Upload folder using huggingface_hub

att_conv_playground/i_small_heads_wd_2_latent_masks_seed_1339/args.json

@@ -1 +1 @@
-{"hellaswag": true, "attention_kind": "selective", "log_dir": "att_conv_playground/i_small_heads_wd_2_latent_masks_seed_1339", "resume_checkpoint": null, "resume_optimizer": false, "add_a_head": false, "add_head_to_start": true, "new_head_init": "normal", "n_heads": 24, "n_embd": 264, "head_dim": 11, "protect_bos_token": true, "prevent_from_masking_myself": true, "max_steps": 4375, "warmup_steps": 250, "group": "att_conv_playground", "use_wandb": true, "kill_self_after_run": false, "random_seed": 1339, "memory_penalty_epsilon": 0.1, "selection_head_linear_combo": "n_latent_masks", "selection_head_linear_combo_scale": 1.0, "disable_selection_head_linear_combo_bias": false, "assert_latent_matches_no_head": false, "n_latent_masks": 2, "init_latent_masks_to_identity": true, "init_latent_masks_to_inverse": false, "latent_mask_scale": null, "latent_mask_runtime_multiplier": null, "latent_mask_sigmoid": false, "S_layernorm": false, "one_head_per_latent_mask": true, "att_conv": true, "att_conv_init": "eye", "att_conv_scale": 1.0, "att_conv_weight_decay": true, "protection_kind": "none", "leaky_relu_alpha": null, "leaky_relu_bias": null, "use_compile": true, "use_mini_model": false, "upload_to_hf": true, "seq_len": 256, "batch_size": 32, "total_batch_size": 131072, "protection_head_scaling_factor": 1.0, "protection_head_bias": 0.0, "n_sliced_masks": null, "mask_layernorm": false, "residual_attention_masks": false, "compute_base_shapes": false, "base_shapes_savefile": null, "mup": true, "disable_selection": false, "mup_enable_coord_check_logging": false, "max_lr": 0.003, "decay_lr": true, "readout_zero_init": false, "query_zero_init": false, "l1_loss": false, "debugpy": false, "key": "i_small_heads_wd_2_latent_masks", "latent_mask_precision": "float32", "att_conv_precision": "bfloat16"}
+{"hellaswag": true, "attention_kind": "selective", "log_dir": "att_conv_playground/i_small_heads_wd_2_latent_masks_seed_1339", "resume_checkpoint": null, "resume_optimizer": false, "add_a_head": false, "add_head_to_start": true, "new_head_init": "normal", "n_heads": 22, "n_embd": 264, "head_dim": 11, "protect_bos_token": true, "prevent_from_masking_myself": true, "max_steps": 4375, "warmup_steps": 250, "group": "att_conv_playground", "use_wandb": true, "kill_self_after_run": false, "random_seed": 1339, "memory_penalty_epsilon": 0.1, "selection_head_linear_combo": "n_latent_masks", "selection_head_linear_combo_scale": 1.0, "disable_selection_head_linear_combo_bias": false, "assert_latent_matches_no_head": false, "n_latent_masks": 4, "init_latent_masks_to_identity": true, "init_latent_masks_to_inverse": false, "latent_mask_scale": null, "latent_mask_runtime_multiplier": null, "latent_mask_sigmoid": false, "S_layernorm": false, "one_head_per_latent_mask": true, "att_conv": true, "att_conv_init": "eye", "att_conv_scale": 1.0, "att_conv_weight_decay": true, "protection_kind": "none", "leaky_relu_alpha": null, "leaky_relu_bias": null, "use_compile": true, "use_mini_model": false, "upload_to_hf": true, "seq_len": 256, "batch_size": 32, "total_batch_size": 131072, "protection_head_scaling_factor": 1.0, "protection_head_bias": 0.0, "n_sliced_masks": null, "mask_layernorm": false, "residual_attention_masks": false, "compute_base_shapes": false, "base_shapes_savefile": null, "mup": true, "disable_selection": false, "mup_enable_coord_check_logging": false, "max_lr": 0.003, "decay_lr": true, "readout_zero_init": false, "query_zero_init": false, "l1_loss": false, "debugpy": false, "key": "i_small_heads_wd_2_latent_masks", "latent_mask_precision": "float32", "att_conv_precision": "bfloat16"}

att_conv_playground/i_small_heads_wd_2_latent_masks_seed_1339/log2.txt

@@ -1,529 +1,529 @@
 max_steps: 4375
-0 val loss 11.3717
-0 val perplexity 86825.3984
-0 train 11.363577 (lr=4.1958e-06) (hash(x)=44992657)
-10 train 9.922216 (lr=4.6154e-05) (hash(x)=33468021)
-20 train 9.374759 (lr=8.8112e-05) (hash(x)=40941803)
-30 train 8.678726 (lr=1.3007e-04) (hash(x)=36715902)
-40 train 7.961945 (lr=1.7203e-04) (hash(x)=32710993)
-50 train 7.699759 (lr=2.1399e-04) (hash(x)=43839896)
-60 train 7.751177 (lr=2.5594e-04) (hash(x)=40675468)
-70 train 7.417886 (lr=2.9790e-04) (hash(x)=34592631)
-80 train 7.395207 (lr=3.3986e-04) (hash(x)=44444845)
-90 train 7.187118 (lr=3.8182e-04) (hash(x)=41965258)
-100 val loss 7.0550
-100 val perplexity 1158.6124
-100 train 6.913579 (lr=4.2378e-04) (hash(x)=41284750)
-110 train 6.851485 (lr=4.6573e-04) (hash(x)=41118734)
-120 train 6.700687 (lr=5.0769e-04) (hash(x)=37537547)
-130 train 6.547397 (lr=5.4965e-04) (hash(x)=43625179)
-140 train 6.525340 (lr=5.9161e-04) (hash(x)=41940760)
-150 train 6.595750 (lr=6.3357e-04) (hash(x)=39210431)
-160 train 6.500293 (lr=6.7552e-04) (hash(x)=41128294)
-170 train 6.408043 (lr=7.1748e-04) (hash(x)=41590227)
-180 train 6.342257 (lr=7.5944e-04) (hash(x)=38084871)
-190 train 6.252754 (lr=8.0140e-04) (hash(x)=34534333)
-200 val loss 6.1526
-200 val perplexity 469.9528
-200 train 6.191759 (lr=8.4336e-04) (hash(x)=41614299)
-210 train 6.071459 (lr=8.8531e-04) (hash(x)=35551597)
-220 train 5.996685 (lr=9.2727e-04) (hash(x)=44615432)
-230 train 6.131633 (lr=9.6923e-04) (hash(x)=37078768)
