Upload folder using huggingface_hub

attention_kindselective_n_heads4_seed1345/args.json

@@ -1 +1 @@
-{"hellaswag": true, "attention_kind": "selective", "log_dir": "wider_is_better_9/attention_kindselective_n_heads4_seed1345", "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": 1345, "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_1345", "n_embd": 256}
+{"hellaswag": true, "attention_kind": "selective", "log_dir": "wider_is_better_9/attention_kindselective_n_heads4_seed1345", "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": 1345, "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_1345", "n_embd": 256}

attention_kindselective_n_heads4_seed1345/dataloader_08749.pt

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

attention_kindselective_n_heads4_seed1345/log2.txt

@@ -1,1056 +1,405 @@
-max_steps: 17500
-max_steps: 17500
-0 val loss 11.2682
-0 val perplexity 78293.8047
-0 val loss 11.2682
-0 val perplexity 78293.8047
-0 train 11.260221 (lr=1.8750e-07) (hash(x)=134751525)
-0 train 11.260221 (lr=2.0000e-07) (hash(x)=134751525)
-100 val loss 9.9224
-100 val perplexity 20381.3145
-100 train 9.978566 (lr=1.8937e-05) (hash(x)=144692155)
-100 val loss 9.9026
-100 val perplexity 19981.9570
-100 train 9.958472 (lr=2.0200e-05) (hash(x)=144692155)
-200 val loss 8.7106
-200 val perplexity 6066.6841
-200 train 8.693931 (lr=3.7500e-05) (hash(x)=150358957)
-200 val loss 8.8364
-200 val perplexity 6880.2979
-200 train 8.824292 (lr=4.0000e-05) (hash(x)=150358957)
-300 val loss 7.7947
-300 val perplexity 2427.6169
-300 train 7.770412 (lr=3.7497e-05) (hash(x)=150268299)
-300 val loss 7.9373
-300 val perplexity 2799.7073
-300 train 7.926304 (lr=3.9997e-05) (hash(x)=150268299)
-400 val loss 7.5322
-400 val perplexity 1867.2103
-400 train 7.378184 (lr=3.7489e-05) (hash(x)=126319983)
-400 val loss 7.6340
-400 val perplexity 2067.3862
-400 train 7.495752 (lr=3.9988e-05) (hash(x)=126319983)
-500 val loss 7.4340
-500 val perplexity 1692.6014
-500 train 7.467863 (lr=3.7475e-05) (hash(x)=153934799)
-500 val loss 7.5062
-500 val perplexity 1819.3467
-500 train 7.542481 (lr=3.9973e-05) (hash(x)=153934799)
-600 val loss 7.3549
-600 val perplexity 1563.9165
-600 train 7.397041 (lr=3.7456e-05) (hash(x)=159305874)
-600 val loss 7.4469
-600 val perplexity 1714.4636
-600 train 7.486856 (lr=3.9953e-05) (hash(x)=159305874)
-700 val loss 7.2917
-700 val perplexity 1468.0907
-700 train 7.144725 (lr=3.7430e-05) (hash(x)=141986303)
-700 val loss 7.3974
-700 val perplexity 1631.6576
-700 train 7.242340 (lr=3.9926e-05) (hash(x)=141986303)
-800 val loss 7.2317
-800 val perplexity 1382.5919
-800 train 7.207565 (lr=3.7400e-05) (hash(x)=149855631)
-800 val loss 7.3438
-800 val perplexity 1546.6282
-800 train 7.328330 (lr=3.9893e-05) (hash(x)=149855631)
-900 val loss 7.1985
-900 val perplexity 1337.3608
-900 train 7.079328 (lr=3.7364e-05) (hash(x)=149870240)
-900 val loss 7.3186
-900 val perplexity 1508.1108
-900 train 7.203938 (lr=3.9855e-05) (hash(x)=149870240)
-1000 val loss 7.1332
-1000 val perplexity 1252.9331
-1000 train 7.201777 (lr=3.7322e-05) (hash(x)=150706760)
-1000 val loss 7.2615
-1000 val perplexity 1424.3921
-1000 train 7.330913 (lr=3.9810e-05) (hash(x)=150706760)
-1100 val loss 7.0820
-1100 val perplexity 1190.3313
-1100 train 6.918931 (lr=3.7275e-05) (hash(x)=139020621)
-1100 val loss 7.2093
-1100 val perplexity 1351.9862
-1100 train 7.051281 (lr=3.9760e-05) (hash(x)=139020621)
-1200 val loss 7.0396
-1200 val perplexity 1140.9326
-1200 train 6.899639 (lr=3.7223e-05) (hash(x)=146858163)
-1200 val loss 7.1719
-1200 val perplexity 1302.3617
-1200 train 7.031198 (lr=3.9704e-05) (hash(x)=146858163)
-1300 val loss 6.9959
-1300 val perplexity 1092.1119
-1300 train 7.059475 (lr=3.7164e-05) (hash(x)=156383408)
-1300 val loss 7.1213
-1300 val perplexity 1238.0621
-1300 train 7.194174 (lr=3.9642e-05) (hash(x)=156383408)
-1400 val loss 6.9603
-1400 val perplexity 1053.9180
-1400 train 6.890877 (lr=3.7101e-05) (hash(x)=144262072)
-1400 val loss 7.0776
-1400 val perplexity 1185.1265
-1400 train 7.007119 (lr=3.9574e-05) (hash(x)=144262072)
-1500 val loss 6.9248
-1500 val perplexity 1017.1529
-1500 train 6.774604 (lr=3.7032e-05) (hash(x)=144528048)
-1500 val loss 7.0352
-1500 val perplexity 1135.9320
-1500 train 6.889645 (lr=3.9501e-05) (hash(x)=144528048)
-1600 val loss 6.8806
-1600 val perplexity 973.1651
-1600 train 6.797646 (lr=3.6958e-05) (hash(x)=135443880)
-1600 val loss 6.9865
-1600 val perplexity 1081.8743
-1600 train 6.900005 (lr=3.9421e-05) (hash(x)=135443880)
-1700 val loss 6.8431
-1700 val perplexity 937.4041
-1700 train 6.766635 (lr=3.6878e-05) (hash(x)=145067496)
-1700 val loss 6.9683
-1700 val perplexity 1062.3906
-1700 train 6.887623 (lr=3.9336e-05) (hash(x)=145067496)
-1800 val loss 6.8003
-1800 val perplexity 898.1185
-1800 train 6.762501 (lr=3.6793e-05) (hash(x)=153147388)
-1800 val loss 6.9060
-1800 val perplexity 998.2911
-1800 train 6.871210 (lr=3.9246e-05) (hash(x)=153147388)
-1900 val loss 6.7612
-1900 val perplexity 863.6940
-1900 train 6.646096 (lr=3.6702e-05) (hash(x)=151331979)
-1900 val loss 6.8692
-1900 val perplexity 962.1613
-1900 train 6.756928 (lr=3.9149e-05) (hash(x)=151331979)
-2000 val loss 6.7419
-2000 val perplexity 847.1882
-2000 train 6.551958 (lr=3.6606e-05) (hash(x)=151448445)
-2000 val loss 6.8355
-2000 val perplexity 930.3309
-2000 train 6.634272 (lr=3.9047e-05) (hash(x)=151448445)
-2100 val loss 6.6880
-2100 val perplexity 802.7498
-2100 train 6.586328 (lr=3.6505e-05) (hash(x)=157223072)
-2100 val loss 6.8038
-2100 val perplexity 901.2855
-2100 train 6.707023 (lr=3.8939e-05) (hash(x)=157223072)
-2200 val loss 6.6446
-2200 val perplexity 768.5997
-2200 train 6.666820 (lr=3.6399e-05) (hash(x)=153161010)
-2200 val loss 6.7681
-2200 val perplexity 869.6474
-2200 train 6.781376 (lr=3.8826e-05) (hash(x)=153161010)
-2300 val loss 6.6056
-2300 val perplexity 739.1978
-2300 train 6.688467 (lr=3.6288e-05) (hash(x)=156906742)
-2300 val loss 6.7296
-2300 val perplexity 836.7805
-2300 train 6.817736 (lr=3.8707e-05) (hash(x)=156906742)
-2400 val loss 6.5766
-2400 val perplexity 718.1187
-2400 train 6.470441 (lr=3.6171e-05) (hash(x)=143227423)
-2400 val loss 6.7280
-2400 val perplexity 835.5054
-2400 train 6.619276 (lr=3.8583e-05) (hash(x)=143227423)
-2500 val loss 6.5469
-2500 val perplexity 697.0684
-2500 train 6.497835 (lr=3.6049e-05) (hash(x)=151827079)
-2500 val loss 6.6696
-2500 val perplexity 788.0593
-2500 train 6.607266 (lr=3.8453e-05) (hash(x)=151827079)
-2600 val loss 6.5227
-2600 val perplexity 680.4191
-2600 train 6.965666 (lr=3.5923e-05) (hash(x)=176373796)
-2600 val loss 6.6474
-2600 val perplexity 770.7552
-2600 train 7.123261 (lr=3.8317e-05) (hash(x)=176373796)
-2700 val loss 6.4933
-2700 val perplexity 660.6890
-2700 train 6.428716 (lr=3.5791e-05) (hash(x)=148437519)
