Upload folder using huggingface_hub

attention_kindselective_n_heads2_seed1341/args.json

@@ -1 +1 @@
-{"hellaswag": true, "attention_kind": "selective", "log_dir": "wider_is_better_4/attention_kindselective_n_heads2_seed1341", "resume_checkpoint": null, "resume_optimizer": false, "add_a_head": false, "add_head_to_start": true, "new_head_init": "normal", "n_heads": 2, "protect_bos_token": true, "prevent_from_masking_myself": true, "max_steps": 10000, "warmup_steps": 200, "group": "wider_is_better_4", "use_wandb": true, "kill_self_after_run": false, "random_seed": 1341, "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": 0.00015, "decay_lr": true, "readout_zero_init": false, "query_zero_init": false, "l1_loss": false, "debugpy": false, "key": "1.5e-4_30720_2_1341", "n_embd": 128}
+{"hellaswag": true, "attention_kind": "selective", "log_dir": "wider_is_better_4/attention_kindselective_n_heads2_seed1341", "resume_checkpoint": null, "resume_optimizer": false, "add_a_head": false, "add_head_to_start": true, "new_head_init": "normal", "n_heads": 2, "protect_bos_token": true, "prevent_from_masking_myself": true, "max_steps": 10000, "warmup_steps": 200, "group": "wider_is_better_4", "use_wandb": true, "kill_self_after_run": false, "random_seed": 1341, "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": 5e-05, "decay_lr": true, "readout_zero_init": false, "query_zero_init": false, "l1_loss": false, "debugpy": false, "key": "0.5e-4_30720_2_1341", "n_embd": 128}