-240 train 5.840449 (lr=1.0112e-03) (hash(x)=40644078)
-250 train 5.889553 (lr=1.0531e-03) (hash(x)=44636066)
-260 train 5.777018 (lr=1.0951e-03) (hash(x)=43247952)
-270 train 5.736223 (lr=1.1371e-03) (hash(x)=37989370)
-280 train 5.658681 (lr=1.1790e-03) (hash(x)=36229730)
-290 train 5.593200 (lr=1.2210e-03) (hash(x)=35032996)
-300 val loss 5.7312
-300 val perplexity 308.3363
-300 train 5.473669 (lr=1.2629e-03) (hash(x)=33868636)
-310 train 5.389345 (lr=1.3049e-03) (hash(x)=39782705)
-320 train 5.486483 (lr=1.3469e-03) (hash(x)=43525765)
-330 train 5.566754 (lr=1.3888e-03) (hash(x)=38074926)
-340 train 5.623736 (lr=1.4308e-03) (hash(x)=50125702)
-350 train 5.597556 (lr=1.4727e-03) (hash(x)=35062222)
-360 train 5.524805 (lr=1.5147e-03) (hash(x)=39671001)
-370 train 5.572453 (lr=1.5566e-03) (hash(x)=35748364)
-380 train 5.367682 (lr=1.5986e-03) (hash(x)=39962666)
-390 train 5.329647 (lr=1.6406e-03) (hash(x)=39213360)
-400 val loss 5.3464
-400 val perplexity 209.8587
-400 train 5.324333 (lr=1.6825e-03) (hash(x)=37919584)
-410 train 5.278996 (lr=1.7245e-03) (hash(x)=42001527)
-420 train 5.268399 (lr=1.7664e-03) (hash(x)=33940005)
-430 train 5.307408 (lr=1.8084e-03) (hash(x)=41749217)
-440 train 5.188647 (lr=1.8503e-03) (hash(x)=45219739)
-450 train 5.177632 (lr=1.8923e-03) (hash(x)=38515090)
-460 train 5.087607 (lr=1.9343e-03) (hash(x)=37518607)
-470 train 4.881213 (lr=1.9762e-03) (hash(x)=46830716)
-480 train 5.026694 (lr=2.0182e-03) (hash(x)=49262598)
-490 train 4.819833 (lr=2.0601e-03) (hash(x)=41370559)
-500 val loss 5.1326
-500 val perplexity 169.4548
-500 train 4.905719 (lr=2.1021e-03) (hash(x)=35237459)
-510 train 5.010959 (lr=2.1441e-03) (hash(x)=35070579)
-520 train 5.061775 (lr=2.1860e-03) (hash(x)=37355110)
-530 train 5.144450 (lr=2.2280e-03) (hash(x)=39398044)
-540 train 5.065890 (lr=2.2699e-03) (hash(x)=39156676)
-550 train 5.013740 (lr=2.3119e-03) (hash(x)=38553909)
-560 train 5.080887 (lr=2.3538e-03) (hash(x)=36555371)
-570 train 4.975443 (lr=2.3958e-03) (hash(x)=42048991)
-580 train 4.926690 (lr=2.4378e-03) (hash(x)=37359399)
-590 train 4.973100 (lr=2.4797e-03) (hash(x)=39254663)
-600 val loss 4.9274
-600 val perplexity 138.0236
-600 train 4.933433 (lr=2.5217e-03) (hash(x)=38982539)
-610 train 4.852407 (lr=2.5636e-03) (hash(x)=35859145)
-620 train 4.837459 (lr=2.6056e-03) (hash(x)=35593542)
-630 train 4.754174 (lr=2.6476e-03) (hash(x)=39124418)
-640 train 4.744407 (lr=2.6895e-03) (hash(x)=39107374)
-650 train 4.673471 (lr=2.7315e-03) (hash(x)=34529151)
-660 train 4.607458 (lr=2.7734e-03) (hash(x)=39483931)
-670 train 4.589476 (lr=2.8154e-03) (hash(x)=43008010)
-680 train 4.502759 (lr=2.8573e-03) (hash(x)=33469328)
-690 train 4.547626 (lr=2.8993e-03) (hash(x)=35025253)
-700 val loss 4.7850
-700 val perplexity 119.6962
-700 train 4.515065 (lr=2.9413e-03) (hash(x)=43280188)
-710 train 4.817590 (lr=2.9832e-03) (hash(x)=40999263)
-720 train 4.786658 (lr=3.0000e-03) (hash(x)=38304742)
-730 train 4.792525 (lr=2.9999e-03) (hash(x)=34705188)
-740 train 4.745008 (lr=2.9997e-03) (hash(x)=40028964)
-750 train 4.689537 (lr=2.9994e-03) (hash(x)=39051591)
-760 train 4.738353 (lr=2.9990e-03) (hash(x)=41286198)
-770 train 4.677357 (lr=2.9985e-03) (hash(x)=44512967)
-780 train 4.705936 (lr=2.9979e-03) (hash(x)=45171627)
-790 train 4.664905 (lr=2.9972e-03) (hash(x)=46345673)
-800 val loss 4.6640
-800 val perplexity 106.0645
-800 train 4.625056 (lr=2.9964e-03) (hash(x)=35529545)
-810 train 4.543311 (lr=2.9955e-03) (hash(x)=42321793)
-820 train 4.543081 (lr=2.9945e-03) (hash(x)=49327569)
-830 train 4.616230 (lr=2.9934e-03) (hash(x)=41325057)
-840 train 4.561257 (lr=2.9922e-03) (hash(x)=37305614)
-850 train 4.519702 (lr=2.9909e-03) (hash(x)=38000800)
-860 train 4.635437 (lr=2.9896e-03) (hash(x)=44047697)
-870 train 4.637448 (lr=2.9881e-03) (hash(x)=41576206)
-880 train 4.608762 (lr=2.9865e-03) (hash(x)=44685458)
-890 train 4.589599 (lr=2.9848e-03) (hash(x)=40398597)
-900 val loss 4.5638
-900 val perplexity 95.9509
-900 train 4.526895 (lr=2.9830e-03) (hash(x)=39312673)
-910 train 4.544037 (lr=2.9811e-03) (hash(x)=38881875)
-920 train 4.566689 (lr=2.9792e-03) (hash(x)=39140166)
-930 train 4.465323 (lr=2.9771e-03) (hash(x)=39075781)
-940 train 4.481410 (lr=2.9749e-03) (hash(x)=46887071)
-950 train 4.476902 (lr=2.9726e-03) (hash(x)=39740603)
-960 train 4.521322 (lr=2.9703e-03) (hash(x)=39590268)
-970 train 4.476286 (lr=2.9678e-03) (hash(x)=43580052)
-980 train 4.420864 (lr=2.9652e-03) (hash(x)=40226735)
-990 train 4.520546 (lr=2.9626e-03) (hash(x)=40723896)
-1000 val loss 4.4791
-1000 val perplexity 88.1552
-1000 train 4.480749 (lr=2.9598e-03) (hash(x)=39408050)
-1010 train 4.672426 (lr=2.9570e-03) (hash(x)=32247950)
-1020 train 4.555251 (lr=2.9540e-03) (hash(x)=39049938)
-1030 train 4.519478 (lr=2.9510e-03) (hash(x)=37453736)
-1040 train 4.429589 (lr=2.9478e-03) (hash(x)=36649666)
-1050 train 4.493746 (lr=2.9446e-03) (hash(x)=41404097)
-1060 train 4.484105 (lr=2.9412e-03) (hash(x)=40701599)
-1070 train 4.501267 (lr=2.9378e-03) (hash(x)=41787393)
-1080 train 4.388685 (lr=2.9343e-03) (hash(x)=42325465)
-1090 train 4.426929 (lr=2.9307e-03) (hash(x)=45018925)
-1100 val loss 4.4183
-1100 val perplexity 82.9554
-1100 train 4.446635 (lr=2.9270e-03) (hash(x)=53751982)
-1110 train 4.370152 (lr=2.9231e-03) (hash(x)=43118376)