-2700 val loss 6.6203
-2700 val perplexity 750.1442
-2700 train 6.563433 (lr=3.8177e-05) (hash(x)=148437519)
-2800 val loss 6.4604
-2800 val perplexity 639.2952
-2800 train 6.459748 (lr=3.5654e-05) (hash(x)=155989503)
-2800 val loss 6.5978
-2800 val perplexity 733.4659
-2800 train 6.597178 (lr=3.8031e-05) (hash(x)=155989503)
-2900 val loss 6.4426
-2900 val perplexity 628.0563
-2900 train 6.569895 (lr=3.5512e-05) (hash(x)=159227845)
-2900 val loss 6.5750
-2900 val perplexity 716.9130
-2900 train 6.709556 (lr=3.7879e-05) (hash(x)=159227845)
-3000 val loss 6.4088
-3000 val perplexity 607.1508
-3000 train 6.432460 (lr=3.5365e-05) (hash(x)=156507542)
-3000 val loss 6.5458
-3000 val perplexity 696.3286
-3000 train 6.572825 (lr=3.7723e-05) (hash(x)=156507542)
-3100 val loss 6.3846
-3100 val perplexity 592.6638
-3100 train 6.370049 (lr=3.5214e-05) (hash(x)=155771797)
-3100 val loss 6.5187
-3100 val perplexity 677.6685
-3100 train 6.503825 (lr=3.7561e-05) (hash(x)=155771797)
-3200 val loss 6.3627
-3200 val perplexity 579.7885
-3200 train 6.287634 (lr=3.5057e-05) (hash(x)=156078901)
-3200 val loss 6.4918
-3200 val perplexity 659.7232
-3200 train 6.411802 (lr=3.7394e-05) (hash(x)=156078901)
-3300 val loss 6.3464
-3300 val perplexity 570.4557
-3300 train 6.214350 (lr=3.4896e-05) (hash(x)=143722886)
-3300 val loss 6.4636
-3300 val perplexity 641.3925
-3300 train 6.330358 (lr=3.7222e-05) (hash(x)=143722886)
-3400 val loss 6.3244
-3400 val perplexity 558.0221
-3400 train 6.250166 (lr=3.4730e-05) (hash(x)=143793373)
-3400 val loss 6.4451
-3400 val perplexity 629.5942
-3400 train 6.376811 (lr=3.7045e-05) (hash(x)=143793373)
-3500 val loss 6.3115
-3500 val perplexity 550.8678
-3500 train 6.593259 (lr=3.4560e-05) (hash(x)=147278080)
-3500 val loss 6.4321
-3500 val perplexity 621.4998
-3500 train 6.691866 (lr=3.6864e-05) (hash(x)=147278080)
-3600 val loss 6.2929
-3600 val perplexity 540.7174
-3600 train 7.051870 (lr=3.4384e-05) (hash(x)=175299205)
-3600 val loss 6.4133
-3600 val perplexity 609.8760
-3600 train 7.152054 (lr=3.6677e-05) (hash(x)=175299205)
-3700 val loss 6.2706
-3700 val perplexity 528.8101
-3700 train 6.318336 (lr=3.4205e-05) (hash(x)=150047891)
-3700 val loss 6.3813
-3700 val perplexity 590.7053
-3700 train 6.425822 (lr=3.6485e-05) (hash(x)=150047891)
-3800 val loss 6.2556
-3800 val perplexity 520.9250
-3800 train 6.285592 (lr=3.4021e-05) (hash(x)=153419668)
-3800 val loss 6.3678
-3800 val perplexity 582.7953
-3800 train 6.395144 (lr=3.6289e-05) (hash(x)=153419668)
-3900 val loss 6.2362
-3900 val perplexity 510.8896
-3900 train 6.351056 (lr=3.3832e-05) (hash(x)=153147630)
-3900 val loss 6.3476
-3900 val perplexity 571.1245
-3900 train 6.456195 (lr=3.6088e-05) (hash(x)=153147630)
-4000 val loss 6.2256
-4000 val perplexity 505.5310
-4000 train 6.058341 (lr=3.3639e-05) (hash(x)=142428269)
-4000 val loss 6.3408
-4000 val perplexity 567.2731
-4000 train 6.177061 (lr=3.5882e-05) (hash(x)=142428269)
-4100 val loss 6.2155
-4100 val perplexity 500.4389
-4100 train 6.109726 (lr=3.3442e-05) (hash(x)=148836993)
-4100 val loss 6.3292
-4100 val perplexity 560.7189
-4100 train 6.216930 (lr=3.5671e-05) (hash(x)=148836993)
-4200 val loss 6.1891
-4200 val perplexity 487.4299
-4200 train 6.215916 (lr=3.3240e-05) (hash(x)=159557219)
-4200 val loss 6.3076
-4200 val perplexity 548.7442
-4200 train 6.353669 (lr=3.5457e-05) (hash(x)=159557219)
-4300 val loss 6.1722
-4300 val perplexity 479.2188
-4300 train 6.305759 (lr=3.3035e-05) (hash(x)=154137450)
-4300 val loss 6.3024
-4300 val perplexity 545.8607
-4300 train 6.428835 (lr=3.5237e-05) (hash(x)=154137450)
-4400 val loss 6.1598
-4400 val perplexity 473.3203
-4400 train 6.318167 (lr=3.2825e-05) (hash(x)=153336405)
-4400 val loss 6.2828
-4400 val perplexity 535.2630
-4400 train 6.430600 (lr=3.5014e-05) (hash(x)=153336405)
-4500 val loss 6.1431
-4500 val perplexity 465.5147
-4500 train 6.117867 (lr=3.2611e-05) (hash(x)=145729145)
-4500 val loss 6.2663
-4500 val perplexity 526.5511
-4500 train 6.233288 (lr=3.4786e-05) (hash(x)=145729145)
-4600 val loss 6.1397
-4600 val perplexity 463.9298
-4600 train 6.154843 (lr=3.2394e-05) (hash(x)=157789272)
-4600 val loss 6.2642
-4600 val perplexity 525.4109
-4600 train 6.261403 (lr=3.4553e-05) (hash(x)=157789272)
-4700 val loss 6.1223
-4700 val perplexity 455.9128
-4700 train 6.113758 (lr=3.2172e-05) (hash(x)=146146781)
-4700 val loss 6.2476
-4700 val perplexity 516.7608
-4700 train 6.227163 (lr=3.4317e-05) (hash(x)=146146781)
-4800 val loss 6.1178
-4800 val perplexity 453.8862
-4800 train 6.142168 (lr=3.1947e-05) (hash(x)=149454469)
-4800 val loss 6.2516
-4800 val perplexity 518.8306
-4800 train 6.256900 (lr=3.4077e-05) (hash(x)=149454469)
-4900 val loss 6.0944
-4900 val perplexity 443.3552
-4900 train 6.184642 (lr=3.1718e-05) (hash(x)=155966467)
-4900 val loss 6.2232
-4900 val perplexity 504.3185
-4900 train 6.321077 (lr=3.3832e-05) (hash(x)=155966467)
-5000 val loss 6.0871
-5000 val perplexity 440.1455
-5000 train 6.048105 (lr=3.1485e-05) (hash(x)=147184546)
-5000 val loss 6.2208
-5000 val perplexity 503.0935
-5000 train 6.176530 (lr=3.3584e-05) (hash(x)=147184546)
-5100 val loss 6.0836
-5100 val perplexity 438.5894
-5100 train 6.055791 (lr=3.1249e-05) (hash(x)=149019501)
-5100 val loss 6.2195
-5100 val perplexity 502.4680
-5100 train 6.189746 (lr=3.3332e-05) (hash(x)=149019501)
-5200 val loss 6.0647
-5200 val perplexity 430.4049
-5200 train 5.981862 (lr=3.1009e-05) (hash(x)=143521586)
-5200 val loss 6.2022
-5200 val perplexity 493.8430
-5200 train 6.108462 (lr=3.3076e-05) (hash(x)=143521586)
-5300 val loss 6.0555
-5300 val perplexity 426.4686
-5300 train 5.918194 (lr=3.0766e-05) (hash(x)=143870567)
-5300 val loss 6.1932
-5300 val perplexity 489.3878
-5300 train 6.047366 (lr=3.2817e-05) (hash(x)=143870567)
-5400 val loss 6.0455
-5400 val perplexity 422.2164
-5400 train 6.074172 (lr=3.0519e-05) (hash(x)=150658725)
-5400 val loss 6.1730
-5400 val perplexity 479.6065
-5400 train 6.186630 (lr=3.2554e-05) (hash(x)=150658725)
-5500 val loss 6.0326
-5500 val perplexity 416.8026
-5500 train 6.064359 (lr=3.0269e-05) (hash(x)=147645530)
-5500 val loss 6.1636
-5500 val perplexity 475.1183
-5500 train 6.185577 (lr=3.2287e-05) (hash(x)=147645530)
-5600 val loss 6.0319
-5600 val perplexity 416.4993
-5600 train 6.074053 (lr=3.0016e-05) (hash(x)=146088410)
-5600 val loss 6.1565
-5600 val perplexity 471.7789
-5600 train 6.184332 (lr=3.2017e-05) (hash(x)=146088410)
-5700 val loss 6.0128
-5700 val perplexity 408.6373
-5700 train 6.120058 (lr=2.9760e-05) (hash(x)=151193083)
-5700 val loss 6.1430
-5700 val perplexity 465.4441
-5700 train 6.244153 (lr=3.1744e-05) (hash(x)=151193083)
-5800 val loss 6.0042
-5800 val perplexity 405.1129
-5800 train 5.773274 (lr=2.9501e-05) (hash(x)=137474865)
-5800 val loss 6.1343
-5800 val perplexity 461.3980