attention_kindselective_n_heads2_seed1341/log2.txt

@@ -1,303 +1,303 @@
 max_steps: 10000
 0 val loss 11.7722
 0 val perplexity 129595.2031
-0 train 11.774795 (lr=7.5000e-07) (hash(x)=145079536)
-100 val loss 9.3037
-100 val perplexity 10979.0723
-100 train 9.244246 (lr=7.5750e-05) (hash(x)=154745873)
-200 val loss 7.7460
-200 val perplexity 2312.2744
-200 train 7.946997 (lr=1.5000e-04) (hash(x)=155800595)
-300 val loss 7.5960
-300 val perplexity 1990.2469
-300 train 7.554796 (lr=1.4997e-04) (hash(x)=148595389)
-400 val loss 7.5968
-400 val perplexity 1991.7423
-400 train 7.505088 (lr=1.4986e-04) (hash(x)=145606733)
-500 val loss 7.5520
-500 val perplexity 1904.6262
-500 train 7.350472 (lr=1.4969e-04) (hash(x)=138221231)
-600 val loss 7.5281
-600 val perplexity 1859.5831
-600 train 7.521947 (lr=1.4945e-04) (hash(x)=150367139)
-700 val loss 7.5113
-700 val perplexity 1828.5997
-700 train 7.653578 (lr=1.4913e-04) (hash(x)=155579314)
-800 val loss 7.5076
-800 val perplexity 1821.8347
-800 train 7.586863 (lr=1.4876e-04) (hash(x)=155747374)
-900 val loss 7.4474
-900 val perplexity 1715.4008
-900 train 7.648261 (lr=1.4831e-04) (hash(x)=159334575)
-1000 val loss 7.4414
-1000 val perplexity 1705.1790
-1000 train 7.342261 (lr=1.4779e-04) (hash(x)=140604760)
-1100 val loss 7.3735
-1100 val perplexity 1593.1841
-1100 train 7.223920 (lr=1.4721e-04) (hash(x)=146208052)
-1200 val loss 7.3531
-1200 val perplexity 1560.9684
-1200 train 7.281145 (lr=1.4656e-04) (hash(x)=148404734)
-1300 val loss 7.2945
-1300 val perplexity 1472.1868
-1300 train 7.225808 (lr=1.4585e-04) (hash(x)=155681970)
-1400 val loss 7.2784
-1400 val perplexity 1448.7200
-1400 train 7.236976 (lr=1.4507e-04) (hash(x)=148115934)
-1500 val loss 7.2136
-1500 val perplexity 1357.7581
-1500 train 7.340710 (lr=1.4422e-04) (hash(x)=157074034)
-1600 val loss 7.1754
-1600 val perplexity 1306.8774
-1600 train 6.969009 (lr=1.4332e-04) (hash(x)=137464699)
-1700 val loss 7.1214
-1700 val perplexity 1238.1949
-1700 train 7.288368 (lr=1.4235e-04) (hash(x)=166955614)
-1800 val loss 7.0839
-1800 val perplexity 1192.5714
-1800 train 6.921106 (lr=1.4131e-04) (hash(x)=143886042)
-1900 val loss 7.0449
-1900 val perplexity 1146.9973
-1900 train 7.622898 (lr=1.4022e-04) (hash(x)=193610391)
-2000 val loss 6.9766
-2000 val perplexity 1071.3042
-2000 train 7.168832 (lr=1.3907e-04) (hash(x)=163799796)
-2100 val loss 6.9072
-2100 val perplexity 999.4552
-2100 train 6.886005 (lr=1.3786e-04) (hash(x)=154107339)
-2200 val loss 6.8528
-2200 val perplexity 946.5522
-2200 train 6.839452 (lr=1.3660e-04) (hash(x)=144592844)
-2300 val loss 6.8093
-2300 val perplexity 906.2578
-2300 train 6.876617 (lr=1.3527e-04) (hash(x)=154535861)
-2400 val loss 6.7855
-2400 val perplexity 884.9128
-2400 train 7.465554 (lr=1.3390e-04) (hash(x)=204706354)
-2500 val loss 6.7782
-2500 val perplexity 878.4916
-2500 train 6.612565 (lr=1.3247e-04) (hash(x)=143783202)
-2600 val loss 6.7483
-2600 val perplexity 852.6260
-2600 train 6.657750 (lr=1.3099e-04) (hash(x)=150862210)
-2700 val loss 6.7156
-2700 val perplexity 825.1681
-2700 train 6.771245 (lr=1.2946e-04) (hash(x)=150753426)
-2800 val loss 6.6992
-2800 val perplexity 811.7590
-2800 train 6.700248 (lr=1.2788e-04) (hash(x)=147766811)
-2900 val loss 6.6682
-2900 val perplexity 786.9391
-2900 train 6.603729 (lr=1.2626e-04) (hash(x)=151177814)
-3000 val loss 6.6481
-3000 val perplexity 771.3284
-3000 train 6.530660 (lr=1.2459e-04) (hash(x)=135925327)
-3100 val loss 6.6459
-3100 val perplexity 769.5980
-3100 train 6.306259 (lr=1.2287e-04) (hash(x)=124842353)
-3200 val loss 6.6309
-3200 val perplexity 758.1389
-3200 train 6.726900 (lr=1.2112e-04) (hash(x)=160440642)