-1120 train 4.402332 (lr=2.9192e-03) (hash(x)=41489942)
-1130 train 4.420882 (lr=2.9152e-03) (hash(x)=48020487)
-1140 train 4.280365 (lr=2.9112e-03) (hash(x)=43266072)
-1150 train 4.487528 (lr=2.9070e-03) (hash(x)=41790132)
-1160 train 4.445304 (lr=2.9027e-03) (hash(x)=33739090)
-1170 train 4.420757 (lr=2.8983e-03) (hash(x)=43885030)
-1180 train 4.368628 (lr=2.8939e-03) (hash(x)=39046714)
-1190 train 4.399640 (lr=2.8893e-03) (hash(x)=44387023)
-1200 val loss 4.3480
-1200 val perplexity 77.3216
-1200 train 4.260426 (lr=2.8847e-03) (hash(x)=36667141)
-1210 train 4.338835 (lr=2.8800e-03) (hash(x)=45487179)
-1220 train 4.315816 (lr=2.8751e-03) (hash(x)=43456413)
-1230 train 4.248469 (lr=2.8702e-03) (hash(x)=42473632)
-1240 train 4.327480 (lr=2.8652e-03) (hash(x)=42966315)
-1250 train 4.178415 (lr=2.8601e-03) (hash(x)=42930524)
-1260 train 4.320940 (lr=2.8550e-03) (hash(x)=43467676)
-1270 train 4.246199 (lr=2.8497e-03) (hash(x)=40872580)
-1280 train 4.317531 (lr=2.8443e-03) (hash(x)=40619157)
-1290 train 4.327679 (lr=2.8389e-03) (hash(x)=44406165)
-1300 val loss 4.3118
-1300 val perplexity 74.5715
-1300 train 4.291611 (lr=2.8333e-03) (hash(x)=40808029)
-1310 train 4.469441 (lr=2.8277e-03) (hash(x)=46125736)
-1320 train 4.235446 (lr=2.8220e-03) (hash(x)=41698487)
-1330 train 4.268370 (lr=2.8162e-03) (hash(x)=39302878)
-1340 train 4.383691 (lr=2.8103e-03) (hash(x)=41826369)
-1350 train 4.202989 (lr=2.8044e-03) (hash(x)=37061413)
-1360 train 4.325859 (lr=2.7983e-03) (hash(x)=40770974)
-1370 train 4.238734 (lr=2.7922e-03) (hash(x)=38738718)
-1380 train 4.229860 (lr=2.7860e-03) (hash(x)=34961558)
-1390 train 4.285625 (lr=2.7797e-03) (hash(x)=38559342)
-1400 val loss 4.2974
-1400 val perplexity 73.5120
-1400 train 4.240549 (lr=2.7733e-03) (hash(x)=38840020)
-1410 train 4.395767 (lr=2.7668e-03) (hash(x)=44448100)
-1420 train 4.446609 (lr=2.7603e-03) (hash(x)=39199838)
-1430 train 4.242622 (lr=2.7536e-03) (hash(x)=41669873)
-1440 train 4.373518 (lr=2.7469e-03) (hash(x)=53253836)
-1450 train 4.293159 (lr=2.7401e-03) (hash(x)=41320631)
-1460 train 4.212480 (lr=2.7333e-03) (hash(x)=48462803)
-1470 train 4.253891 (lr=2.7263e-03) (hash(x)=42356864)
-1480 train 4.305854 (lr=2.7193e-03) (hash(x)=41118068)
-1490 train 4.219227 (lr=2.7121e-03) (hash(x)=37242854)
-1500 val loss 4.2488
-1500 val perplexity 70.0222
-1500 train 4.264964 (lr=2.7050e-03) (hash(x)=40571031)
-1510 train 4.219374 (lr=2.6977e-03) (hash(x)=35384608)
-1520 train 4.243370 (lr=2.6903e-03) (hash(x)=35341056)
-1530 train 4.140860 (lr=2.6829e-03) (hash(x)=36471779)
-1540 train 4.172465 (lr=2.6754e-03) (hash(x)=40356153)
-1550 train 4.217282 (lr=2.6678e-03) (hash(x)=40133377)
-1560 train 4.143442 (lr=2.6602e-03) (hash(x)=33722160)
-1570 train 4.294497 (lr=2.6525e-03) (hash(x)=39555295)
-1580 train 4.189536 (lr=2.6447e-03) (hash(x)=35950666)
-1590 train 4.345695 (lr=2.6368e-03) (hash(x)=36528535)
-1600 val loss 4.2286
-1600 val perplexity 68.6207
-1600 train 4.230598 (lr=2.6289e-03) (hash(x)=42613017)
-1610 train 4.376665 (lr=2.6208e-03) (hash(x)=42845965)
-1620 train 4.284088 (lr=2.6128e-03) (hash(x)=36137074)
-1630 train 4.227170 (lr=2.6046e-03) (hash(x)=37560569)
-1640 train 4.325464 (lr=2.5964e-03) (hash(x)=43680614)
-1650 train 4.180369 (lr=2.5881e-03) (hash(x)=37281218)
-1660 train 4.206783 (lr=2.5797e-03) (hash(x)=36165222)
-1670 train 4.159068 (lr=2.5713e-03) (hash(x)=50222513)
-1680 train 4.044919 (lr=2.5628e-03) (hash(x)=35239173)
-1690 train 4.103105 (lr=2.5542e-03) (hash(x)=40287349)
-1700 val loss 4.2301
-1700 val perplexity 68.7223
-1700 train 4.290835 (lr=2.5455e-03) (hash(x)=37444780)
-1710 train 4.271445 (lr=2.5368e-03) (hash(x)=38962723)
-1720 train 4.277743 (lr=2.5281e-03) (hash(x)=33882305)
-1730 train 4.255969 (lr=2.5192e-03) (hash(x)=36501540)
-1740 train 4.253308 (lr=2.5103e-03) (hash(x)=38723266)
-1750 train 4.188325 (lr=2.5014e-03) (hash(x)=38772789)
-1760 train 4.264967 (lr=2.4924e-03) (hash(x)=37707898)
-1770 train 4.122534 (lr=2.4833e-03) (hash(x)=30529327)
-1780 train 4.212236 (lr=2.4741e-03) (hash(x)=37885464)
-1790 train 4.224483 (lr=2.4649e-03) (hash(x)=39013967)
-1800 val loss 4.1900
-1800 val perplexity 66.0248
-1800 train 4.262849 (lr=2.4556e-03) (hash(x)=34906955)
-1810 train 4.135907 (lr=2.4463e-03) (hash(x)=40790682)
-1820 train 4.010628 (lr=2.4369e-03) (hash(x)=39912223)
-1830 train 4.132839 (lr=2.4275e-03) (hash(x)=39304486)
-1840 train 4.100446 (lr=2.4180e-03) (hash(x)=37247975)
-1850 train 4.091725 (lr=2.4084e-03) (hash(x)=41340614)
-1860 train 4.249487 (lr=2.3988e-03) (hash(x)=39924627)
-1870 train 4.231539 (lr=2.3891e-03) (hash(x)=40235993)
-1880 train 4.119097 (lr=2.3794e-03) (hash(x)=37693476)
-1890 train 4.238847 (lr=2.3696e-03) (hash(x)=40449864)
-1900 val loss 4.1672
-1900 val perplexity 64.5368
-1900 train 4.147777 (lr=2.3598e-03) (hash(x)=37907749)
-1910 train 4.121119 (lr=2.3499e-03) (hash(x)=34547901)
-1920 train 4.150450 (lr=2.3400e-03) (hash(x)=37771788)
-1930 train 4.025002 (lr=2.3300e-03) (hash(x)=36049454)
-1940 train 4.118981 (lr=2.3200e-03) (hash(x)=29564336)
-1950 train 4.109178 (lr=2.3099e-03) (hash(x)=41689281)
-1960 train 4.021831 (lr=2.2998e-03) (hash(x)=41976699)
-1970 train 4.217256 (lr=2.2896e-03) (hash(x)=41538621)
-1980 train 4.101467 (lr=2.2793e-03) (hash(x)=36490902)
-1990 train 4.133924 (lr=2.2691e-03) (hash(x)=42739315)