-5800 train 5.897436 (lr=3.1468e-05) (hash(x)=137474865)
-5900 val loss 5.9983
-5900 val perplexity 402.7342
-5900 train 5.959085 (lr=2.9239e-05) (hash(x)=142426763)
-5900 val loss 6.1302
-5900 val perplexity 459.5471
-5900 train 6.072005 (lr=3.1188e-05) (hash(x)=142426763)
-6000 val loss 5.9785
-6000 val perplexity 394.8582
-6000 train 5.758265 (lr=2.8974e-05) (hash(x)=140492670)
-6000 val loss 6.1079
-6000 val perplexity 449.4054
-6000 train 5.878311 (lr=3.0905e-05) (hash(x)=140492670)
-6100 val loss 5.9674
-6100 val perplexity 390.4943
-6100 train 5.966070 (lr=2.8706e-05) (hash(x)=175687796)
-6100 val loss 6.1001
-6100 val perplexity 445.8966
-6100 train 6.160385 (lr=3.0620e-05) (hash(x)=175687796)
-6200 val loss 5.9568
-6200 val perplexity 386.3677
-6200 train 6.023569 (lr=2.8436e-05) (hash(x)=165841139)
-6200 val loss 6.0909
-6200 val perplexity 441.8349
-6200 train 6.148386 (lr=3.0332e-05) (hash(x)=165841139)
-6300 val loss 5.9504
-6300 val perplexity 383.9209
-6300 train 6.004202 (lr=2.8163e-05) (hash(x)=145328367)
-6300 val loss 6.0847
-6300 val perplexity 439.0967
-6300 train 6.139287 (lr=3.0040e-05) (hash(x)=145328367)
-6400 val loss 5.9501
-6400 val perplexity 383.7911
-6400 train 5.937871 (lr=2.7888e-05) (hash(x)=148213323)
-6400 val loss 6.0794
-6400 val perplexity 436.7498
-6400 train 6.061519 (lr=2.9747e-05) (hash(x)=148213323)
-6500 val loss 5.9341
-6500 val perplexity 377.6818
-6500 train 6.011780 (lr=2.7610e-05) (hash(x)=153344426)
-6500 val loss 6.0682
-6500 val perplexity 431.9211
-6500 train 6.160855 (lr=2.9450e-05) (hash(x)=153344426)
-6600 val loss 5.9266
-6600 val perplexity 374.8727
-6600 train 5.903677 (lr=2.7330e-05) (hash(x)=145840902)
-6600 val loss 6.0594
-6600 val perplexity 428.1175
-6600 train 6.020699 (lr=2.9152e-05) (hash(x)=145840902)
-6700 val loss 5.9211
-6700 val perplexity 372.8060
-6700 train 5.757521 (lr=2.7047e-05) (hash(x)=142812723)
-6700 val loss 6.0461
-6700 val perplexity 422.4782
-6700 train 5.886734 (lr=2.8851e-05) (hash(x)=142812723)
-6800 val loss 5.9144
-6800 val perplexity 370.3436
-6800 train 5.761705 (lr=2.6763e-05) (hash(x)=147420964)
-6800 val loss 6.0408
-6800 val perplexity 420.2113
-6800 train 5.889901 (lr=2.8547e-05) (hash(x)=147420964)
-6900 val loss 5.9056
-6900 val perplexity 367.0766
-6900 train 5.940935 (lr=2.6477e-05) (hash(x)=152637292)
-6900 val loss 6.0352
-6900 val perplexity 417.8958
-6900 train 6.087862 (lr=2.8242e-05) (hash(x)=152637292)
-7000 val loss 5.8875
-7000 val perplexity 360.5145
-7000 train 5.894612 (lr=2.6188e-05) (hash(x)=144728159)
-7000 val loss 6.0211
-7000 val perplexity 412.0505
-7000 train 6.019482 (lr=2.7934e-05) (hash(x)=144728159)
-7100 val loss 5.8794
-7100 val perplexity 357.5976
-7100 train 5.857129 (lr=2.5898e-05) (hash(x)=143450899)
-7100 val loss 6.0116
-7100 val perplexity 408.1296
-7100 train 5.993532 (lr=2.7624e-05) (hash(x)=143450899)
-7200 val loss 5.8710
-7200 val perplexity 354.5907
-7200 train 5.953189 (lr=2.5606e-05) (hash(x)=156229897)
-7200 val loss 6.0054
-7200 val perplexity 405.6198
-7200 train 6.080842 (lr=2.7313e-05) (hash(x)=156229897)
-7300 val loss 5.8619
-7300 val perplexity 351.3760
-7300 train 5.745842 (lr=2.5312e-05) (hash(x)=147447250)
-7300 val loss 5.9921
-7300 val perplexity 400.2460
-7300 train 5.861655 (lr=2.7000e-05) (hash(x)=147447250)
-7400 val loss 5.8641
-7400 val perplexity 352.1667
-7400 train 5.624749 (lr=2.5017e-05) (hash(x)=134836112)
-7400 val loss 5.9922
-7400 val perplexity 400.2918
-7400 train 5.735302 (lr=2.6685e-05) (hash(x)=134836112)
-7500 val loss 5.8571
-7500 val perplexity 349.7002
-7500 train 5.881787 (lr=2.4721e-05) (hash(x)=155027499)
-7500 val loss 5.9880
-7500 val perplexity 398.6122
-7500 train 6.009743 (lr=2.6369e-05) (hash(x)=155027499)
-7600 val loss 5.8399
-7600 val perplexity 343.7379
-7600 train 5.726958 (lr=2.4423e-05) (hash(x)=145744849)
-7600 val loss 5.9702
-7600 val perplexity 391.5980
-7600 train 5.859680 (lr=2.6051e-05) (hash(x)=145744849)
-7700 val loss 5.8324
-7700 val perplexity 341.1719
-7700 train 5.965559 (lr=2.4124e-05) (hash(x)=158082973)
-7700 val loss 5.9659
-7700 val perplexity 389.9156
-7700 train 6.088557 (lr=2.5732e-05) (hash(x)=158082973)
-7800 val loss 5.8256
-7800 val perplexity 338.8489
-7800 train 5.963749 (lr=2.3823e-05) (hash(x)=161538768)
-7800 val loss 5.9585
-7800 val perplexity 387.0208
-7800 train 6.099160 (lr=2.5411e-05) (hash(x)=161538768)
-7900 val loss 5.8157
-7900 val perplexity 335.5297
-7900 train 5.912204 (lr=2.3522e-05) (hash(x)=152103064)
-7900 val loss 5.9481
-7900 val perplexity 383.0178
-7900 train 6.041970 (lr=2.5090e-05) (hash(x)=152103064)
-8000 val loss 5.8101
-8000 val perplexity 333.6431
-8000 train 5.795048 (lr=2.3219e-05) (hash(x)=152409525)
-8000 val loss 5.9415
-8000 val perplexity 380.4964
-8000 train 5.935862 (lr=2.4767e-05) (hash(x)=152409525)
-8100 val loss 5.8058
-8100 val perplexity 332.2193
-8100 train 5.785800 (lr=2.2916e-05) (hash(x)=150986182)
-8100 val loss 5.9357
-8100 val perplexity 378.3020
-8100 train 5.913834 (lr=2.4444e-05) (hash(x)=150986182)
-8200 val loss 5.7943
-8200 val perplexity 328.4197
-8200 train 5.795736 (lr=2.2612e-05) (hash(x)=159555650)
-8200 val loss 5.9304
-8200 val perplexity 376.2982
-8200 train 5.935581 (lr=2.4120e-05) (hash(x)=159555650)
-8300 val loss 5.7917
-8300 val perplexity 327.5845
-8300 train 5.967180 (lr=2.2308e-05) (hash(x)=139602404)
-8300 val loss 5.9256
-8300 val perplexity 374.5074
-8300 train 6.084064 (lr=2.3795e-05) (hash(x)=139602404)
-8400 val loss 5.7826
-8400 val perplexity 324.5954
-8400 train 5.679160 (lr=2.2002e-05) (hash(x)=154432574)
-8400 val loss 5.9134
-8400 val perplexity 369.9438
-8400 train 5.807495 (lr=2.3469e-05) (hash(x)=154432574)
-8500 val loss 5.7777
-8500 val perplexity 323.0103
-8500 train 5.736130 (lr=2.1697e-05) (hash(x)=142554189)
-8500 val loss 5.9038
-8500 val perplexity 366.4180
-8500 train 5.842204 (lr=2.3143e-05) (hash(x)=142554189)
-8600 val loss 5.7696
-8600 val perplexity 320.4031
-8600 train 5.718740 (lr=2.1391e-05) (hash(x)=151300791)
-8600 val loss 5.9026
-8600 val perplexity 366.0033
-8600 train 5.853130 (lr=2.2817e-05) (hash(x)=151300791)
-8700 val loss 5.7677
-8700 val perplexity 319.8089
-8700 train 5.769819 (lr=2.1085e-05) (hash(x)=148488790)
-8700 val loss 5.8968
-8700 val perplexity 363.8705
-8700 train 5.895819 (lr=2.2490e-05) (hash(x)=148488790)
-8800 val loss 5.7722
-8800 val perplexity 321.2419
-8800 train 5.793202 (lr=2.0778e-05) (hash(x)=156797847)
-8800 val loss 5.8967
-8800 val perplexity 363.8233
-8800 train 5.924081 (lr=2.2163e-05) (hash(x)=156797847)
-8900 val loss 5.7454
-8900 val perplexity 312.7444
-8900 train 5.563432 (lr=2.0472e-05) (hash(x)=143919631)
-8900 val loss 5.8758
-8900 val perplexity 356.3207
-8900 train 5.694329 (lr=2.1837e-05) (hash(x)=143919631)
-9000 val loss 5.7429