-3300 val loss 6.6013
-3300 val perplexity 736.0597
-3300 train 6.514308 (lr=1.1932e-04) (hash(x)=143017131)
-3400 val loss 6.5759
-3400 val perplexity 717.5570
-3400 train 6.911897 (lr=1.1749e-04) (hash(x)=148143474)
-3500 val loss 6.5747
-3500 val perplexity 716.7311
-3500 train 6.394977 (lr=1.1562e-04) (hash(x)=148732639)
-3600 val loss 6.5536
-3600 val perplexity 701.7585
-3600 train 6.547552 (lr=1.1372e-04) (hash(x)=146333803)
-3700 val loss 6.5360
-3700 val perplexity 689.5156
-3700 train 6.359855 (lr=1.1179e-04) (hash(x)=143611445)
-3800 val loss 6.5232
-3800 val perplexity 680.7549
-3800 train 6.323216 (lr=1.0982e-04) (hash(x)=141272115)
-3900 val loss 6.5171
-3900 val perplexity 676.5939
-3900 train 6.454644 (lr=1.0783e-04) (hash(x)=149161848)
-4000 val loss 6.4992
-4000 val perplexity 664.5805
-4000 train 6.553896 (lr=1.0581e-04) (hash(x)=152036955)
-4100 val loss 6.4889
-4100 val perplexity 657.7863
-4100 train 6.545446 (lr=1.0377e-04) (hash(x)=153260968)
-4200 val loss 6.4873
-4200 val perplexity 656.7577
-4200 train 6.374079 (lr=1.0171e-04) (hash(x)=147629478)
-4300 val loss 6.4743
-4300 val perplexity 648.2596
-4300 train 6.486774 (lr=9.9622e-05) (hash(x)=161701080)
-4400 val loss 6.4651
-4400 val perplexity 642.3095
-4400 train 6.521642 (lr=9.7520e-05) (hash(x)=155290524)
-4500 val loss 6.4584
-4500 val perplexity 638.0176
-4500 train 6.259203 (lr=9.5403e-05) (hash(x)=144726593)
-4600 val loss 6.4602
-4600 val perplexity 639.1897
-4600 train 6.407053 (lr=9.3273e-05) (hash(x)=149517202)
-4700 val loss 6.4416
-4700 val perplexity 627.4361
-4700 train 6.434956 (lr=9.1132e-05) (hash(x)=139403008)
-4800 val loss 6.4160
-4800 val perplexity 611.5692
-4800 train 6.517553 (lr=8.8982e-05) (hash(x)=153387238)
-4900 val loss 6.4117
-4900 val perplexity 608.9391
-4900 train 6.209703 (lr=8.6825e-05) (hash(x)=139881569)
-5000 val loss 6.3954
-5000 val perplexity 599.0758
-5000 train 6.397029 (lr=8.4663e-05) (hash(x)=152602580)
-5100 val loss 6.4324
-5100 val perplexity 621.6495
-5100 train 6.492046 (lr=8.2500e-05) (hash(x)=150598460)
-5200 val loss 6.4197
-5200 val perplexity 613.7912
-5200 train 6.316842 (lr=8.0337e-05) (hash(x)=147660496)
-5300 val loss 6.3985
-5300 val perplexity 600.9280
-5300 train 6.150064 (lr=7.8175e-05) (hash(x)=147693848)
-5400 val loss 6.4180
-5400 val perplexity 612.7855
-5400 train 6.179458 (lr=7.6018e-05) (hash(x)=136048145)
-5500 val loss 6.3870
-5500 val perplexity 594.0621
-5500 train 6.369116 (lr=7.3868e-05) (hash(x)=140313123)
-5600 val loss 6.3757
-5600 val perplexity 587.4236
-5600 train 6.401426 (lr=7.1727e-05) (hash(x)=156997993)
-5700 val loss 6.3713
-5700 val perplexity 584.8242
-5700 train 6.384679 (lr=6.9597e-05) (hash(x)=143513880)
-5800 val loss 6.3648
-5800 val perplexity 580.9998
-5800 train 6.479723 (lr=6.7480e-05) (hash(x)=153736628)
-5900 val loss 6.3639
-5900 val perplexity 580.4908
-5900 train 6.517274 (lr=6.5378e-05) (hash(x)=166385417)
-6000 val loss 6.3563
-6000 val perplexity 576.1237
-6000 train 6.592653 (lr=6.3294e-05) (hash(x)=153216517)
-6100 val loss 6.3545
-6100 val perplexity 575.0741
-6100 train 6.244929 (lr=6.1230e-05) (hash(x)=145387545)
-6200 val loss 6.3445
-6200 val perplexity 569.3646
-6200 train 6.456944 (lr=5.9188e-05) (hash(x)=156306460)
-6300 val loss 6.3371
-6300 val perplexity 565.1736
-6300 train 6.238484 (lr=5.7169e-05) (hash(x)=148610447)
-6400 val loss 6.3339
-6400 val perplexity 563.3510
-6400 train 6.519849 (lr=5.5177e-05) (hash(x)=158343535)
-6500 val loss 6.3335
-6500 val perplexity 563.1171
-6500 train 6.496376 (lr=5.3213e-05) (hash(x)=157742557)
-6600 val loss 6.3244
