6.661233 (lr=4.7299e-05) (hash(x)=171180926)
+3900 val loss 5.9952
+3900 val perplexity 401.5043
+3900 train 5.910914 (lr=2.6907e-05) (hash(x)=143754617)
+1900 val loss 6.4539
+1900 val perplexity 635.1964
+1900 train 6.375162 (lr=4.6878e-05) (hash(x)=141769419)
+4000 val loss 5.9766
+4000 val perplexity 394.0837
+4000 train 5.954946 (lr=2.6244e-05) (hash(x)=156930722)
+2000 val loss 6.4109
+2000 val perplexity 608.4214
+2000 train 6.352161 (lr=4.6428e-05) (hash(x)=151963443)
+4100 val loss 5.9643
+4100 val perplexity 389.2686
+4100 train 5.786150 (lr=2.5575e-05) (hash(x)=147500519)
+2100 val loss 6.3943
+2100 val perplexity 598.4083
+2100 train 6.443270 (lr=4.5951e-05) (hash(x)=162947470)
+4200 val loss 5.9559
+4200 val perplexity 386.0066
+4200 train 5.804139 (lr=2.4900e-05) (hash(x)=143232237)
+2200 val loss 6.3348
+2200 val perplexity 563.8331
+2200 train 6.577147 (lr=4.5448e-05) (hash(x)=154954810)
+4300 val loss 5.9418
+4300 val perplexity 380.6191
+4300 train 5.853108 (lr=2.4222e-05) (hash(x)=146811670)
+2300 val loss 6.3036
+2300 val perplexity 546.5355
+2300 train 6.291760 (lr=4.4918e-05) (hash(x)=151878111)
+4400 val loss 5.9268
+4400 val perplexity 374.9412
+4400 train 5.956989 (lr=2.3540e-05) (hash(x)=158418746)
+2400 val loss 6.2758
+2400 val perplexity 531.5452
+2400 train 6.167399 (lr=4.4363e-05) (hash(x)=158661057)
+4500 val loss 5.9192
+4500 val perplexity 372.1174
+4500 train 5.882172 (lr=2.2856e-05) (hash(x)=156695778)
+2500 val loss 6.2377
+2500 val perplexity 511.6937
+2500 train 6.265279 (lr=4.3784e-05) (hash(x)=150925584)
+4600 val loss 5.9054
+4600 val perplexity 367.0164
+4600 train 5.931722 (lr=2.2171e-05) (hash(x)=147791497)
+2600 val loss 6.2134
+2600 val perplexity 499.3786
+2600 train 6.122252 (lr=4.3181e-05) (hash(x)=144515755)
+4700 val loss 5.8916
+4700 val perplexity 361.9958
+4700 train 5.770754 (lr=2.1486e-05) (hash(x)=155533088)
+2700 val loss 6.1906
+2700 val perplexity 488.1596
+2700 train 6.177955 (lr=4.2555e-05) (hash(x)=153109144)
+4800 val loss 5.8876
+4800 val perplexity 360.5260
+4800 train 5.742024 (lr=2.0801e-05) (hash(x)=138350044)
+2800 val loss 6.1570
+2800 val perplexity 472.0003
+2800 train 6.059934 (lr=4.1908e-05) (hash(x)=151152897)
+4900 val loss 5.8709
+4900 val perplexity 354.5802
+4900 train 5.874714 (lr=2.0118e-05) (hash(x)=143735284)
+2900 val loss 6.1354
+2900 val perplexity 461.9008
+2900 train 6.076121 (lr=4.1240e-05) (hash(x)=145800210)
+5000 val loss 5.8621
+5000 val perplexity 351.4452
+5000 train 5.823290 (lr=1.9438e-05) (hash(x)=154976463)
+3000 val loss 6.1252
+3000 val perplexity 457.2171
+3000 train 5.920308 (lr=4.0551e-05) (hash(x)=141997485)
+5100 val loss 5.8545
+5100 val perplexity 348.8145
+5100 train 5.741332 (lr=1.8762e-05) (hash(x)=149894982)
+3100 val loss 6.0910