-9000 val perplexity 311.9829
-9000 train 5.697790 (lr=2.0165e-05) (hash(x)=148706835)
-9000 val loss 5.8706
-9000 val perplexity 354.4645
-9000 train 5.813400 (lr=2.1510e-05) (hash(x)=148706835)
-9100 val loss 5.7340
-9100 val perplexity 309.2187
-9100 train 5.746788 (lr=1.9859e-05) (hash(x)=153074259)
-9100 val loss 5.8633
-9100 val perplexity 351.8923
-9100 train 5.881595 (lr=2.1183e-05) (hash(x)=153074259)
-9200 val loss 5.7314
-9200 val perplexity 308.3887
-9200 train 5.648998 (lr=1.9553e-05) (hash(x)=151127077)
-9200 val loss 5.8608
-9200 val perplexity 351.0108
-9200 train 5.790888 (lr=2.0857e-05) (hash(x)=151127077)
-9300 val loss 5.7295
-9300 val perplexity 307.8009
-9300 train 5.615701 (lr=1.9248e-05) (hash(x)=152170602)
-9300 val loss 5.8557
-9300 val perplexity 349.2050
-9300 train 5.747551 (lr=2.0531e-05) (hash(x)=152170602)
-9400 val loss 5.7279
-9400 val perplexity 307.3193
-9400 train 5.753187 (lr=1.8942e-05) (hash(x)=153874803)
-9400 val loss 5.8605
-9400 val perplexity 350.9137
-9400 train 5.890156 (lr=2.0205e-05) (hash(x)=153874803)
-9500 val loss 5.7154
-9500 val perplexity 303.4984
-9500 train 5.708488 (lr=1.8638e-05) (hash(x)=157021298)
-9500 val loss 5.8374
-9500 val perplexity 342.8917
-9500 train 5.833831 (lr=1.9880e-05) (hash(x)=157021298)
-9600 val loss 5.7075
-9600 val perplexity 301.1231
-9600 train 5.551288 (lr=1.8334e-05) (hash(x)=135620494)
-9600 val loss 5.8302
-9600 val perplexity 340.4203
-9600 train 5.669508 (lr=1.9556e-05) (hash(x)=135620494)
-9700 val loss 5.7031
-9700 val perplexity 299.7987
-9700 train 5.422754 (lr=1.8031e-05) (hash(x)=142434463)
-9700 val loss 5.8272
-9700 val perplexity 339.4169
-9700 train 5.546177 (lr=1.9233e-05) (hash(x)=142434463)
-9800 val loss 5.7007
-9800 val perplexity 299.0790
-9800 train 5.567791 (lr=1.7728e-05) (hash(x)=149075541)
-9800 val loss 5.8215
-9800 val perplexity 337.4643
-9800 train 5.689050 (lr=1.8910e-05) (hash(x)=149075541)
-9900 val loss 5.6956
-9900 val perplexity 297.5502
-9900 train 5.564990 (lr=1.7427e-05) (hash(x)=141901425)
-9900 val loss 5.8226
-9900 val perplexity 337.8496
-9900 train 5.691305 (lr=1.8589e-05) (hash(x)=141901425)
-10000 val loss 5.7064
-10000 val perplexity 300.7795
-10000 train 5.710242 (lr=1.7126e-05) (hash(x)=133062699)
-10000 val loss 5.8309
-10000 val perplexity 340.6516
-10000 train 5.793503 (lr=1.8268e-05) (hash(x)=133062699)
-10100 val loss 5.6820
-10100 val perplexity 293.5367
-10100 train 5.713012 (lr=1.6827e-05) (hash(x)=157339196)
-10100 val loss 5.8057
-10100 val perplexity 332.1971
-10100 train 5.830470 (lr=1.7949e-05) (hash(x)=157339196)
-10200 val loss 5.6760
-10200 val perplexity 291.7882
-10200 train 5.457125 (lr=1.6529e-05) (hash(x)=131082121)
-10200 val loss 5.8028
-10200 val perplexity 331.2259
-10200 train 5.567506 (lr=1.7631e-05) (hash(x)=131082121)
-10300 val loss 5.6741
-10300 val perplexity 291.2323
-10300 train 5.780905 (lr=1.6233e-05) (hash(x)=144553757)
-10300 val loss 5.7975
-10300 val perplexity 329.4636
-10300 train 5.897439 (lr=1.7315e-05) (hash(x)=144553757)
-10400 val loss 5.6704
-10400 val perplexity 290.1580
-10400 train 5.475269 (lr=1.5938e-05) (hash(x)=143398861)
-10400 val loss 5.7914
-10400 val perplexity 327.4641
-10400 train 5.582107 (lr=1.7000e-05) (hash(x)=143398861)
-10500 val loss 5.6653
-10500 val perplexity 288.6734
-10500 train 5.667819 (lr=1.5644e-05) (hash(x)=157633809)
-10500 val loss 5.7876
-10500 val perplexity 326.2262
-10500 train 5.782300 (lr=1.6687e-05) (hash(x)=157633809)
-10600 val loss 5.6602
-10600 val perplexity 287.1956
-10600 train 5.656303 (lr=1.5352e-05) (hash(x)=147294315)
-10600 val loss 5.7819
-10600 val perplexity 324.3848
-10600 train 5.784723 (lr=1.6376e-05) (hash(x)=147294315)
-10700 val loss 5.6542
-10700 val perplexity 285.4958
-10700 train 5.770455 (lr=1.5062e-05) (hash(x)=157735028)
-10700 val loss 5.7786
-10700 val perplexity 323.2980
-10700 train 5.899136 (lr=1.6066e-05) (hash(x)=157735028)
-10800 val loss 5.6488
-10800 val perplexity 283.9470
-10800 train 5.618337 (lr=1.4773e-05) (hash(x)=149846113)
-10800 val loss 5.7691
-10800 val perplexity 320.2588
-10800 train 5.736881 (lr=1.5758e-05) (hash(x)=149846113)
-10900 val loss 5.6446
-10900 val perplexity 282.7607
-10900 train 6.028375 (lr=1.4487e-05) (hash(x)=157353556)
-10900 val loss 5.7660
-10900 val perplexity 319.2617
-10900 train 6.155986 (lr=1.5453e-05) (hash(x)=157353556)
-11000 val loss 5.6426
-11000 val perplexity 282.2033
-11000 train 5.940796 (lr=1.4203e-05) (hash(x)=160672627)
-11000 val loss 5.7646
-11000 val perplexity 318.8192
-11000 train 6.067262 (lr=1.5149e-05) (hash(x)=160672627)
-11100 val loss 5.6409
-11100 val perplexity 281.7201
-11100 train 5.861447 (lr=1.3920e-05) (hash(x)=166692335)
-11100 val loss 5.7601
-11100 val perplexity 317.3687
-11100 train 5.980080 (lr=1.4848e-05) (hash(x)=166692335)
-11200 val loss 5.6315
-11200 val perplexity 279.0858
-11200 train 5.687639 (lr=1.3640e-05) (hash(x)=158899869)
-11200 val loss 5.7544
-11200 val perplexity 315.5762
-11200 train 5.817999 (lr=1.4550e-05) (hash(x)=158899869)
-11300 val loss 5.6277
-11300 val perplexity 278.0259
-11300 train 5.896500 (lr=1.3362e-05) (hash(x)=166306828)
-11300 val loss 5.7467
-11300 val perplexity 313.1556
-11300 train 6.023516 (lr=1.4253e-05) (hash(x)=166306828)
-11400 val loss 5.6253
-11400 val perplexity 277.3638
-11400 train 5.544091 (lr=1.3087e-05) (hash(x)=149545840)
-11400 val loss 5.7425
-11400 val perplexity 311.8488
-11400 train 5.663081 (lr=1.3960e-05) (hash(x)=149545840)
-11500 val loss 5.6223
-11500 val perplexity 276.5126
-11500 train 5.603643 (lr=1.2814e-05) (hash(x)=139076910)
-11500 val loss 5.7426
-11500 val perplexity 311.8870
-11500 train 5.702015 (lr=1.3668e-05) (hash(x)=139076910)
-11600 val loss 5.6191
-11600 val perplexity 275.6353
-11600 train 5.457228 (lr=1.2544e-05) (hash(x)=151871279)
-11600 val loss 5.7380
-11600 val perplexity 310.4578
-11600 train 5.567941 (lr=1.3380e-05) (hash(x)=151871279)
-11700 val loss 5.6158
-11700 val perplexity 274.7234
-11700 train 5.563904 (lr=1.2276e-05) (hash(x)=152940196)
-11700 val loss 5.7336
-11700 val perplexity 309.0824
-11700 train 5.681150 (lr=1.3095e-05) (hash(x)=152940196)
-11800 val loss 5.6118
-11800 val perplexity 273.6394
-11800 train 5.697263 (lr=1.2011e-05) (hash(x)=150754129)
-11800 val loss 5.7298
-11800 val perplexity 307.9215
-11800 train 5.819008 (lr=1.2812e-05) (hash(x)=150754129)
-11900 val loss 5.6070
-11900 val perplexity 272.3301
-11900 train 5.570013 (lr=1.1749e-05) (hash(x)=148982597)
-11900 val loss 5.7252
-11900 val perplexity 306.4939
-11900 train 5.677073 (lr=1.2532e-05) (hash(x)=148982597)
-12000 val loss 5.6033
-12000 val perplexity 271.3333
-12000 train 5.634357 (lr=1.1490e-05) (hash(x)=167798390)
-12000 val loss 5.7203
-12000 val perplexity 305.0081