-6600 val perplexity 558.0327
-6600 train 6.256868 (lr=5.1279e-05) (hash(x)=150457308)
-6700 val loss 6.3217
-6700 val perplexity 556.5438
-6700 train 6.202847 (lr=4.9377e-05) (hash(x)=142426205)
-6800 val loss 6.3153
-6800 val perplexity 552.9587
-6800 train 6.127337 (lr=4.7509e-05) (hash(x)=151522525)
-6900 val loss 6.3105
-6900 val perplexity 550.3406
-6900 train 6.256704 (lr=4.5676e-05) (hash(x)=152029630)
-7000 val loss 6.2970
-7000 val perplexity 542.9250
-7000 train 6.171764 (lr=4.3882e-05) (hash(x)=142978954)
-7100 val loss 6.2922
-7100 val perplexity 540.3380
-7100 train 6.182000 (lr=4.2128e-05) (hash(x)=148064359)
-7200 val loss 6.2844
-7200 val perplexity 536.1356
-7200 train 6.192896 (lr=4.0414e-05) (hash(x)=155771845)
-7300 val loss 6.2826
-7300 val perplexity 535.1908
-7300 train 6.097363 (lr=3.8745e-05) (hash(x)=141639844)
-7400 val loss 6.2840
-7400 val perplexity 535.9132
-7400 train 6.340342 (lr=3.7120e-05) (hash(x)=153466021)
-7500 val loss 6.2818
-7500 val perplexity 534.7566
-7500 train 6.249422 (lr=3.5541e-05) (hash(x)=148843587)
-7600 val loss 6.2678
-7600 val perplexity 527.3170
-7600 train 6.355673 (lr=3.4011e-05) (hash(x)=152053933)
-7700 val loss 6.2628
-7700 val perplexity 524.6606
-7700 train 6.090435 (lr=3.2531e-05) (hash(x)=141564003)
-7800 val loss 6.2601
-7800 val perplexity 523.2859
-7800 train 6.137829 (lr=3.1102e-05) (hash(x)=148617843)
-7900 val loss 6.2531
-7900 val perplexity 519.6293
-7900 train 6.211982 (lr=2.9726e-05) (hash(x)=146417668)
-8000 val loss 6.2608
-8000 val perplexity 523.6473
-8000 train 6.075680 (lr=2.8405e-05) (hash(x)=150165488)
-8100 val loss 6.2577
-8100 val perplexity 522.0248
-8100 train 6.238003 (lr=2.7138e-05) (hash(x)=144818855)
-8200 val loss 6.2450
-8200 val perplexity 515.4420
-8200 train 6.336446 (lr=2.5929e-05) (hash(x)=143074954)
-8300 val loss 6.2406
-8300 val perplexity 513.1754
-8300 train 6.135967 (lr=2.4778e-05) (hash(x)=145172494)
-8400 val loss 6.2397
-8400 val perplexity 512.6929
-8400 train 6.257989 (lr=2.3686e-05) (hash(x)=152284393)
-8500 val loss 6.2402
-8500 val perplexity 512.9636
-8500 train 6.090525 (lr=2.2655e-05) (hash(x)=147044705)
-8600 val loss 6.2379
-8600 val perplexity 511.7964
-8600 train 6.147628 (lr=2.1685e-05) (hash(x)=147762608)
-8700 val loss 6.2356
-8700 val perplexity 510.5859
-8700 train 6.242692 (lr=2.0777e-05) (hash(x)=141994567)
-8800 val loss 6.2281
-8800 val perplexity 506.7967
-8800 train 6.295036 (lr=1.9933e-05) (hash(x)=152552896)
-8900 val loss 6.2227
-8900 val perplexity 504.0461
-8900 train 6.078781 (lr=1.9153e-05) (hash(x)=139714454)
-9000 val loss 6.2215
-9000 val perplexity 503.4492
-9000 train 6.259016 (lr=1.8439e-05) (hash(x)=152274419)
-9100 val loss 6.2211
-9100 val perplexity 503.2584
-9100 train 6.327313 (lr=1.7790e-05) (hash(x)=160536885)
-9200 val loss 6.2220
-9200 val perplexity 503.6907
-9200 train 6.066282 (lr=1.7208e-05) (hash(x)=138386899)
-9300 val loss 6.2184
-9300 val perplexity 501.9156
-9300 train 6.252558 (lr=1.6692e-05) (hash(x)=149957457)
-9400 val loss 6.2129
-9400 val perplexity 499.1484
-9400 train 6.262452 (lr=1.6245e-05) (hash(x)=161671548)
-9500 val loss 6.2120
-9500 val perplexity 498.6978
-9500 train 6.317385 (lr=1.5865e-05) (hash(x)=171857400)
-9600 val loss 6.2096
-9600 val perplexity 497.4986
-9600 train 5.995666 (lr=1.5554e-05) (hash(x)=135378471)
-9700 val loss 6.2061
-9700 val perplexity 495.7690
-9700 train 6.124307 (lr=1.5312e-05) (hash(x)=149514591)
-9800 val loss 6.2041
-9800 val perplexity 494.7568
-9800 train 6.190481 (lr=1.5139e-05) (hash(x)=152679072)
-9900 val loss 6.2029
-9900 val perplexity 494.1768
-9900 train 6.138339 (lr=1.5035e-05) (hash(x)=150415193)