-2000 val loss 4.1470
-2000 val perplexity 63.2408
-2000 train 4.148038 (lr=2.2588e-03) (hash(x)=38600074)
-2010 train 4.161505 (lr=2.2484e-03) (hash(x)=34555488)
-2020 train 4.178379 (lr=2.2380e-03) (hash(x)=37227345)
-2030 train 4.130076 (lr=2.2275e-03) (hash(x)=38172386)
-2040 train 4.244411 (lr=2.2170e-03) (hash(x)=39340786)
-2050 train 4.088753 (lr=2.2065e-03) (hash(x)=37237298)
-2060 train 4.324434 (lr=2.1959e-03) (hash(x)=35894782)
-2070 train 4.149876 (lr=2.1853e-03) (hash(x)=44161861)
-2080 train 4.162563 (lr=2.1746e-03) (hash(x)=38062558)
-2090 train 4.012234 (lr=2.1639e-03) (hash(x)=38608301)
-2100 val loss 4.1251
-2100 val perplexity 61.8755
-2100 train 4.045103 (lr=2.1532e-03) (hash(x)=47611537)
-2110 train 4.043654 (lr=2.1424e-03) (hash(x)=38076651)
-2120 train 4.052896 (lr=2.1316e-03) (hash(x)=39870893)
-2130 train 4.200476 (lr=2.1208e-03) (hash(x)=48123321)
-2140 train 4.227482 (lr=2.1099e-03) (hash(x)=44124977)
-2150 train 4.132596 (lr=2.0990e-03) (hash(x)=42181374)
-2160 train 4.162151 (lr=2.0881e-03) (hash(x)=41837878)
-2170 train 4.176892 (lr=2.0771e-03) (hash(x)=42161375)
-2180 train 4.143106 (lr=2.0661e-03) (hash(x)=41792986)
-2190 train 4.112856 (lr=2.0550e-03) (hash(x)=41872416)
-2200 val loss 4.1050
-2200 val perplexity 60.6451
-2200 train 4.088569 (lr=2.0440e-03) (hash(x)=37395985)
-2210 train 4.141064 (lr=2.0329e-03) (hash(x)=40616096)
-2220 train 4.037234 (lr=2.0217e-03) (hash(x)=40118423)
-2230 train 3.955050 (lr=2.0106e-03) (hash(x)=39641586)
-2240 train 4.042272 (lr=1.9994e-03) (hash(x)=34720116)
-2250 train 3.986983 (lr=1.9882e-03) (hash(x)=43641508)
-2260 train 4.015029 (lr=1.9770e-03) (hash(x)=31179786)
-2270 train 3.975202 (lr=1.9657e-03) (hash(x)=42721932)
-2280 train 4.307312 (lr=1.9544e-03) (hash(x)=38474505)
-2290 train 4.129827 (lr=1.9431e-03) (hash(x)=41398545)
-2300 val loss 4.0890
-2300 val perplexity 59.6830
-2300 train 4.058106 (lr=1.9318e-03) (hash(x)=44131094)
-2310 train 4.136691 (lr=1.9205e-03) (hash(x)=41667948)
-2320 train 4.216187 (lr=1.9091e-03) (hash(x)=36339232)
-2330 train 4.097335 (lr=1.8977e-03) (hash(x)=41630644)
-2340 train 4.184237 (lr=1.8863e-03) (hash(x)=41278955)
-2350 train 4.061925 (lr=1.8749e-03) (hash(x)=39491976)
-2360 train 4.120791 (lr=1.8635e-03) (hash(x)=41771238)
-2370 train 4.212980 (lr=1.8520e-03) (hash(x)=45374570)
-2380 train 4.096973 (lr=1.8406e-03) (hash(x)=41419267)
-2390 train 4.127297 (lr=1.8291e-03) (hash(x)=38869169)
-2400 val loss 4.0747
-2400 val perplexity 58.8337
-2400 train 4.028738 (lr=1.8176e-03) (hash(x)=38619293)
-2410 train 3.848051 (lr=1.8061e-03) (hash(x)=41600240)
-2420 train 3.982481 (lr=1.7946e-03) (hash(x)=40891045)
-2430 train 3.973238 (lr=1.7830e-03) (hash(x)=41023249)
-2440 train 3.942607 (lr=1.7715e-03) (hash(x)=33813452)
-2450 train 3.894412 (lr=1.7600e-03) (hash(x)=38464119)
-2460 train 3.793348 (lr=1.7484e-03) (hash(x)=40699982)
-2470 train 3.798558 (lr=1.7368e-03) (hash(x)=38254854)
-2480 train 4.089022 (lr=1.7253e-03) (hash(x)=42736069)
-2490 train 4.096173 (lr=1.7137e-03) (hash(x)=41928525)
-2500 val loss 4.0552
-2500 val perplexity 57.6992
-2500 train 4.051444 (lr=1.7021e-03) (hash(x)=45864011)
-2510 train 4.064780 (lr=1.6906e-03) (hash(x)=41380960)
-2520 train 4.060643 (lr=1.6790e-03) (hash(x)=38366545)
-2530 train 4.070795 (lr=1.6674e-03) (hash(x)=41045176)
-2540 train 4.094515 (lr=1.6558e-03) (hash(x)=43435705)
-2550 train 4.050467 (lr=1.6442e-03) (hash(x)=40693090)
-2560 train 3.964399 (lr=1.6326e-03) (hash(x)=40502478)
-2570 train 4.002869 (lr=1.6210e-03) (hash(x)=38610920)
-2580 train 3.924541 (lr=1.6094e-03) (hash(x)=38333499)
-2590 train 4.000230 (lr=1.5979e-03) (hash(x)=41982736)
-2600 val loss 4.0374
-2600 val perplexity 56.6800
-2600 train 3.979413 (lr=1.5863e-03) (hash(x)=37724702)
-2610 train 3.916890 (lr=1.5747e-03) (hash(x)=39564630)
-2620 train 3.860455 (lr=1.5632e-03) (hash(x)=38465803)
-2630 train 3.785778 (lr=1.5516e-03) (hash(x)=36955007)
-2640 train 3.820798 (lr=1.5400e-03) (hash(x)=32994568)
-2650 train 4.001617 (lr=1.5285e-03) (hash(x)=41165765)
-2660 train 4.017964 (lr=1.5170e-03) (hash(x)=42934000)
-2670 train 4.212893 (lr=1.5054e-03) (hash(x)=42726316)
-2680 train 4.099064 (lr=1.4939e-03) (hash(x)=35880318)
-2690 train 3.982738 (lr=1.4824e-03) (hash(x)=41747329)
-2700 val loss 4.0287
-2700 val perplexity 56.1880
-2700 train 4.103619 (lr=1.4709e-03) (hash(x)=40259630)
-2710 train 4.233391 (lr=1.4594e-03) (hash(x)=39895798)
-2720 train 3.959761 (lr=1.4480e-03) (hash(x)=36146683)
-2730 train 3.940617 (lr=1.4365e-03) (hash(x)=36181984)
-2740 train 3.921445 (lr=1.4251e-03) (hash(x)=43700349)
-2750 train 3.987399 (lr=1.4137e-03) (hash(x)=41610597)
-2760 train 3.975067 (lr=1.4023e-03) (hash(x)=31183639)
-2770 train 3.970547 (lr=1.3909e-03) (hash(x)=37722489)
-2780 train 3.913685 (lr=1.3795e-03) (hash(x)=47290688)
-2790 train 3.952569 (lr=1.3682e-03) (hash(x)=41205574)
-2800 val loss 4.0120
-2800 val perplexity 55.2587
-2800 train 3.805958 (lr=1.3569e-03) (hash(x)=40257962)
-2810 train 3.780186 (lr=1.3456e-03) (hash(x)=39529014)
-2820 train 3.799081 (lr=1.3343e-03) (hash(x)=42244749)
-2830 train 3.854383 (lr=1.3230e-03) (hash(x)=32820090)
-2840 train 3.813854 (lr=1.3118e-03) (hash(x)=40315769)
-2850 train 4.049391 (lr=1.3006e-03) (hash(x)=41524462)
-2860 train 4.043042 (lr=1.2894e-03) (hash(x)=38365734)