-12000 train 5.760694 (lr=1.2256e-05) (hash(x)=167798390)
-12100 val loss 5.6032
-12100 val perplexity 271.2964
-12100 train 5.685317 (lr=1.1234e-05) (hash(x)=146590810)
-12100 val loss 5.7247
-12100 val perplexity 306.3307
-12100 train 5.802352 (lr=1.1983e-05) (hash(x)=146590810)
-12200 val loss 5.5975
-12200 val perplexity 269.7564
-12200 train 5.834640 (lr=1.0981e-05) (hash(x)=159875780)
-12200 val loss 5.7183
-12200 val perplexity 304.3986
-12200 train 5.944863 (lr=1.1713e-05) (hash(x)=159875780)
-12300 val loss 5.5976
-12300 val perplexity 269.7828
-12300 train 5.475709 (lr=1.0731e-05) (hash(x)=135887861)
-12300 val loss 5.7144
-12300 val perplexity 303.1967
-12300 train 5.577959 (lr=1.1446e-05) (hash(x)=135887861)
-12400 val loss 5.5885
-12400 val perplexity 267.3324
-12400 train 5.658318 (lr=1.0484e-05) (hash(x)=150484221)
-12400 val loss 5.7083
-12400 val perplexity 301.3469
-12400 train 5.766685 (lr=1.1183e-05) (hash(x)=150484221)
-12500 val loss 5.5868
-12500 val perplexity 266.8819
-12500 train 5.594129 (lr=1.0241e-05) (hash(x)=157629455)
-12500 val loss 5.7041
-12500 val perplexity 300.0880
-12500 train 5.714652 (lr=1.0924e-05) (hash(x)=157629455)
-12600 val loss 5.5830
-12600 val perplexity 265.8598
-12600 train 5.483679 (lr=1.0001e-05) (hash(x)=147308692)
-12600 val loss 5.7032
-12600 val perplexity 299.8161
-12600 train 5.604730 (lr=1.0668e-05) (hash(x)=147308692)
-12700 val loss 5.5829
-12700 val perplexity 265.8332
-12700 train 5.424228 (lr=9.7649e-06) (hash(x)=144614722)
-12700 val loss 5.6991
-12700 val perplexity 298.6121
-12700 train 5.527828 (lr=1.0416e-05) (hash(x)=144614722)
-12800 val loss 5.5785
-12800 val perplexity 264.6794
-12800 train 5.373883 (lr=9.5322e-06) (hash(x)=137633606)
-12800 val loss 5.6981
-12800 val perplexity 298.2974
-12800 train 5.482248 (lr=1.0168e-05) (hash(x)=137633606)
-12900 val loss 5.5781
-12900 val perplexity 264.5634
-12900 train 5.610073 (lr=9.3031e-06) (hash(x)=155534772)
-12900 val loss 5.6962
-12900 val perplexity 297.7232
-13000 val loss 5.5729
-13000 val perplexity 263.2035
-12900 train 5.737048 (lr=9.9233e-06) (hash(x)=155534772)
-13000 train 5.583975 (lr=9.0777e-06) (hash(x)=148501470)
-13000 val loss 5.6919
-13000 val perplexity 296.4686
-13000 train 5.698692 (lr=9.6829e-06) (hash(x)=148501470)
-13100 val loss 5.5755
-13100 val perplexity 263.8708
-13100 train 5.618188 (lr=8.8562e-06) (hash(x)=154570618)
-13200 val loss 5.5710
-13200 val perplexity 262.7007
-13200 train 5.668798 (lr=8.6385e-06) (hash(x)=154153766)
-13100 val loss 5.6918
-13100 val perplexity 296.4400
-13100 train 5.738413 (lr=9.4466e-06) (hash(x)=154570618)
-13300 val loss 5.5658
-13300 val perplexity 261.3461
-13300 train 5.324534 (lr=8.4248e-06) (hash(x)=162760888)
-13200 val loss 5.6868
-13200 val perplexity 294.9540
-13200 train 5.783911 (lr=9.2144e-06) (hash(x)=154153766)
-13400 val loss 5.5625
-13400 val perplexity 260.4643
-13400 train 5.388410 (lr=8.2151e-06) (hash(x)=143223528)
-13300 val loss 5.6839
-13300 val perplexity 294.0870
-13300 train 5.475991 (lr=8.9865e-06) (hash(x)=162760888)
-13500 val loss 5.5673
-13500 val perplexity 261.7356
-13500 train 5.720043 (lr=8.0095e-06) (hash(x)=156416986)
-13400 val loss 5.6781
-13400 val perplexity 292.4062
-13400 train 5.499577 (lr=8.7628e-06) (hash(x)=143223528)
-13600 val loss 5.5630
-13600 val perplexity 260.6043
-13600 train 5.567687 (lr=7.8081e-06) (hash(x)=147521418)
-13500 val loss 5.6845
-13500 val perplexity 294.2740
-13500 train 5.829630 (lr=8.5435e-06) (hash(x)=156416986)
-13700 val loss 5.5587
-13700 val perplexity 259.4802
-13700 train 5.537975 (lr=7.6109e-06) (hash(x)=151127796)
-13600 val loss 5.6802
-13600 val perplexity 293.0212
-13600 train 5.681534 (lr=8.3286e-06) (hash(x)=147521418)
-13800 val loss 5.5565
-13800 val perplexity 258.9199
-13800 train 5.622880 (lr=7.4180e-06) (hash(x)=156863042)
-13700 val loss 5.6766
-13700 val perplexity 291.9636
-13700 train 5.645567 (lr=8.1183e-06) (hash(x)=151127796)
-13900 val loss 5.5517
-13900 val perplexity 257.6819
-13900 train 5.521236 (lr=7.2294e-06) (hash(x)=170986545)
-13800 val loss 5.6731
-13800 val perplexity 290.9450
-13800 train 5.732980 (lr=7.9125e-06) (hash(x)=156863042)
-14000 val loss 5.5492
-14000 val perplexity 257.0268
-14000 train 6.024418 (lr=7.0452e-06) (hash(x)=170744741)
-13900 val loss 5.6690
-13900 val perplexity 289.7548
-13900 train 5.673453 (lr=7.7114e-06) (hash(x)=170986545)
-14100 val loss 5.5483
-14100 val perplexity 256.8030
-14100 train 5.961041 (lr=6.8656e-06) (hash(x)=142328767)
-14000 val loss 5.6654
-14000 val perplexity 288.6993
-14000 train 6.136712 (lr=7.5149e-06) (hash(x)=170744741)
-14200 val loss 5.5467
-14200 val perplexity 256.3835
-14200 train 5.688292 (lr=6.6904e-06) (hash(x)=156141076)
-14100 val loss 5.6645
-14100 val perplexity 288.4526
-14100 train 6.060987 (lr=7.3233e-06) (hash(x)=142328767)
-14300 val loss 5.5492
-14300 val perplexity 257.0196
-14300 train 5.515243 (lr=6.5199e-06) (hash(x)=155447298)
-14200 val loss 5.6627
-14200 val perplexity 287.9176
-14200 train 5.813386 (lr=7.1365e-06) (hash(x)=156141076)
-14400 val loss 5.5428
-14400 val perplexity 255.3856
-14400 train 5.466012 (lr=6.3540e-06) (hash(x)=151991844)
-14300 val loss 5.6665
-14300 val perplexity 289.0223
-14300 train 5.643676 (lr=6.9546e-06) (hash(x)=155447298)
-14500 val loss 5.5391
-14500 val perplexity 254.4577
-14500 train 5.525082 (lr=6.1928e-06) (hash(x)=149773009)
-14400 val loss 5.6593
-14400 val perplexity 286.9611
-14400 train 5.591360 (lr=6.7776e-06) (hash(x)=151991844)
-14600 val loss 5.5368
-14600 val perplexity 253.8520
-14600 train 5.477211 (lr=6.0364e-06) (hash(x)=153836753)
-14500 val loss 5.6561
-14500 val perplexity 286.0430
-14500 train 5.631670 (lr=6.6057e-06) (hash(x)=149773009)
-14700 val loss 5.5346
-14700 val perplexity 253.2984
-14700 train 5.670206 (lr=5.8848e-06) (hash(x)=155466746)
-14600 val loss 5.6531
-14600 val perplexity 285.1770
-14600 train 5.608214 (lr=6.4389e-06) (hash(x)=153836753)
-14800 val loss 5.5342
-14800 val perplexity 253.2075
-14800 train 5.455852 (lr=5.7381e-06) (hash(x)=150110933)
-14700 val loss 5.6515
-14700 val perplexity 284.7285
-14700 train 5.781851 (lr=6.2771e-06) (hash(x)=155466746)
-14900 val loss 5.5353
-14900 val perplexity 253.4765
-14900 train 5.424331 (lr=5.5962e-06) (hash(x)=144558528)
-14800 val loss 5.6519
-14800 val perplexity 284.8269
-14800 train 5.579660 (lr=6.1206e-06) (hash(x)=150110933)
-15000 val loss 5.5324
-15000 val perplexity 252.7622
-15000 train 5.461942 (lr=5.4593e-06) (hash(x)=148315702)
-14900 val loss 5.6529
-14900 val perplexity 285.1064
-14900 train 5.540158 (lr=5.9693e-06) (hash(x)=144558528)
-15100 val loss 5.5290
-15100 val perplexity 251.8855
-15100 train 5.360836 (lr=5.3275e-06) (hash(x)=148452301)
-15000 val loss 5.6492
-15000 val perplexity 284.0755