+3100 val perplexity 441.8601
+3100 train 6.004350 (lr=3.9844e-05) (hash(x)=154049740)
+5200 val loss 5.8469
+5200 val perplexity 346.1438
+5200 train 5.637977 (lr=1.8090e-05) (hash(x)=159326689)
+3200 val loss 6.0723
+3200 val perplexity 433.6939
+3200 train 6.009145 (lr=3.9119e-05) (hash(x)=150471842)
+5300 val loss 5.8343
+5300 val perplexity 341.8247
+5300 train 5.844418 (lr=1.7424e-05) (hash(x)=159484800)
+3300 val loss 6.0593
+3300 val perplexity 428.0873
+3300 train 6.000336 (lr=3.8377e-05) (hash(x)=149048126)
+5400 val loss 5.8222
+5400 val perplexity 337.7180
+5400 train 5.717206 (lr=1.6765e-05) (hash(x)=140385615)
+3400 val loss 6.0372
+3400 val perplexity 418.7006
+3400 train 6.162151 (lr=3.7619e-05) (hash(x)=161261339)
+5500 val loss 5.8150
+5500 val perplexity 335.2764
+5500 train 5.650219 (lr=1.6113e-05) (hash(x)=148498335)
+3500 val loss 6.0272
+3500 val perplexity 414.5575
+3500 train 5.989676 (lr=3.6847e-05) (hash(x)=157495564)
+5600 val loss 5.8132
+5600 val perplexity 334.6795
+5600 train 5.548153 (lr=1.5469e-05) (hash(x)=151907614)
+3600 val loss 6.0075
+3600 val perplexity 406.4727
+3600 train 5.910073 (lr=3.6061e-05) (hash(x)=144352932)
+5700 val loss 5.8038
+5700 val perplexity 331.5614
+5700 train 5.771846 (lr=1.4836e-05) (hash(x)=155192267)
+3700 val loss 5.9869
+3700 val perplexity 398.1840
+3700 train 5.958940 (lr=3.5263e-05) (hash(x)=149389012)
+5800 val loss 5.7946
+5800 val perplexity 328.5269
+5800 train 5.841525 (lr=1.4212e-05) (hash(x)=153132158)
+3800 val loss 5.9682
+3800 val perplexity 390.7858
+3800 train 5.906829 (lr=3.4453e-05) (hash(x)=146607620)
+5900 val loss 5.7887
+5900 val perplexity 326.5968
+5900 train 5.762638 (lr=1.3600e-05) (hash(x)=161446764)
+3900 val loss 5.9571
+3900 val perplexity 386.4784
+3900 train 5.873564 (lr=3.3633e-05) (hash(x)=143754617)
+6000 val loss 5.7856
+6000 val perplexity 325.5639
+6000 train 5.599146 (lr=1.3000e-05) (hash(x)=151512446)
+4000 val loss 5.9406
+4000 val perplexity 380.1711
+4000 train 5.916785 (lr=3.2805e-05) (hash(x)=156930722)
+6100 val loss 5.7740
+6100 val perplexity 321.8256
+6100 train 5.901221 (lr=1.2413e-05) (hash(x)=188094053)
+4100 val loss 5.9264
+4100 val perplexity 374.7864
+4100 train 5.754153 (lr=3.1968e-05) (hash(x)=147500519)
+6200 val loss 5.7654
+6200 val perplexity 319.0692
+6200 train 5.657475 (lr=1.1840e-05) (hash(x)=149389789)
+4200 val loss 5.9170
+4200 val perplexity 371.2794
+4200 train 5.758588 (lr=3.1126e-05) (hash(x)=143232237)
+6300 val loss 5.7603
+6300 val perplexity 317.4521
+6300 train 5.576626 (lr=1.1282e-05) (hash(x)=138212820)
+4300 val loss 5.8994
+4300 val perplexity 364.8297
+4300 train 5.808607 (lr=3.0277e-05) (hash(x)=146811670)
+6400 val loss 5.7599
+6400 val perplexity 317.3077
+6400 train 5.555570 (lr=1.0739e-05) (hash(x)=146535423)