-2870 train 4.177433 (lr=1.2783e-03) (hash(x)=37682602)
-2880 train 4.027946 (lr=1.2671e-03) (hash(x)=39162991)
-2890 train 3.886543 (lr=1.2560e-03) (hash(x)=33316384)
-2900 val loss 3.9878
-2900 val perplexity 53.9357
-2900 train 3.934094 (lr=1.2450e-03) (hash(x)=37271132)
-2910 train 4.123441 (lr=1.2339e-03) (hash(x)=35586242)
-2920 train 4.012568 (lr=1.2229e-03) (hash(x)=33320586)
-2930 train 4.025620 (lr=1.2119e-03) (hash(x)=43531361)
-2940 train 3.913611 (lr=1.2010e-03) (hash(x)=37368286)
-2950 train 3.950181 (lr=1.1901e-03) (hash(x)=40363394)
-2960 train 3.918155 (lr=1.1792e-03) (hash(x)=42253792)
-2970 train 3.860552 (lr=1.1684e-03) (hash(x)=38072598)
-2980 train 3.932182 (lr=1.1576e-03) (hash(x)=41470557)
-2990 train 3.858858 (lr=1.1468e-03) (hash(x)=42600033)
-3000 val loss 3.9912
-3000 val perplexity 54.1222
-3000 train 3.814529 (lr=1.1361e-03) (hash(x)=46890983)
-3010 train 3.680441 (lr=1.1254e-03) (hash(x)=37986759)
-3020 train 3.801029 (lr=1.1147e-03) (hash(x)=36270703)
-3030 train 4.068460 (lr=1.1041e-03) (hash(x)=38228599)
-3040 train 3.918266 (lr=1.0935e-03) (hash(x)=44344296)
-3050 train 4.007051 (lr=1.0830e-03) (hash(x)=39962297)
-3060 train 3.984616 (lr=1.0725e-03) (hash(x)=39817394)
-3070 train 3.982054 (lr=1.0620e-03) (hash(x)=41763868)
-3080 train 4.066197 (lr=1.0516e-03) (hash(x)=42343051)
-3090 train 4.064052 (lr=1.0412e-03) (hash(x)=39204893)
-3100 val loss 3.9691
-3100 val perplexity 52.9344
-3100 train 3.964641 (lr=1.0309e-03) (hash(x)=39353599)
-3110 train 4.035765 (lr=1.0207e-03) (hash(x)=36118073)
-3120 train 3.741359 (lr=1.0104e-03) (hash(x)=44070271)
-3130 train 3.763264 (lr=1.0002e-03) (hash(x)=43368151)
-3140 train 3.840649 (lr=9.9011e-04) (hash(x)=40612203)
-3150 train 3.977300 (lr=9.8002e-04) (hash(x)=39598447)
-3160 train 4.026806 (lr=9.6999e-04) (hash(x)=43897682)
-3170 train 4.014703 (lr=9.6000e-04) (hash(x)=39029243)
-3180 train 4.017778 (lr=9.5007e-04) (hash(x)=38586684)
-3190 train 3.969115 (lr=9.4019e-04) (hash(x)=39299115)
-3200 val loss 3.9483
-3200 val perplexity 51.8474
-3200 train 3.997803 (lr=9.3036e-04) (hash(x)=41064949)
-3210 train 3.983777 (lr=9.2058e-04) (hash(x)=41752822)
-3220 train 3.991479 (lr=9.1085e-04) (hash(x)=37484805)
-3230 train 3.953899 (lr=9.0118e-04) (hash(x)=40514919)
-3240 train 4.007973 (lr=8.9157e-04) (hash(x)=42440343)
-3250 train 3.910555 (lr=8.8201e-04) (hash(x)=32464136)
-3260 train 3.926669 (lr=8.7251e-04) (hash(x)=47270558)
-3270 train 3.866263 (lr=8.6307e-04) (hash(x)=39307546)
-3280 train 3.903190 (lr=8.5368e-04) (hash(x)=40410299)
-3290 train 3.904997 (lr=8.4435e-04) (hash(x)=40541497)
-3300 val loss 3.9433
-3300 val perplexity 51.5865
-3300 train 3.800336 (lr=8.3508e-04) (hash(x)=38089139)
-3310 train 3.837263 (lr=8.2588e-04) (hash(x)=42006576)
-3320 train 3.929136 (lr=8.1673e-04) (hash(x)=40422556)
-3330 train 3.721211 (lr=8.0764e-04) (hash(x)=39021643)
-3340 train 3.982839 (lr=7.9862e-04) (hash(x)=40434305)
-3350 train 4.005759 (lr=7.8966e-04) (hash(x)=40833559)
-3360 train 4.008427 (lr=7.8076e-04) (hash(x)=40045231)
-3370 train 3.968306 (lr=7.7192e-04) (hash(x)=34668317)
-3380 train 3.967199 (lr=7.6315e-04) (hash(x)=42738568)
-3390 train 4.063785 (lr=7.5445e-04) (hash(x)=38980166)
-3400 val loss 3.9229
-3400 val perplexity 50.5446
-3400 train 3.973525 (lr=7.4581e-04) (hash(x)=43878176)
-3410 train 4.014478 (lr=7.3724e-04) (hash(x)=38201991)
-3420 train 3.991537 (lr=7.2874e-04) (hash(x)=34905889)
-3430 train 3.991725 (lr=7.2030e-04) (hash(x)=41680405)
-3440 train 3.953187 (lr=7.1193e-04) (hash(x)=39948184)
-3450 train 4.002751 (lr=7.0363e-04) (hash(x)=33623728)
-3460 train 3.877424 (lr=6.9541e-04) (hash(x)=41027647)
-3470 train 3.910935 (lr=6.8725e-04) (hash(x)=42779482)
-3480 train 3.923709 (lr=6.7916e-04) (hash(x)=41579288)
-3490 train 3.778706 (lr=6.7114e-04) (hash(x)=37174846)
-3500 val loss 3.9204
-3500 val perplexity 50.4205
-3500 train 3.658134 (lr=6.6320e-04) (hash(x)=37082090)
-3510 train 3.853891 (lr=6.5533e-04) (hash(x)=43131340)
-3520 train 3.772217 (lr=6.4753e-04) (hash(x)=42457250)
-3530 train 3.998982 (lr=6.3981e-04) (hash(x)=39504492)
-3540 train 3.943151 (lr=6.3216e-04) (hash(x)=36225997)
-3550 train 4.074574 (lr=6.2458e-04) (hash(x)=40912087)
-3560 train 3.912014 (lr=6.1708e-04) (hash(x)=41545925)
-3570 train 4.000611 (lr=6.0966e-04) (hash(x)=32553193)
-3580 train 3.847172 (lr=6.0231e-04) (hash(x)=38799856)
-3590 train 3.893481 (lr=5.9504e-04) (hash(x)=40772075)
-3600 val loss 3.9020
-3600 val perplexity 49.5003
-3600 train 3.936508 (lr=5.8785e-04) (hash(x)=39299903)
-3610 train 3.983123 (lr=5.8074e-04) (hash(x)=39503247)
-3620 train 3.876462 (lr=5.7370e-04) (hash(x)=38730668)
-3630 train 3.897536 (lr=5.6675e-04) (hash(x)=40503276)
-3640 train 3.900669 (lr=5.5987e-04) (hash(x)=36504573)
-3650 train 3.922709 (lr=5.5308e-04) (hash(x)=44903075)
-3660 train 3.828653 (lr=5.4636e-04) (hash(x)=33800148)
-3670 train 3.741939 (lr=5.3973e-04) (hash(x)=39404541)
-3680 train 3.687747 (lr=5.3318e-04) (hash(x)=44777432)
-3690 train 3.784186 (lr=5.2671e-04) (hash(x)=45649913)
-3700 val loss 3.9036
-3700 val perplexity 49.5806
-3700 train 3.911320 (lr=5.2033e-04) (hash(x)=33812537)
-3710 train 4.057878 (lr=5.1402e-04) (hash(x)=36161222)
-3720 train 3.933919 (lr=5.0780e-04) (hash(x)=43690616)