-15000 train 5.591082 (lr=5.8233e-06) (hash(x)=148315702)
-15200 val loss 5.5284
-15200 val perplexity 251.7516
-15200 train 5.794895 (lr=5.2006e-06) (hash(x)=162181769)
-15100 val loss 5.6469
-15100 val perplexity 283.4043
-15100 train 5.490116 (lr=5.6826e-06) (hash(x)=148452301)
-15300 val loss 5.5254
-15300 val perplexity 250.9756
-15300 train 5.455128 (lr=5.0789e-06) (hash(x)=147256707)
-15200 val loss 5.6460
-15200 val perplexity 283.1549
-15200 train 5.907623 (lr=5.5473e-06) (hash(x)=162181769)
-15400 val loss 5.5246
-15400 val perplexity 250.7757
-15400 train 5.608875 (lr=4.9622e-06) (hash(x)=154947797)
-15300 val loss 5.6429
-15300 val perplexity 282.2780
-15300 train 5.572362 (lr=5.4175e-06) (hash(x)=147256707)
-15500 val loss 5.5235
-15500 val perplexity 250.5068
-15500 train 5.526404 (lr=4.8508e-06) (hash(x)=164200136)
-15400 val loss 5.6435
-15400 val perplexity 282.4595
-15400 train 5.747571 (lr=5.2931e-06) (hash(x)=154947797)
-15600 val loss 5.5229
-15600 val perplexity 250.3518
-15600 train 5.393709 (lr=4.7445e-06) (hash(x)=145086131)
-15500 val loss 5.6418
-15500 val perplexity 281.9762
-15500 train 5.676428 (lr=5.1742e-06) (hash(x)=164200136)
-15700 val loss 5.5191
-15700 val perplexity 249.4096
-15700 train 5.456976 (lr=4.6435e-06) (hash(x)=152070006)
-15600 val loss 5.6410
-15600 val perplexity 281.7361
-15600 train 5.505874 (lr=5.0608e-06) (hash(x)=145086131)
-15800 val loss 5.5173
-15800 val perplexity 248.9668
-15800 train 5.554524 (lr=4.5478e-06) (hash(x)=150969760)
-15700 val loss 5.6360
-15700 val perplexity 280.3287
-15700 train 5.572722 (lr=4.9531e-06) (hash(x)=152070006)
-15900 val loss 5.5166
-15900 val perplexity 248.7816
-15900 train 5.541933 (lr=4.4573e-06) (hash(x)=149495681)
-15800 val loss 5.6344
-15800 val perplexity 279.8850
-15800 train 5.664589 (lr=4.8509e-06) (hash(x)=150969760)
-16000 val loss 5.5158
-16000 val perplexity 248.5788
-16000 train 5.636119 (lr=4.3722e-06) (hash(x)=159383533)
-15900 val loss 5.6345
-15900 val perplexity 279.9112
-15900 train 5.652789 (lr=4.7545e-06) (hash(x)=149495681)
-16100 val loss 5.5158
-16100 val perplexity 248.5813
-16100 train 5.390795 (lr=4.2924e-06) (hash(x)=154365611)
-16000 val loss 5.6332
-16000 val perplexity 279.5606
-16000 train 5.748144 (lr=4.6637e-06) (hash(x)=159383533)
-16200 val loss 5.5139
-16200 val perplexity 248.1075
-16200 train 5.667452 (lr=4.2180e-06) (hash(x)=154284815)
-16100 val loss 5.6329
-16100 val perplexity 279.4662
-16100 train 5.498175 (lr=4.5786e-06) (hash(x)=154365611)
-16300 val loss 5.5127
-16300 val perplexity 247.8266
-16300 train 5.508811 (lr=4.1491e-06) (hash(x)=148416573)
-16200 val loss 5.6316
-16200 val perplexity 279.0958
-16200 train 5.793496 (lr=4.4993e-06) (hash(x)=154284815)
-16400 val loss 5.5107
-16400 val perplexity 247.3319
-16400 train 5.460778 (lr=4.0856e-06) (hash(x)=150447087)
-16300 val loss 5.6309
-16300 val perplexity 278.9182
-16300 train 5.620351 (lr=4.4257e-06) (hash(x)=148416573)
-16500 val loss 5.5105
-16500 val perplexity 247.2626
-16500 train 5.507597 (lr=4.0275e-06) (hash(x)=156112678)
-16400 val loss 5.6282
-16400 val perplexity 278.1597
-16400 train 5.578074 (lr=4.3579e-06) (hash(x)=150447087)
-16600 val loss 5.5083
-16600 val perplexity 246.7281
-16600 train 5.449524 (lr=3.9749e-06) (hash(x)=150449728)
-16500 val loss 5.6285
-16500 val perplexity 278.2431
-16500 train 5.620706 (lr=4.2960e-06) (hash(x)=156112678)
-16700 val loss 5.5066
-16700 val perplexity 246.3001
-16700 train 5.568088 (lr=3.9278e-06) (hash(x)=153905935)
-16600 val loss 5.6267
-16600 val perplexity 277.7373
-16600 train 5.564746 (lr=4.2399e-06) (hash(x)=150449728)
-16800 val loss 5.5031
-16800 val perplexity 245.4518
-16800 train 5.478307 (lr=3.8862e-06) (hash(x)=144086573)
-16700 val loss 5.6246
-16700 val perplexity 277.1668
-16700 train 5.682582 (lr=4.1896e-06) (hash(x)=153905935)
-16900 val loss 5.5039
-16900 val perplexity 245.6386
-16900 train 5.485327 (lr=3.8501e-06) (hash(x)=145863379)
-16800 val loss 5.6222
-16800 val perplexity 276.4983
-16800 train 5.592798 (lr=4.1452e-06) (hash(x)=144086573)
-17000 val loss 5.5038
-17000 val perplexity 245.6192
-17000 train 5.595400 (lr=3.8195e-06) (hash(x)=149054226)
-16900 val loss 5.6215
-16900 val perplexity 276.3130
-16900 train 5.600308 (lr=4.1067e-06) (hash(x)=145863379)
-17100 val loss 5.5031
-17100 val perplexity 245.4549
-17100 train 5.401844 (lr=3.7945e-06) (hash(x)=147109203)
-17000 val loss 5.6219
-17000 val perplexity 276.4100
-17000 train 5.686682 (lr=4.0741e-06) (hash(x)=149054226)
-17200 val loss 5.5015
-17200 val perplexity 245.0698
-17200 train 5.427037 (lr=3.7750e-06) (hash(x)=141837353)
-17100 val loss 5.6218
-17100 val perplexity 276.3966
-17100 train 5.527103 (lr=4.0475e-06) (hash(x)=147109203)
-17300 val loss 5.4995
-17300 val perplexity 244.5776
-17300 train 5.518600 (lr=3.7611e-06) (hash(x)=148542064)
-17200 val loss 5.6190
-17200 val perplexity 275.6230
-17200 train 5.551311 (lr=4.0267e-06) (hash(x)=141837353)
-17400 val loss 5.4973
-17400 val perplexity 244.0231
-17400 train 5.341631 (lr=3.7528e-06) (hash(x)=140939390)
-17300 val loss 5.6178
-17300 val perplexity 275.2757
-17300 train 5.632535 (lr=4.0119e-06) (hash(x)=148542064)
-17499 val loss 5.4967
-17499 val perplexity 243.8866
-17400 val loss 5.6163
-17400 val perplexity 274.8719
-17400 train 5.445556 (lr=4.0030e-06) (hash(x)=140939390)
-17499 val loss 5.6147
-17499 val perplexity 274.4340
+max_steps: 8750
+0 val loss 11.2669
+0 val perplexity 78187.7734
+2100 val loss 6.3234
+2100 val perplexity 557.4394
+2100 train 6.341493 (lr=3.6761e-05) (hash(x)=159721084)
+0 train 11.258430 (lr=1.0000e-07) (hash(x)=134751525)
+2200 val loss 6.2898
+2200 val perplexity 539.0543
+2200 train 6.273652 (lr=3.6358e-05) (hash(x)=144907001)
+100 val loss 10.0339
+100 val perplexity 22786.2734
+100 train 10.057615 (lr=1.0100e-05) (hash(x)=150358957)
+2300 val loss 6.2708
+2300 val perplexity 528.8981
+2300 train 6.209044 (lr=3.5935e-05) (hash(x)=154541837)
+200 val loss 9.4317
+200 val perplexity 12478.0625
+200 train 9.452061 (lr=2.0100e-05) (hash(x)=126319983)
+2400 val loss 6.2405
+2400 val perplexity 513.1238
+2400 train 6.335445 (lr=3.5491e-05) (hash(x)=161706790)
+300 val loss 8.2208
+300 val perplexity 3717.3279
+300 train 8.155610 (lr=3.0100e-05) (hash(x)=159305874)
+2500 val loss 6.2065
+2500 val perplexity 495.9534
+2500 train 6.129488 (lr=3.5027e-05) (hash(x)=143556162)
+400 val loss 7.6034
+400 val perplexity 2005.0027
+400 train 7.581938 (lr=4.0100e-05) (hash(x)=149855631)
+2600 val loss 6.1888
+2600 val perplexity 487.2436
+2600 train 6.040008 (lr=3.4545e-05) (hash(x)=142425078)
+500 val loss 7.4183
+500 val perplexity 1666.2664
+2700 val loss 6.1632
+2700 val perplexity 474.9361
+500 train 7.447924 (lr=5.0000e-05) (hash(x)=150706760)