-9999 val loss 6.2036
-9999 val perplexity 494.5280
+0 train 11.774795 (lr=2.5000e-07) (hash(x)=145079536)
+100 val loss 10.1967
+100 val perplexity 26815.2441
+100 train 10.143024 (lr=2.5250e-05) (hash(x)=154745873)
+200 val loss 8.4253
+200 val perplexity 4561.1460
+200 train 8.489046 (lr=5.0000e-05) (hash(x)=155800595)
+300 val loss 7.7144
+300 val perplexity 2240.3523
+300 train 7.665938 (lr=4.9988e-05) (hash(x)=148595389)
+400 val loss 7.5699
+400 val perplexity 1938.9187
+400 train 7.476192 (lr=4.9954e-05) (hash(x)=145606733)
+500 val loss 7.5245
+500 val perplexity 1852.8209
+500 train 7.344064 (lr=4.9896e-05) (hash(x)=138221231)
+600 val loss 7.4760
+600 val perplexity 1765.1185
+600 train 7.476169 (lr=4.9815e-05) (hash(x)=150367139)
+700 val loss 7.4224
+700 val perplexity 1672.9674
+700 train 7.559739 (lr=4.9712e-05) (hash(x)=155579314)
+800 val loss 7.3439
+800 val perplexity 1546.7035
+800 train 7.421778 (lr=4.9585e-05) (hash(x)=155747374)
+900 val loss 7.2824
+900 val perplexity 1454.4935
+900 train 7.492918 (lr=4.9436e-05) (hash(x)=159334575)
+1000 val loss 7.2183
+1000 val perplexity 1364.1555
+1000 train 7.090861 (lr=4.9264e-05) (hash(x)=140604760)
+1100 val loss 7.1651
+1100 val perplexity 1293.5281
+1100 train 6.987471 (lr=4.9070e-05) (hash(x)=146208052)
+1200 val loss 7.0891
+1200 val perplexity 1198.8540
+1200 train 6.984112 (lr=4.8854e-05) (hash(x)=148404734)
+1300 val loss 7.0370
+1300 val perplexity 1137.9492
+1300 train 6.965640 (lr=4.8616e-05) (hash(x)=155681970)
+1400 val loss 6.9815
+1400 val perplexity 1076.5845
+1400 train 6.933797 (lr=4.8356e-05) (hash(x)=148115934)
+1500 val loss 6.9014
+1500 val perplexity 993.6962
+1500 train 7.060465 (lr=4.8074e-05) (hash(x)=157074034)
+1600 val loss 6.8488
+1600 val perplexity 942.7806
+1600 train 6.613491 (lr=4.7772e-05) (hash(x)=137464699)
+1700 val loss 6.7931
+1700 val perplexity 891.7070
+1700 train 6.929006 (lr=4.7448e-05) (hash(x)=166955614)
+1800 val loss 6.7475
+1800 val perplexity 851.9409
+1800 train 6.565822 (lr=4.7105e-05) (hash(x)=143886042)
+1900 val loss 6.7145
+1900 val perplexity 824.3010
+1900 train 7.249403 (lr=4.6741e-05) (hash(x)=193610391)
+2000 val loss 6.6721
+2000 val perplexity 790.0560
+2000 train 6.862562 (lr=4.6357e-05) (hash(x)=163799796)
+2100 val loss 6.6205
+2100 val perplexity 750.3120
+2100 train 6.607678 (lr=4.5954e-05) (hash(x)=154107339)
+2200 val loss 6.5971
+2200 val perplexity 732.9810
+2200 train 6.600620 (lr=4.5532e-05) (hash(x)=144592844)
+2300 val loss 6.5494
+2300 val perplexity 698.8473
+2300 train 6.613638 (lr=4.5091e-05) (hash(x)=154535861)
+2400 val loss 6.5211
+2400 val perplexity 679.3499
+2400 train 7.256778 (lr=4.4633e-05) (hash(x)=204706354)
+2500 val loss 6.5063
+2500 val perplexity 669.3500
+2500 train 6.299736 (lr=4.4156e-05) (hash(x)=143783202)
+2600 val loss 6.4746
+2600 val perplexity 648.4813
+2600 train 6.395596 (lr=4.3663e-05) (hash(x)=150862210)
+2700 val loss 6.4500
+2700 val perplexity 632.7221
+2700 train 6.488371 (lr=4.3153e-05) (hash(x)=150753426)
+2800 val loss 6.4271
+2800 val perplexity 618.3502
+2800 train 6.448690 (lr=4.2627e-05) (hash(x)=147766811)
+2900 val loss 6.4014
+2900 val perplexity 602.6843
+2900 train 6.334109 (lr=4.2085e-05) (hash(x)=151177814)
+3000 val loss 6.3806
+3000 val perplexity 590.3071
+3000 train 6.256360 (lr=4.1529e-05) (hash(x)=135925327)
+3100 val loss 6.3712
+3100 val perplexity 584.7452
+3100 train 6.015045 (lr=4.0957e-05) (hash(x)=124842353)
+3200 val loss 6.3610
+3200 val perplexity 578.8417
+3200 train 6.456865 (lr=4.0373e-05) (hash(x)=160440642)
+3300 val loss 6.3370
+3300 val perplexity 565.1111