-3730 train 3.941305 (lr=5.0167e-04) (hash(x)=35462313)
-3740 train 4.237491 (lr=4.9562e-04) (hash(x)=44674254)
-3750 train 3.919392 (lr=4.8965e-04) (hash(x)=44276297)
-3760 train 3.964192 (lr=4.8377e-04) (hash(x)=38492178)
-3770 train 3.966371 (lr=4.7798e-04) (hash(x)=44347531)
-3780 train 3.881049 (lr=4.7227e-04) (hash(x)=33792059)
-3790 train 3.978416 (lr=4.6665e-04) (hash(x)=38549318)
-3800 val loss 3.8848
-3800 val perplexity 48.6579
-3800 train 3.856662 (lr=4.6112e-04) (hash(x)=41437448)
-3810 train 3.760668 (lr=4.5567e-04) (hash(x)=36656768)
-3820 train 3.922438 (lr=4.5031e-04) (hash(x)=46387893)
-3830 train 3.918259 (lr=4.4504e-04) (hash(x)=44312129)
-3840 train 3.843921 (lr=4.3986e-04) (hash(x)=40341896)
-3850 train 3.894210 (lr=4.3477e-04) (hash(x)=43589160)
-3860 train 3.875674 (lr=4.2977e-04) (hash(x)=40388945)
-3870 train 3.770464 (lr=4.2486e-04) (hash(x)=33336679)
-3880 train 3.864926 (lr=4.2004e-04) (hash(x)=37643311)
-3890 train 3.888263 (lr=4.1530e-04) (hash(x)=38272927)
-3900 val loss 3.8792
-3900 val perplexity 48.3860
-3900 train 3.833359 (lr=4.1066e-04) (hash(x)=41925748)
-3910 train 3.923103 (lr=4.0611e-04) (hash(x)=38377253)
-3920 train 3.932636 (lr=4.0166e-04) (hash(x)=41616611)
-3930 train 3.854330 (lr=3.9729e-04) (hash(x)=32301827)
-3940 train 3.906909 (lr=3.9302e-04) (hash(x)=47697363)
-3950 train 3.858984 (lr=3.8884e-04) (hash(x)=37867767)
-3960 train 3.860146 (lr=3.8475e-04) (hash(x)=37187295)
-3970 train 3.915128 (lr=3.8076e-04) (hash(x)=41952752)
-3980 train 3.885388 (lr=3.7685e-04) (hash(x)=38358660)
-3990 train 3.756501 (lr=3.7305e-04) (hash(x)=40207878)
-4000 val loss 3.8731
-4000 val perplexity 48.0901
-4000 train 3.842866 (lr=3.6933e-04) (hash(x)=39134015)
-4010 train 3.912690 (lr=3.6572e-04) (hash(x)=38313135)
-4020 train 3.843382 (lr=3.6219e-04) (hash(x)=40710513)
-4030 train 3.969596 (lr=3.5876e-04) (hash(x)=35530165)
-4040 train 3.895833 (lr=3.5543e-04) (hash(x)=39494066)
-4050 train 3.878303 (lr=3.5219e-04) (hash(x)=40861777)
-4060 train 3.866653 (lr=3.4905e-04) (hash(x)=36151939)
-4070 train 3.827314 (lr=3.4600e-04) (hash(x)=39008241)
-4080 train 3.916543 (lr=3.4305e-04) (hash(x)=40395746)
-4090 train 3.953969 (lr=3.4019e-04) (hash(x)=38855480)
-4100 val loss 3.8613
-4100 val perplexity 47.5253
-4100 train 3.893692 (lr=3.3744e-04) (hash(x)=45013254)
-4110 train 3.895224 (lr=3.3477e-04) (hash(x)=39934818)
-4120 train 3.898216 (lr=3.3221e-04) (hash(x)=33417705)
-4130 train 3.835839 (lr=3.2974e-04) (hash(x)=41267978)
-4140 train 3.829896 (lr=3.2737e-04) (hash(x)=38961609)
-4150 train 3.831160 (lr=3.2510e-04) (hash(x)=42173878)
-4160 train 3.794766 (lr=3.2292e-04) (hash(x)=36867405)
-4170 train 3.773464 (lr=3.2085e-04) (hash(x)=38991954)
-4180 train 3.835734 (lr=3.1887e-04) (hash(x)=39032592)
-4190 train 3.812086 (lr=3.1699e-04) (hash(x)=61329402)
-4200 val loss 3.8609
-4200 val perplexity 47.5074
-4200 train 3.824891 (lr=3.1520e-04) (hash(x)=40455200)
-4210 train 3.854889 (lr=3.1352e-04) (hash(x)=40913417)
-4220 train 3.887470 (lr=3.1193e-04) (hash(x)=38809704)
-4230 train 3.867836 (lr=3.1044e-04) (hash(x)=39560204)
-4240 train 3.873663 (lr=3.0905e-04) (hash(x)=41496549)
-4250 train 3.912349 (lr=3.0776e-04) (hash(x)=33994768)
-4260 train 3.898098 (lr=3.0657e-04) (hash(x)=39658687)
-4270 train 3.774222 (lr=3.0548e-04) (hash(x)=42583064)
-4280 train 3.840446 (lr=3.0449e-04) (hash(x)=40635811)
-4290 train 3.738791 (lr=3.0359e-04) (hash(x)=35313841)
-4300 val loss 3.8585
-4300 val perplexity 47.3941
-4300 train 3.921301 (lr=3.0280e-04) (hash(x)=39332245)
-4310 train 3.818811 (lr=3.0210e-04) (hash(x)=43732999)
-4320 train 3.821961 (lr=3.0150e-04) (hash(x)=40831835)
-4330 train 3.961745 (lr=3.0101e-04) (hash(x)=41813368)
-4340 train 3.863962 (lr=3.0061e-04) (hash(x)=36952586)
-4350 train 3.885816 (lr=3.0031e-04) (hash(x)=42470745)
-4360 train 3.813696 (lr=3.0011e-04) (hash(x)=39581495)
-4370 train 4.003681 (lr=3.0001e-04) (hash(x)=35914538)
-4374 val loss 3.8514
-4374 val perplexity 47.0584
+0 val loss 11.2835
+0 val perplexity 79495.7422
+0 train 11.290979 (lr=4.1958e-06) (hash(x)=44992657)
+10 train 9.967012 (lr=4.6154e-05) (hash(x)=33468021)
+20 train 9.386133 (lr=8.8112e-05) (hash(x)=40941803)
+30 train 8.686039 (lr=1.3007e-04) (hash(x)=36715902)
+40 train 7.944033 (lr=1.7203e-04) (hash(x)=32710993)
+50 train 7.708426 (lr=2.1399e-04) (hash(x)=43839896)
+60 train 7.768104 (lr=2.5594e-04) (hash(x)=40675468)
+70 train 7.454903 (lr=2.9790e-04) (hash(x)=34592631)
+80 train 7.425917 (lr=3.3986e-04) (hash(x)=44444845)
+90 train 7.210289 (lr=3.8182e-04) (hash(x)=41965258)
+100 val loss 7.0900
+100 val perplexity 1199.8834
+100 train 6.943618 (lr=4.2378e-04) (hash(x)=41284750)
+110 train 6.879869 (lr=4.6573e-04) (hash(x)=41118734)
+120 train 6.713333 (lr=5.0769e-04) (hash(x)=37537547)
+130 train 6.566779 (lr=5.4965e-04) (hash(x)=43625179)
+140 train 6.535774 (lr=5.9161e-04) (hash(x)=41940760)
+150 train 6.596593 (lr=6.3357e-04) (hash(x)=39210431)
+160 train 6.543444 (lr=6.7552e-04) (hash(x)=41128294)
+170 train 6.445559 (lr=7.1748e-04) (hash(x)=41590227)
+180 train 6.361904 (lr=7.5944e-04) (hash(x)=38084871)
+190 train 6.270483 (lr=8.0140e-04) (hash(x)=34534333)
+200 val loss 6.1927
+200 val perplexity 489.1622
+200 train 6.234326 (lr=8.4336e-04) (hash(x)=41614299)
+210 train 6.105705 (lr=8.8531e-04) (hash(x)=35551597)