+2700 train 6.187916 (lr=3.4044e-05) (hash(x)=163195606)
+600 val loss 7.2856
+600 val perplexity 1459.1943
+2800 val loss 6.1321
+2800 val perplexity 460.3968
+600 train 7.214978 (lr=4.9984e-05) (hash(x)=146858163)
+2800 train 6.040635 (lr=3.3526e-05) (hash(x)=154078337)
+700 val loss 7.1724
+700 val perplexity 1302.9692
+2900 val loss 6.1184
+2900 val perplexity 454.1573
+700 train 7.041670 (lr=4.9935e-05) (hash(x)=144262072)
+2900 train 5.980464 (lr=3.2992e-05) (hash(x)=145186687)
+800 val loss 7.0515
+800 val perplexity 1154.5875
+3000 val loss 6.0893
+3000 val perplexity 441.1168
+800 train 7.024981 (lr=4.9853e-05) (hash(x)=135443880)
+3000 train 6.053969 (lr=3.2441e-05) (hash(x)=149726716)
+900 val loss 6.9356
+900 val perplexity 1028.2451
+3100 val loss 6.0699
+3100 val perplexity 432.6577
+900 train 6.879653 (lr=4.9739e-05) (hash(x)=153147388)
+3100 train 5.996583 (lr=3.1875e-05) (hash(x)=151122509)
+1000 val loss 6.8611
+1000 val perplexity 954.4516
+3200 val loss 6.0574
+3200 val perplexity 427.2504
+1000 train 6.789762 (lr=4.9593e-05) (hash(x)=151448445)
+3200 train 6.024183 (lr=3.1295e-05) (hash(x)=152172187)
+1100 val loss 6.7471
+1100 val perplexity 851.6025
+1100 train 6.763339 (lr=4.9415e-05) (hash(x)=153161010)
+3300 val loss 6.0366
+3300 val perplexity 418.4794
+3300 train 5.972900 (lr=3.0702e-05) (hash(x)=150581974)
+1200 val loss 6.6606
+1200 val perplexity 781.0002
+1200 train 6.489265 (lr=4.9205e-05) (hash(x)=143227423)
+3400 val loss 6.0252
+3400 val perplexity 413.7378
+3400 train 5.951941 (lr=3.0095e-05) (hash(x)=160061666)
+1300 val loss 6.5786
+1300 val perplexity 719.5532
+1300 train 7.032114 (lr=4.8964e-05) (hash(x)=176373796)
+3500 val loss 6.0040
+3500 val perplexity 405.0391
+3500 train 5.953021 (lr=2.9477e-05) (hash(x)=150556913)
+1400 val loss 6.5175
+1400 val perplexity 676.9121
+1400 train 6.442500 (lr=4.8691e-05) (hash(x)=155989503)
+3600 val loss 5.9863
+3600 val perplexity 397.9272
+3600 train 5.898867 (lr=2.8849e-05) (hash(x)=148311961)
+1500 val loss 6.4455
+1500 val perplexity 629.8690
+1500 train 6.425520 (lr=4.8388e-05) (hash(x)=156507542)
+3700 val loss 5.9755
+3700 val perplexity 393.6454
+3700 train 5.869819 (lr=2.8210e-05) (hash(x)=150305284)
+1600 val loss 6.4007
+1600 val perplexity 602.2720
+1600 train 6.387810 (lr=4.8055e-05) (hash(x)=156078901)
+3800 val loss 5.9550
+3800 val perplexity 385.6592
+3800 train 5.966496 (lr=2.7562e-05) (hash(x)=148428531)
+3900 val loss 5.9399
+3900 val perplexity 379.8931
+1700 val loss 6.3578
+1700 val perplexity 576.9627
+3900 train 5.812899 (lr=2.6907e-05) (hash(x)=142448374)
+1700 train 6.265946 (lr=4.7691e-05) (hash(x)=156889457)
+4000 val loss 5.9339
+4000 val perplexity 377.6096
+1800 val loss 6.3287
+1800 val perplexity 560.4460
+4000 train 5.860096 (lr=2.6244e-05) (hash(x)=158002288)
+1800 train 6.571303 (lr=4.7299e-05) (hash(x)=162477906)
+4100 val loss 5.9118
+4100 val perplexity 369.3702
+1900 val loss 6.2829
+1900 val perplexity 535.3409
+4100 train 5.855021 (lr=2.5575e-05) (hash(x)=151388626)
+1900 train 6.331224 (lr=4.6878e-05) (hash(x)=158648033)
+2000 val loss 6.2529
+2000 val perplexity 519.5218
+4200 val loss 5.8998
+4200 val perplexity 364.9668
+2000 train 6.176389 (lr=4.6428e-05) (hash(x)=156732586)
+4200 train 5.820235 (lr=2.4900e-05) (hash(x)=145123294)
+2100 val loss 6.2218
+2100 val perplexity 503.5863
+4300 val loss 5.8852
+4300 val perplexity 359.6642
+2100 train 6.233637 (lr=4.5951e-05) (hash(x)=159721084)
+4300 train 5.760603 (lr=2.4222e-05) (hash(x)=145869150)
+2200 val loss 6.1882
+2200 val perplexity 486.9799
+4400 val loss 5.8783
+4400 val perplexity 357.2179
+2200 train 6.180151 (lr=4.5448e-05) (hash(x)=144907001)
+4400 train 5.910197 (lr=2.3540e-05) (hash(x)=146923320)
+2300 val loss 6.1754
+2300 val perplexity 480.7523
+2300 train 6.120876 (lr=4.4918e-05) (hash(x)=154541837)
+4500 val loss 5.8562
+4500 val perplexity 349.4057
+4500 train 5.689988 (lr=2.2856e-05) (hash(x)=146720819)
+2400 val loss 6.1511
+2400 val perplexity 469.2356
+2400 train 6.247717 (lr=4.4363e-05) (hash(x)=161706790)
+4600 val loss 5.8494
+4600 val perplexity 347.0210
+4600 train 5.598522 (lr=2.2171e-05) (hash(x)=140342909)
+2500 val loss 6.1159
+2500 val perplexity 452.9982
+2500 train 6.041896 (lr=4.3784e-05) (hash(x)=143556162)
+4700 val loss 5.8390
+4700 val perplexity 343.4363
+4700 train 5.973941 (lr=2.1486e-05) (hash(x)=163866463)
+2600 val loss 6.1075
+2600 val perplexity 449.2150
+2600 train 5.955640 (lr=4.3181e-05) (hash(x)=142425078)
+4800 val loss 5.8233
+4800 val perplexity 338.0736
+4800 train 5.517035 (lr=2.0801e-05) (hash(x)=133459145)
+2700 val loss 6.0728
+2700 val perplexity 433.9062
+2700 train 6.102422 (lr=4.2555e-05) (hash(x)=163195606)
+4900 val loss 5.8158
+4900 val perplexity 335.5680
+4900 train 5.618517 (lr=2.0118e-05) (hash(x)=143144356)
+2800 val loss 6.0459
+2800 val perplexity 422.3680
+2800 train 5.959752 (lr=4.1908e-05) (hash(x)=154078337)
+5000 val loss 5.8271
+5000 val perplexity 339.3646
+5000 train 5.733089 (lr=1.9438e-05) (hash(x)=132636494)
+2900 val loss 6.0318
+2900 val perplexity 416.4836
+2900 train 5.897262 (lr=4.1240e-05) (hash(x)=145186687)
+5100 val loss 5.7928
+5100 val perplexity 327.9449
+5100 train 5.715573 (lr=1.8762e-05) (hash(x)=157278728)
+3000 val loss 6.0084
+3000 val perplexity 406.8332
+3000 train 5.973816 (lr=4.0551e-05) (hash(x)=149726716)
+5200 val loss 5.7863
+5200 val perplexity 325.8200
+5200 train 5.641362 (lr=1.8090e-05) (hash(x)=148198434)
+3100 val loss 5.9891
+3100 val perplexity 399.0505
+3100 train 5.912086 (lr=3.9844e-05) (hash(x)=151122509)
+5300 val loss 5.7748
+5300 val perplexity 322.0783
+5300 train 5.702882 (lr=1.7424e-05) (hash(x)=138380906)
+3200 val loss 5.9763
+3200 val perplexity 393.9663
+3200 train 5.948274 (lr=3.9119e-05) (hash(x)=152172187)
+5400 val loss 5.7647
+5400 val perplexity 318.8321
+5400 train 5.852244 (lr=1.6765e-05) (hash(x)=163441464)
+3300 val loss 5.9555
+3300 val perplexity 385.8792
+3300 train 5.893881 (lr=3.8377e-05) (hash(x)=150581974)
+5500 val loss 5.7589
+5500 val perplexity 316.9842
+5500 train 5.707340 (lr=1.6113e-05) (hash(x)=154347714)
+3400 val loss 5.9440
+3400 val perplexity 381.4698
+3400 train 5.869471 (lr=3.7619e-05) (hash(x)=160061666)
+5600 val loss 5.7507
+5600 val perplexity 314.4107
+5600 train 5.793334 (lr=1.5469e-05) (hash(x)=148449981)
+3500 val loss 5.9191
+3500 val perplexity 372.0731
+3500 train 5.870534 (lr=3.6847e-05) (hash(x)=150556913)
+5700 val loss 5.7394
+5700 val perplexity 310.8636
+5700 train 5.631087 (lr=1.4836e-05) (hash(x)=139300274)
+3600 val loss 5.9062
+3600 val perplexity 367.3229
+3600 train 5.819812 (lr=3.6061e-05) (hash(x)=148311961)
+5800 val loss 5.7361
+5800 val perplexity 309.8442
+5800 train 5.725385 (lr=1.4212e-05) (hash(x)=154857144)
+3700 val loss 5.8894
+3700 val perplexity 361.1877