+3300 train 6.252707 (lr=3.9775e-05) (hash(x)=143017131)
+3400 val loss 6.3147
+3400 val perplexity 552.6184
+3400 train 6.692260 (lr=3.9164e-05) (hash(x)=148143474)
+3500 val loss 6.3072
+3500 val perplexity 548.5033
+3500 train 6.128360 (lr=3.8541e-05) (hash(x)=148732639)
+3600 val loss 6.2925
+3600 val perplexity 540.4887
+3600 train 6.286038 (lr=3.7907e-05) (hash(x)=146333803)
+3700 val loss 6.2816
+3700 val perplexity 534.6521
+3700 train 6.090304 (lr=3.7262e-05) (hash(x)=143611445)
+3800 val loss 6.2777
+3800 val perplexity 532.5809
+3800 train 6.061575 (lr=3.6608e-05) (hash(x)=141272115)
+3900 val loss 6.2569
+3900 val perplexity 521.6227
+3900 train 6.195599 (lr=3.5944e-05) (hash(x)=149161848)
+4000 val loss 6.2324
+4000 val perplexity 508.9707
+4000 train 6.309767 (lr=3.5271e-05) (hash(x)=152036955)
+4100 val loss 6.2213
+4100 val perplexity 503.3548
+4100 train 6.268515 (lr=3.4590e-05) (hash(x)=153260968)
+4200 val loss 6.2097
+4200 val perplexity 497.5497
+4200 train 6.101007 (lr=3.3902e-05) (hash(x)=147629478)
+4300 val loss 6.1937
+4300 val perplexity 489.6672
+4300 train 6.167351 (lr=3.3207e-05) (hash(x)=161701080)
+4400 val loss 6.1911
+4400 val perplexity 488.3745
+4400 train 6.231541 (lr=3.2507e-05) (hash(x)=155290524)
+4500 val loss 6.1868
+4500 val perplexity 486.3091
+4500 train 5.971059 (lr=3.1801e-05) (hash(x)=144726593)
+4600 val loss 6.1760
+4600 val perplexity 481.0795
+4600 train 6.129531 (lr=3.1091e-05) (hash(x)=149517202)
+4700 val loss 6.1587
+4700 val perplexity 472.8268
+4700 train 6.162160 (lr=3.0377e-05) (hash(x)=139403008)
+4800 val loss 6.1452
+4800 val perplexity 466.4960
+4800 train 6.247012 (lr=2.9661e-05) (hash(x)=153387238)
+4900 val loss 6.1427
+4900 val perplexity 465.2972
+4900 train 5.897235 (lr=2.8942e-05) (hash(x)=139881569)
+5000 val loss 6.1239
+5000 val perplexity 456.6504
+5000 train 6.141126 (lr=2.8221e-05) (hash(x)=152602580)
+5100 val loss 6.1204
+5100 val perplexity 455.0395
+5100 train 6.160268 (lr=2.7500e-05) (hash(x)=150598460)
+5200 val loss 6.1112
+5200 val perplexity 450.8787
+5200 train 5.984895 (lr=2.6779e-05) (hash(x)=147660496)
+5300 val loss 6.1046
+5300 val perplexity 447.9008
+5300 train 5.845158 (lr=2.6058e-05) (hash(x)=147693848)
+5400 val loss 6.1049
+5400 val perplexity 448.0625
+5400 train 5.862473 (lr=2.5339e-05) (hash(x)=136048145)
+5500 val loss 6.0878
+5500 val perplexity 440.4329
+5500 train 6.121768 (lr=2.4623e-05) (hash(x)=140313123)
+5600 val loss 6.0770
+5600 val perplexity 435.7415
+5600 train 6.093797 (lr=2.3909e-05) (hash(x)=156997993)
+5700 val loss 6.0683
+5700 val perplexity 431.9543
+5700 train 6.087464 (lr=2.3199e-05) (hash(x)=143513880)
+5800 val loss 6.0614
+5800 val perplexity 428.9725
+5800 train 6.172104 (lr=2.2493e-05) (hash(x)=153736628)
+5900 val loss 6.0547
+5900 val perplexity 426.1208
+5900 train 6.190841 (lr=2.1793e-05) (hash(x)=166385417)
+6000 val loss 6.0458
+6000 val perplexity 422.3309
+6000 train 6.277092 (lr=2.1098e-05) (hash(x)=153216517)
+6100 val loss 6.0446
+6100 val perplexity 421.8097
+6100 train 5.932088 (lr=2.0410e-05) (hash(x)=145387545)
+6200 val loss 6.0460
+6200 val perplexity 422.4171
+6200 train 6.175024 (lr=1.9729e-05) (hash(x)=156306460)
+6300 val loss 6.0292
+6300 val perplexity 415.3701
+6300 train 5.960823 (lr=1.9056e-05) (hash(x)=148610447)
+6400 val loss 6.0213
+6400 val perplexity 412.0996
+6400 train 6.243478 (lr=1.8392e-05) (hash(x)=158343535)
+6500 val loss 6.0134
+6500 val perplexity 408.8782
+6500 train 6.185011 (lr=1.7738e-05) (hash(x)=157742557)
+6600 val loss 6.0124
+6600 val perplexity 408.4781
+6600 train 5.903574 (lr=1.7093e-05) (hash(x)=150457308)