+220 train 6.018891 (lr=9.2727e-04) (hash(x)=44615432)
+230 train 6.168056 (lr=9.6923e-04) (hash(x)=37078768)
+240 train 5.860798 (lr=1.0112e-03) (hash(x)=40644078)
+250 train 5.914165 (lr=1.0531e-03) (hash(x)=44636066)
+260 train 5.808303 (lr=1.0951e-03) (hash(x)=43247952)
+270 train 5.739007 (lr=1.1371e-03) (hash(x)=37989370)
+280 train 5.688848 (lr=1.1790e-03) (hash(x)=36229730)
+290 train 5.625371 (lr=1.2210e-03) (hash(x)=35032996)
+300 val loss 5.7157
+300 val perplexity 303.5841
+300 train 5.452739 (lr=1.2629e-03) (hash(x)=33868636)
+310 train 5.389076 (lr=1.3049e-03) (hash(x)=39782705)
+320 train 5.493956 (lr=1.3469e-03) (hash(x)=43525765)
+330 train 5.583786 (lr=1.3888e-03) (hash(x)=38074926)
+340 train 5.623821 (lr=1.4308e-03) (hash(x)=50125702)
+350 train 5.601887 (lr=1.4727e-03) (hash(x)=35062222)
+360 train 5.536216 (lr=1.5147e-03) (hash(x)=39671001)
+370 train 5.570259 (lr=1.5566e-03) (hash(x)=35748364)
+380 train 5.366518 (lr=1.5986e-03) (hash(x)=39962666)
+390 train 5.338472 (lr=1.6406e-03) (hash(x)=39213360)
+400 val loss 5.3521
+400 val perplexity 211.0562
+400 train 5.336286 (lr=1.6825e-03) (hash(x)=37919584)
+410 train 5.288481 (lr=1.7245e-03) (hash(x)=42001527)
+420 train 5.276488 (lr=1.7664e-03) (hash(x)=33940005)
+430 train 5.319012 (lr=1.8084e-03) (hash(x)=41749217)
+440 train 5.209821 (lr=1.8503e-03) (hash(x)=45219739)
+450 train 5.202966 (lr=1.8923e-03) (hash(x)=38515090)
+460 train 5.107697 (lr=1.9343e-03) (hash(x)=37518607)
+470 train 4.909622 (lr=1.9762e-03) (hash(x)=46830716)
+480 train 5.021257 (lr=2.0182e-03) (hash(x)=49262598)
+490 train 4.841714 (lr=2.0601e-03) (hash(x)=41370559)
+500 val loss 5.1417
+500 val perplexity 171.0144
+500 train 4.923532 (lr=2.1021e-03) (hash(x)=35237459)
+510 train 5.049324 (lr=2.1441e-03) (hash(x)=35070579)
+520 train 5.089362 (lr=2.1860e-03) (hash(x)=37355110)
+530 train 5.168279 (lr=2.2280e-03) (hash(x)=39398044)
+540 train 5.100167 (lr=2.2699e-03) (hash(x)=39156676)
+550 train 5.044903 (lr=2.3119e-03) (hash(x)=38553909)
+560 train 5.109780 (lr=2.3538e-03) (hash(x)=36555371)
+570 train 5.016411 (lr=2.3958e-03) (hash(x)=42048991)
+580 train 4.954191 (lr=2.4378e-03) (hash(x)=37359399)
+590 train 4.982177 (lr=2.4797e-03) (hash(x)=39254663)
+600 val loss 4.9401
+600 val perplexity 139.7894
+600 train 4.951568 (lr=2.5217e-03) (hash(x)=38982539)
+610 train 4.876301 (lr=2.5636e-03) (hash(x)=35859145)
+620 train 4.862639 (lr=2.6056e-03) (hash(x)=35593542)
+630 train 4.757588 (lr=2.6476e-03) (hash(x)=39124418)
+640 train 4.755217 (lr=2.6895e-03) (hash(x)=39107374)
+650 train 4.696900 (lr=2.7315e-03) (hash(x)=34529151)
+660 train 4.609609 (lr=2.7734e-03) (hash(x)=39483931)
+670 train 4.610886 (lr=2.8154e-03) (hash(x)=43008010)
+680 train 4.519474 (lr=2.8573e-03) (hash(x)=33469328)
+690 train 4.562234 (lr=2.8993e-03) (hash(x)=35025253)
+700 val loss 4.8037
+700 val perplexity 121.9615
+700 train 4.525724 (lr=2.9413e-03) (hash(x)=43280188)
+710 train 4.817435 (lr=2.9832e-03) (hash(x)=40999263)
+720 train 4.784417 (lr=3.0000e-03) (hash(x)=38304742)
+730 train 4.795308 (lr=2.9999e-03) (hash(x)=34705188)
+740 train 4.757912 (lr=2.9997e-03) (hash(x)=40028964)
+750 train 4.705203 (lr=2.9994e-03) (hash(x)=39051591)
+760 train 4.748432 (lr=2.9990e-03) (hash(x)=41286198)
+770 train 4.693416 (lr=2.9985e-03) (hash(x)=44512967)
+780 train 4.718927 (lr=2.9979e-03) (hash(x)=45171627)
+790 train 4.684438 (lr=2.9972e-03) (hash(x)=46345673)
+800 val loss 4.6812
+800 val perplexity 107.8989
+800 train 4.642310 (lr=2.9964e-03) (hash(x)=35529545)
+810 train 4.542624 (lr=2.9955e-03) (hash(x)=42321793)
+820 train 4.568599 (lr=2.9945e-03) (hash(x)=49327569)
+830 train 4.631002 (lr=2.9934e-03) (hash(x)=41325057)
+840 train 4.571959 (lr=2.9922e-03) (hash(x)=37305614)
+850 train 4.526261 (lr=2.9909e-03) (hash(x)=38000800)
+860 train 4.653804 (lr=2.9896e-03) (hash(x)=44047697)
+870 train 4.645876 (lr=2.9881e-03) (hash(x)=41576206)
+880 train 4.619092 (lr=2.9865e-03) (hash(x)=44685458)
+890 train 4.603628 (lr=2.9848e-03) (hash(x)=40398597)
+900 val loss 4.5726
+900 val perplexity 96.7920
+900 train 4.538272 (lr=2.9830e-03) (hash(x)=39312673)
+910 train 4.555772 (lr=2.9811e-03) (hash(x)=38881875)
+920 train 4.588659 (lr=2.9792e-03) (hash(x)=39140166)
+930 train 4.481109 (lr=2.9771e-03) (hash(x)=39075781)
+940 train 4.498732 (lr=2.9749e-03) (hash(x)=46887071)
+950 train 4.491267 (lr=2.9726e-03) (hash(x)=39740603)
+960 train 4.538983 (lr=2.9703e-03) (hash(x)=39590268)
+970 train 4.495248 (lr=2.9678e-03) (hash(x)=43580052)
+980 train 4.424994 (lr=2.9652e-03) (hash(x)=40226735)
+990 train 4.537102 (lr=2.9626e-03) (hash(x)=40723896)
+1000 val loss 4.4893
+1000 val perplexity 89.0589
+1000 train 4.491249 (lr=2.9598e-03) (hash(x)=39408050)
+1010 train 4.680374 (lr=2.9570e-03) (hash(x)=32247950)
+1020 train 4.570291 (lr=2.9540e-03) (hash(x)=39049938)
+1030 train 4.537564 (lr=2.9510e-03) (hash(x)=37453736)
+1040 train 4.439554 (lr=2.9478e-03) (hash(x)=36649666)
+1050 train 4.504604 (lr=2.9446e-03) (hash(x)=41404097)
+1060 train 4.494673 (lr=2.9412e-03) (hash(x)=40701599)
+1070 train 4.509439 (lr=2.9378e-03) (hash(x)=41787393)
+1080 train 4.399241 (lr=2.9343e-03) (hash(x)=42325465)
+1090 train 4.430777 (lr=2.9307e-03) (hash(x)=45018925)
+1100 val loss 4.4258
+1100 val perplexity 83.5794