+3700 train 5.780119 (lr=3.5263e-05) (hash(x)=150305284)
+5900 val loss 5.7258
+5900 val perplexity 306.6911
+5900 train 5.790616 (lr=1.3600e-05) (hash(x)=151756013)
+3800 val loss 5.8679
+3800 val perplexity 353.4908
+3800 train 5.873760 (lr=3.4453e-05) (hash(x)=148428531)
+6000 val loss 5.7211
+6000 val perplexity 305.2342
+6000 train 5.626347 (lr=1.3000e-05) (hash(x)=145414657)
+3900 val loss 5.8545
+3900 val perplexity 348.8070
+3900 train 5.733078 (lr=3.3633e-05) (hash(x)=142448374)
+6100 val loss 5.7135
+6100 val perplexity 302.9179
+6100 train 5.538174 (lr=1.2413e-05) (hash(x)=139501217)
+4000 val loss 5.8454
+4000 val perplexity 345.6555
+4000 train 5.774595 (lr=3.2805e-05) (hash(x)=158002288)
+6200 val loss 5.7065
+6200 val perplexity 300.8038
+6200 train 5.745558 (lr=1.1840e-05) (hash(x)=155783358)
+4100 val loss 5.8263
+4100 val perplexity 339.1169
+4100 train 5.765561 (lr=3.1968e-05) (hash(x)=151388626)
+6300 val loss 5.6996
+6300 val perplexity 298.7556
+6300 train 5.757732 (lr=1.1282e-05) (hash(x)=157033091)
+4200 val loss 5.8122
+4200 val perplexity 334.3692
+4200 train 5.739853 (lr=3.1126e-05) (hash(x)=145123294)
+6400 val loss 5.6950
+6400 val perplexity 297.3656
+6400 train 5.500960 (lr=1.0739e-05) (hash(x)=144475330)
+4300 val loss 5.8004
+4300 val perplexity 330.4464
+4300 train 5.682630 (lr=3.0277e-05) (hash(x)=145869150)
+6500 val loss 5.6889
+6500 val perplexity 295.5813
+6500 train 5.733134 (lr=1.0213e-05) (hash(x)=124041822)
+4400 val loss 5.7926
+4400 val perplexity 327.8682
+4400 train 5.825141 (lr=2.9425e-05) (hash(x)=146923320)
+6600 val loss 5.6870
+6600 val perplexity 295.0057
+6600 train 5.619920 (lr=9.7032e-06) (hash(x)=142131981)
+4500 val loss 5.7743
+4500 val perplexity 321.9186
+4500 train 5.603869 (lr=2.8571e-05) (hash(x)=146720819)
+6700 val loss 5.6769
+6700 val perplexity 292.0528
+6700 train 5.776531 (lr=9.2116e-06) (hash(x)=154916248)
+4600 val loss 5.7642
+4600 val perplexity 318.6734
+4600 train 5.523923 (lr=2.7714e-05) (hash(x)=140342909)
+6800 val loss 5.6779
+6800 val perplexity 292.3389
+6800 train 5.674693 (lr=8.7387e-06) (hash(x)=147996387)
+4700 val loss 5.7529
+4700 val perplexity 315.0909
+4700 train 5.885762 (lr=2.6857e-05) (hash(x)=163866463)
+6900 val loss 5.6723
+6900 val perplexity 290.7080
+6900 train 5.701598 (lr=8.2849e-06) (hash(x)=153422428)
+4800 val loss 5.7385
+4800 val perplexity 310.5868
+4800 train 5.431343 (lr=2.6002e-05) (hash(x)=133459145)
+7000 val loss 5.6649
+7000 val perplexity 288.5706
+7000 train 5.685080 (lr=7.8510e-06) (hash(x)=174135078)
+4900 val loss 5.7298
+4900 val perplexity 307.9135
+4900 train 5.529056 (lr=2.5148e-05) (hash(x)=143144356)
+7100 val loss 5.6626
+7100 val perplexity 287.8837
+7100 train 5.738641 (lr=7.4377e-06) (hash(x)=160382475)
+5000 val loss 5.7392
+5000 val perplexity 310.8004
+5000 train 5.652639 (lr=2.4298e-05) (hash(x)=132636494)
+7200 val loss 5.6591
+7200 val perplexity 286.8994
+7200 train 5.749727 (lr=7.0455e-06) (hash(x)=155310085)
+5100 val loss 5.7103
+5100 val perplexity 301.9537
+5100 train 5.633508 (lr=2.3452e-05) (hash(x)=157278728)
+7300 val loss 5.6538
+7300 val perplexity 285.3680
+7300 train 5.788727 (lr=6.6749e-06) (hash(x)=159329031)
+5200 val loss 5.7006
+5200 val perplexity 299.0363
+5200 train 5.549832 (lr=2.2613e-05) (hash(x)=148198434)
+7400 val loss 5.6522
+7400 val perplexity 284.9121
+7400 train 5.595225 (lr=6.3266e-06) (hash(x)=151577269)
+5300 val loss 5.6867
+5300 val perplexity 294.9123
+5300 train 5.608719 (lr=2.1780e-05) (hash(x)=138380906)
+7500 val loss 5.6501
+7500 val perplexity 284.3204
+7500 train 5.701279 (lr=6.0010e-06) (hash(x)=160620287)
+5400 val loss 5.6770
+5400 val perplexity 292.0624
+5400 train 5.762171 (lr=2.0956e-05) (hash(x)=163441464)
+7600 val loss 5.6449
+7600 val perplexity 282.8365
+7600 train 5.535135 (lr=5.6986e-06) (hash(x)=143711597)
+5500 val loss 5.6715
+5500 val perplexity 290.4799
+5500 train 5.610294 (lr=2.0141e-05) (hash(x)=154347714)
+7700 val loss 5.6429
+7700 val perplexity 282.2746
+7700 train 5.639419 (lr=5.4198e-06) (hash(x)=149243577)
+5600 val loss 5.6623
+5600 val perplexity 287.8121
+5600 train 5.707816 (lr=1.9337e-05) (hash(x)=148449981)
+7800 val loss 5.6412
+7800 val perplexity 281.8125
+7800 train 5.623954 (lr=5.1650e-06) (hash(x)=147677247)
+5700 val loss 5.6526
+5700 val perplexity 285.0300
+5700 train 5.546151 (lr=1.8545e-05) (hash(x)=139300274)
+7900 val loss 5.6348
+7900 val perplexity 279.9960
+7900 train 5.586359 (lr=4.9347e-06) (hash(x)=145801236)
+5800 val loss 5.6498
+5800 val perplexity 284.2278
+5800 train 5.649865 (lr=1.7765e-05) (hash(x)=154857144)
+8000 val loss 5.6343
+8000 val perplexity 279.8730
+8000 train 5.752988 (lr=4.7291e-06) (hash(x)=158007655)
+5900 val loss 5.6391
+5900 val perplexity 281.2168
+5900 train 5.697678 (lr=1.7000e-05) (hash(x)=151756013)
+8100 val loss 5.6332
+8100 val perplexity 279.5642
+8100 train 5.698683 (lr=4.5486e-06) (hash(x)=158030048)
+6000 val loss 5.6328
+6000 val perplexity 279.4322
+6000 train 5.540736 (lr=1.6250e-05) (hash(x)=145414657)
+8200 val loss 5.6291
+8200 val perplexity 278.3981
+8200 train 5.477635 (lr=4.3933e-06) (hash(x)=150674725)
+6100 val loss 5.6272
+6100 val perplexity 277.8833
+6100 train 5.443038 (lr=1.5516e-05) (hash(x)=139501217)
+8300 val loss 5.6283
+8300 val perplexity 278.1931
+8300 train 5.535476 (lr=4.2636e-06) (hash(x)=150805711)
+6200 val loss 5.6168
+6200 val perplexity 274.9977
+6200 train 5.654116 (lr=1.4800e-05) (hash(x)=155783358)
+8400 val loss 5.6236
+8400 val perplexity 276.8825
+8400 train 5.500546 (lr=4.1596e-06) (hash(x)=138396210)
+6300 val loss 5.6119
+6300 val perplexity 273.6768
+6300 train 5.672994 (lr=1.4102e-05) (hash(x)=157033091)
+8500 val loss 5.6230
+8500 val perplexity 276.7253
+8500 train 5.496522 (lr=4.0815e-06) (hash(x)=156172740)
+6400 val loss 5.6083
+6400 val perplexity 272.6679
+6400 train 5.410939 (lr=1.3424e-05) (hash(x)=144475330)
+8600 val loss 5.6206
+8600 val perplexity 276.0491
+8600 train 5.663106 (lr=4.0294e-06) (hash(x)=154911352)
+6500 val loss 5.6007
+6500 val perplexity 270.6130
+6500 train 5.650802 (lr=1.2766e-05) (hash(x)=124041822)
+8700 val loss 5.6183
+8700 val perplexity 275.4319
+8700 train 5.675045 (lr=4.0033e-06) (hash(x)=153446789)
+8749 val loss 5.6168
+8749 val perplexity 275.0146
+6600 val loss 5.5983
+6600 val perplexity 269.9799
+6600 train 5.537288 (lr=1.2129e-05) (hash(x)=142131981)

attention_kindselective_n_heads4_seed1345/model_08749.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:31cd955fc1721477be23f672ac87f87e451ef7a89edd44713354f527a6a1c59a
+size 92843394

attention_kindselective_n_heads4_seed1345/optimizer_08749.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:362d9c846eda58dcea8e110286fcf8e8e3c0a5e25baf54e68a42fb838089a968
+size 179406214