+6700 val loss 6.0066
+6700 val perplexity 406.1186
+6700 train 5.866361 (lr=1.6459e-05) (hash(x)=142426205)
+6800 val loss 6.0069
+6800 val perplexity 406.2187
+6800 train 5.806967 (lr=1.5836e-05) (hash(x)=151522525)
+6900 val loss 5.9981
+6900 val perplexity 402.6668
+6900 train 5.947786 (lr=1.5225e-05) (hash(x)=152029630)
+7000 val loss 5.9889
+7000 val perplexity 398.9860
+7000 train 5.835184 (lr=1.4627e-05) (hash(x)=142978954)
+7100 val loss 5.9839
+7100 val perplexity 396.9924
+7100 train 5.866302 (lr=1.4043e-05) (hash(x)=148064359)
+7200 val loss 5.9817
+7200 val perplexity 396.0956
+7200 train 5.878535 (lr=1.3471e-05) (hash(x)=155771845)
+7300 val loss 5.9780
+7300 val perplexity 394.6485
+7300 train 5.788899 (lr=1.2915e-05) (hash(x)=141639844)
+7400 val loss 5.9782
+7400 val perplexity 394.7098
+7400 train 6.030018 (lr=1.2373e-05) (hash(x)=153466021)
+7500 val loss 5.9703
+7500 val perplexity 391.6344
+7500 train 5.951066 (lr=1.1847e-05) (hash(x)=148843587)
+7600 val loss 5.9636
+7600 val perplexity 388.9968
+7600 train 6.093093 (lr=1.1337e-05) (hash(x)=152053933)
+7700 val loss 5.9594
+7700 val perplexity 387.3708
+7700 train 5.819657 (lr=1.0844e-05) (hash(x)=141564003)
+7800 val loss 5.9565
+7800 val perplexity 386.2691
+7800 train 5.836637 (lr=1.0367e-05) (hash(x)=148617843)
+7900 val loss 5.9547
+7900 val perplexity 385.5467
+7900 train 5.925571 (lr=9.9088e-06) (hash(x)=146417668)
+8000 val loss 5.9517
+8000 val perplexity 384.4099
+8000 train 5.756671 (lr=9.4682e-06) (hash(x)=150165488)
+8100 val loss 5.9492
+8100 val perplexity 383.4286
+8100 train 5.928113 (lr=9.0461e-06) (hash(x)=144818855)
+8200 val loss 5.9425
+8200 val perplexity 380.8962
+8200 train 6.034739 (lr=8.6430e-06) (hash(x)=143074954)
+8300 val loss 5.9406
+8300 val perplexity 380.1790
+8300 train 5.819485 (lr=8.2593e-06) (hash(x)=145172494)
+8400 val loss 5.9388
+8400 val perplexity 379.4785
+8400 train 5.967767 (lr=7.8953e-06) (hash(x)=152284393)
+8500 val loss 5.9353
+8500 val perplexity 378.1628
+8500 train 5.779359 (lr=7.5515e-06) (hash(x)=147044705)
+8600 val loss 5.9345
+8600 val perplexity 377.8589
+8600 train 5.811585 (lr=7.2282e-06) (hash(x)=147762608)
+8700 val loss 5.9323
+8700 val perplexity 377.0360
+8700 train 5.942141 (lr=6.9257e-06) (hash(x)=141994567)
+8800 val loss 5.9272
+8800 val perplexity 375.1202
+8800 train 6.021777 (lr=6.6444e-06) (hash(x)=152552896)
+8900 val loss 5.9233
+8900 val perplexity 373.6460
+8900 train 5.762463 (lr=6.3845e-06) (hash(x)=139714454)
+9000 val loss 5.9218
+9000 val perplexity 373.0761
+9000 train 5.954716 (lr=6.1462e-06) (hash(x)=152274419)
+9100 val loss 5.9213
+9100 val perplexity 372.9054
+9100 train 6.007237 (lr=5.9300e-06) (hash(x)=160536885)
+9200 val loss 5.9212
+9200 val perplexity 372.8654
+9200 train 5.734900 (lr=5.7359e-06) (hash(x)=138386899)
+9300 val loss 5.9195
+9300 val perplexity 372.2228
+9300 train 5.970142 (lr=5.5641e-06) (hash(x)=149957457)
+9400 val loss 5.9159
+9400 val perplexity 370.8707
+9400 train 5.931352 (lr=5.4149e-06) (hash(x)=161671548)
+9500 val loss 5.9132
+9500 val perplexity 369.8911
+9500 train 6.040483 (lr=5.2884e-06) (hash(x)=171857400)
+9600 val loss 5.9124
+9600 val perplexity 369.6096
+9600 train 5.686675 (lr=5.1847e-06) (hash(x)=135378471)
+9700 val loss 5.9109
+9700 val perplexity 369.0223
+9700 train 5.833959 (lr=5.1040e-06) (hash(x)=149514591)
+9800 val loss 5.9099
+9800 val perplexity 368.6873
+9800 train 5.888211 (lr=5.0462e-06) (hash(x)=152679072)
+9900 val loss 5.9073
+9900 val perplexity 367.6994
+9900 train 5.844470 (lr=5.0116e-06) (hash(x)=150415193)
+9999 val loss 5.9055
+9999 val perplexity 367.0529