+4400 val loss 5.8864
+4400 val perplexity 360.1214
+4400 train 5.923263 (lr=2.9425e-05) (hash(x)=158418746)
+6500 val loss 5.7497
+6500 val perplexity 314.1065
+6500 train 5.620624 (lr=1.0213e-05) (hash(x)=145950843)
+4500 val loss 5.8788
+4500 val perplexity 357.3871
+4500 train 5.841973 (lr=2.8571e-05) (hash(x)=156695778)
+6600 val loss 5.7452
+6600 val perplexity 312.6862
+6600 train 5.642556 (lr=9.7032e-06) (hash(x)=141162902)
+4600 val loss 5.8646
+4600 val perplexity 352.3253
+4600 train 5.884218 (lr=2.7714e-05) (hash(x)=147791497)
+6700 val loss 5.7435
+6700 val perplexity 312.1676
+6700 train 5.656245 (lr=9.2116e-06) (hash(x)=153018737)
+4700 val loss 5.8486
+4700 val perplexity 346.7375
+4700 train 5.732183 (lr=2.6857e-05) (hash(x)=155533088)
+6800 val loss 5.7361
+6800 val perplexity 309.8445
+6800 train 5.663511 (lr=8.7387e-06) (hash(x)=155640155)
+4800 val loss 5.8401
+4800 val perplexity 343.7982
+4800 train 5.699566 (lr=2.6002e-05) (hash(x)=138350044)
+6900 val loss 5.7339
+6900 val perplexity 309.1730
+6900 train 5.701223 (lr=8.2849e-06) (hash(x)=153722115)
+4900 val loss 5.8266
+4900 val perplexity 339.2005
+4900 train 5.827739 (lr=2.5148e-05) (hash(x)=143735284)
+7000 val loss 5.7269
+7000 val perplexity 307.0191
+7000 train 5.635604 (lr=7.8510e-06) (hash(x)=146953450)
+5000 val loss 5.8136
+5000 val perplexity 334.8062
+5000 train 5.771458 (lr=2.4298e-05) (hash(x)=154976463)
+7100 val loss 5.7270
+7100 val perplexity 307.0580
+7100 train 5.561845 (lr=7.4377e-06) (hash(x)=137663885)
+5100 val loss 5.8073
+5100 val perplexity 332.7099
+5100 train 5.691780 (lr=2.3452e-05) (hash(x)=149894982)
+7200 val loss 5.7176
+7200 val perplexity 304.1624
+7200 train 5.780611 (lr=7.0455e-06) (hash(x)=146172950)
+5200 val loss 5.8004
+5200 val perplexity 330.4291
+5200 train 5.584211 (lr=2.2613e-05) (hash(x)=159326689)
+7300 val loss 5.7154
+7300 val perplexity 303.5107
+7300 train 5.605722 (lr=6.6749e-06) (hash(x)=150018163)
+5300 val loss 5.7876
+5300 val perplexity 326.2238
+5300 train 5.793699 (lr=2.1780e-05) (hash(x)=159484800)
+7400 val loss 5.7112
+7400 val perplexity 302.2303
+7400 train 5.707638 (lr=6.3266e-06) (hash(x)=145351166)
+5400 val loss 5.7735
+5400 val perplexity 321.6571
+5400 train 5.678544 (lr=2.0956e-05) (hash(x)=140385615)
+7500 val loss 5.7119
+7500 val perplexity 302.4439
+7500 train 5.492382 (lr=6.0010e-06) (hash(x)=145292116)
+5500 val loss 5.7677
+5500 val perplexity 319.7973
+5500 train 5.603406 (lr=2.0141e-05) (hash(x)=148498335)
+7600 val loss 5.7052
+7600 val perplexity 300.4154
+7600 train 5.625059 (lr=5.6986e-06) (hash(x)=150235132)
+5600 val loss 5.7657
+5600 val perplexity 319.1591
+5600 train 5.504023 (lr=1.9337e-05) (hash(x)=151907614)
+7700 val loss 5.7019
+7700 val perplexity 299.4479
+7700 train 5.611199 (lr=5.4198e-06) (hash(x)=154543455)
+5700 val loss 5.7550
+5700 val perplexity 315.7704
+5700 train 5.722241 (lr=1.8545e-05) (hash(x)=155192267)
+7800 val loss 5.6998
+7800 val perplexity 298.8089
+7800 train 5.596223 (lr=5.1650e-06) (hash(x)=142456852)
+5800 val loss 5.7419
+5800 val perplexity 311.6478
+5800 train 5.790777 (lr=1.7765e-05) (hash(x)=153132158)
+7900 val loss 5.7008
+7900 val perplexity 299.1197
+7900 train 5.484799 (lr=4.9347e-06) (hash(x)=147363479)
+5900 val loss 5.7398
+5900 val perplexity 311.0003
+5900 train 5.717348 (lr=1.7000e-05) (hash(x)=161446764)
+8000 val loss 5.6943
+8000 val perplexity 297.1720
+8000 train 5.726096 (lr=4.7291e-06) (hash(x)=156122973)
+6000 val loss 5.7350
+6000 val perplexity 309.5016
+6000 train 5.536606 (lr=1.6250e-05) (hash(x)=151512446)
+8100 val loss 5.6915
+8100 val perplexity 296.3289
+8100 train 5.735093 (lr=4.5486e-06) (hash(x)=156153179)
+6100 val loss 5.7231
+6100 val perplexity 305.8419
+6100 train 5.840291 (lr=1.5516e-05) (hash(x)=188094053)
+8200 val loss 5.6903
+8200 val perplexity 295.9858
+8200 train 5.730418 (lr=4.3933e-06) (hash(x)=146430698)
+6200 val loss 5.7147
+6200 val perplexity 303.2798
+6200 train 5.604211 (lr=1.4800e-05) (hash(x)=149389789)
+8300 val loss 5.6895
+8300 val perplexity 295.7574
+8300 train 5.553320 (lr=4.2636e-06) (hash(x)=143507257)
+6300 val loss 5.7076
+6300 val perplexity 301.1410
+6300 train 5.530816 (lr=1.4102e-05) (hash(x)=138212820)
+8400 val loss 5.6841
+8400 val perplexity 294.1438
+8400 train 5.753223 (lr=4.1596e-06) (hash(x)=166272643)
+6400 val loss 5.7063
+6400 val perplexity 300.7570
+6400 train 5.495872 (lr=1.3424e-05) (hash(x)=146535423)
+8500 val loss 5.6830
+8500 val perplexity 293.8261
+8500 train 5.585660 (lr=4.0815e-06) (hash(x)=143887848)
+6500 val loss 5.6980
+6500 val perplexity 298.2603
+6500 train 5.569312 (lr=1.2766e-05) (hash(x)=145950843)
+8600 val loss 5.6830
+8600 val perplexity 293.8326
+8600 train 5.728127 (lr=4.0294e-06) (hash(x)=156900341)
+6600 val loss 5.6921
+6600 val perplexity 296.5222
+6600 train 5.591072 (lr=1.2129e-05) (hash(x)=141162902)
+8700 val loss 5.6780
+8700 val perplexity 292.3772
+8700 train 5.916970 (lr=4.0033e-06) (hash(x)=146417632)
+8749 val loss 5.6773
+8749 val perplexity 292.1608
+6700 val loss 5.6888
+6700 val perplexity 295.5327
+6700 train 5.603293 (lr=1.1515e-05) (hash(x)=153018737)

attention_kindselective_n_heads4_seed1339/model_08749.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:939fbe38062ecac4309e19531b7fe8d16c6430ba8e1f3bd0c69d98f5fc7b83c4
+size 92843394

attention_kindselective_n_heads4_seed1339/optimizer_08749.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:3695887ae9806adc82ece1edcfc54ef490826e5da0f736ad5f1acf656aa83198
+size 179406214