attention_kindselective_n_heads2_seed1341/model_02500.pt

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:ea5a9ffee08c2df305cd63f255c66725128c7bb6e3bf9c755dd82a9739aedaf0
+oid sha256:74ba334bd0101aaadc01e26184cae068161abfa0ae4d46c807c3442a5ced88c6
 size 38587970

attention_kindselective_n_heads2_seed1341/model_05000.pt

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:79dea9083038faa3d1bcc51cc54cd099869a120fb53489c85b8689ea587902fb
+oid sha256:b16d132361577f9fe65063258b176bda59304031f30e5218be876604b731f064
 size 38587970

attention_kindselective_n_heads2_seed1341/model_07500.pt

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:30822baa6d6425e32309da513b9d559dcf796e031b8d56a03bb4e4064848c217
+oid sha256:ad2fa889095cb76098814eb3719349a2a88f7883160df398bb236eae2e0dd419
 size 38587970

attention_kindselective_n_heads2_seed1341/model_09999.pt

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:08a2dc8c7d7d2e896fb7212afcaf7421c83e9b52e452c7dd25ea0b5c7c65d081
+oid sha256:7a9877f2d1585c73ca2a2d1529be8d7c23b4402dfb2488a72c5b2e4f1e7f289f
 size 38587970

attention_kindselective_n_heads2_seed1341/optimizer_02500.pt

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:fe779aeeffb6e9af4d60ed5745e01d01e441578cacff5bb2e370062f5a5795f0
+oid sha256:c592f311f13824dc70dda18105f0323da847beb468ad3662628d16c3f1f38bdf
 size 70895430

attention_kindselective_n_heads2_seed1341/optimizer_05000.pt

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:d0a3d427e1b44a10639415b4a0efddb95c41c9007926359e498279a780f501be
+oid sha256:bf007fb4171da723c27811bcb2b34a91e5f5b47f68dfe98b97dd74e61c5ec415
 size 70895430

attention_kindselective_n_heads2_seed1341/optimizer_07500.pt

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:208721297887e24ac34e9421700768e117449cd455e38f2d1f66cadb629b12b3
+oid sha256:da0337e3b4da4c98a4a70a4614120ce958de075a8f71d2460b983614a4d736a7
 size 70895430

attention_kindselective_n_heads2_seed1341/optimizer_09999.pt

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:56ba808dfb6d1a1a89f577bed7918728e41144a989442de66b83e5b73d171f3c
+oid sha256:849ec8ad2b98bcb82237741dbdd657d3a1463e7ec0077e7ae9bca8af3a98c953
 size 70895430