+1100 train 4.454267 (lr=2.9270e-03) (hash(x)=53751982)
+1110 train 4.375455 (lr=2.9231e-03) (hash(x)=43118376)
+1120 train 4.406147 (lr=2.9192e-03) (hash(x)=41489942)
+1130 train 4.437531 (lr=2.9152e-03) (hash(x)=48020487)
+1140 train 4.292471 (lr=2.9112e-03) (hash(x)=43266072)
+1150 train 4.486030 (lr=2.9070e-03) (hash(x)=41790132)
+1160 train 4.459498 (lr=2.9027e-03) (hash(x)=33739090)
+1170 train 4.435995 (lr=2.8983e-03) (hash(x)=43885030)
+1180 train 4.382488 (lr=2.8939e-03) (hash(x)=39046714)
+1190 train 4.415123 (lr=2.8893e-03) (hash(x)=44387023)
+1200 val loss 4.3627
+1200 val perplexity 78.4702
+1200 train 4.278837 (lr=2.8847e-03) (hash(x)=36667141)
+1210 train 4.354375 (lr=2.8800e-03) (hash(x)=45487179)
+1220 train 4.327986 (lr=2.8751e-03) (hash(x)=43456413)
+1230 train 4.258804 (lr=2.8702e-03) (hash(x)=42473632)
+1240 train 4.341355 (lr=2.8652e-03) (hash(x)=42966315)
+1250 train 4.185344 (lr=2.8601e-03) (hash(x)=42930524)
+1260 train 4.330383 (lr=2.8550e-03) (hash(x)=43467676)
+1270 train 4.257578 (lr=2.8497e-03) (hash(x)=40872580)
+1280 train 4.329517 (lr=2.8443e-03) (hash(x)=40619157)
+1290 train 4.345195 (lr=2.8389e-03) (hash(x)=44406165)
+1300 val loss 4.3248
+1300 val perplexity 75.5501
+1300 train 4.308760 (lr=2.8333e-03) (hash(x)=40808029)
+1310 train 4.474747 (lr=2.8277e-03) (hash(x)=46125736)
+1320 train 4.249431 (lr=2.8220e-03) (hash(x)=41698487)
+1330 train 4.280139 (lr=2.8162e-03) (hash(x)=39302878)
+1340 train 4.392772 (lr=2.8103e-03) (hash(x)=41826369)
+1350 train 4.212983 (lr=2.8044e-03) (hash(x)=37061413)
+1360 train 4.337604 (lr=2.7983e-03) (hash(x)=40770974)
+1370 train 4.249659 (lr=2.7922e-03) (hash(x)=38738718)
+1380 train 4.246631 (lr=2.7860e-03) (hash(x)=34961558)
+1390 train 4.296053 (lr=2.7797e-03) (hash(x)=38559342)
+1400 val loss 4.3069
+1400 val perplexity 74.2111
+1400 train 4.249496 (lr=2.7733e-03) (hash(x)=38840020)
+1410 train 4.404141 (lr=2.7668e-03) (hash(x)=44448100)
+1420 train 4.456838 (lr=2.7603e-03) (hash(x)=39199838)
+1430 train 4.253977 (lr=2.7536e-03) (hash(x)=41669873)
+1440 train 4.384237 (lr=2.7469e-03) (hash(x)=53253836)
+1450 train 4.307827 (lr=2.7401e-03) (hash(x)=41320631)
+1460 train 4.223444 (lr=2.7333e-03) (hash(x)=48462803)
+1470 train 4.264635 (lr=2.7263e-03) (hash(x)=42356864)
+1480 train 4.316753 (lr=2.7193e-03) (hash(x)=41118068)
+1490 train 4.232095 (lr=2.7121e-03) (hash(x)=37242854)
+1500 val loss 4.2585
+1500 val perplexity 70.7028
+1500 train 4.271239 (lr=2.7050e-03) (hash(x)=40571031)
+1510 train 4.224005 (lr=2.6977e-03) (hash(x)=35384608)
+1520 train 4.261134 (lr=2.6903e-03) (hash(x)=35341056)
+1530 train 4.147518 (lr=2.6829e-03) (hash(x)=36471779)
+1540 train 4.178072 (lr=2.6754e-03) (hash(x)=40356153)
+1550 train 4.222429 (lr=2.6678e-03) (hash(x)=40133377)
+1560 train 4.149037 (lr=2.6602e-03) (hash(x)=33722160)
+1570 train 4.306149 (lr=2.6525e-03) (hash(x)=39555295)
+1580 train 4.206088 (lr=2.6447e-03) (hash(x)=35950666)
+1590 train 4.360620 (lr=2.6368e-03) (hash(x)=36528535)
+1600 val loss 4.2378
+1600 val perplexity 69.2551
+1600 train 4.245578 (lr=2.6289e-03) (hash(x)=42613017)
+1610 train 4.390632 (lr=2.6208e-03) (hash(x)=42845965)
+1620 train 4.270403 (lr=2.6128e-03) (hash(x)=36137074)
+1630 train 4.234583 (lr=2.6046e-03) (hash(x)=37560569)
+1640 train 4.336964 (lr=2.5964e-03) (hash(x)=43680614)
+1650 train 4.189793 (lr=2.5881e-03) (hash(x)=37281218)
+1660 train 4.212651 (lr=2.5797e-03) (hash(x)=36165222)
+1670 train 4.173405 (lr=2.5713e-03) (hash(x)=50222513)
+1680 train 4.059635 (lr=2.5628e-03) (hash(x)=35239173)
+1690 train 4.114145 (lr=2.5542e-03) (hash(x)=40287349)
+1700 val loss 4.2444
+1700 val perplexity 69.7117
+1700 train 4.296540 (lr=2.5455e-03) (hash(x)=37444780)
+1710 train 4.283964 (lr=2.5368e-03) (hash(x)=38962723)
+1720 train 4.285658 (lr=2.5281e-03) (hash(x)=33882305)
+1730 train 4.265724 (lr=2.5192e-03) (hash(x)=36501540)
+1740 train 4.257672 (lr=2.5103e-03) (hash(x)=38723266)
+1750 train 4.192972 (lr=2.5014e-03) (hash(x)=38772789)
+1760 train 4.271956 (lr=2.4924e-03) (hash(x)=37707898)
+1770 train 4.136791 (lr=2.4833e-03) (hash(x)=30529327)
+1780 train 4.219057 (lr=2.4741e-03) (hash(x)=37885464)
+1790 train 4.232553 (lr=2.4649e-03) (hash(x)=39013967)
+1800 val loss 4.2003
+1800 val perplexity 66.7062
+1800 train 4.269434 (lr=2.4556e-03) (hash(x)=34906955)
+1810 train 4.146118 (lr=2.4463e-03) (hash(x)=40790682)
+1820 train 4.017489 (lr=2.4369e-03) (hash(x)=39912223)
+1830 train 4.142346 (lr=2.4275e-03) (hash(x)=39304486)
+1840 train 4.115877 (lr=2.4180e-03) (hash(x)=37247975)
+1850 train 4.111137 (lr=2.4084e-03) (hash(x)=41340614)
+1860 train 4.262663 (lr=2.3988e-03) (hash(x)=39924627)
+1870 train 4.246150 (lr=2.3891e-03) (hash(x)=40235993)
+1880 train 4.131333 (lr=2.3794e-03) (hash(x)=37693476)
+1890 train 4.249671 (lr=2.3696e-03) (hash(x)=40449864)
+1900 val loss 4.1751
+1900 val perplexity 65.0474
+1900 train 4.158910 (lr=2.3598e-03) (hash(x)=37907749)
+1910 train 4.127439 (lr=2.3499e-03) (hash(x)=34547901)
+1920 train 4.159626 (lr=2.3400e-03) (hash(x)=37771788)
+1930 train 4.031431 (lr=2.3300e-03) (hash(x)=36049454)
+1940 train 4.125335 (lr=2.3200e-03) (hash(x)=29564336)
+1950 train 4.119464 (lr=2.3099e-03) (hash(x)=41689281)
+1960 train 4.030754 (lr=2.2998e-03) (hash(x)=41976699)
+1970 train 4.233627 (lr=2.2896e-03) (hash(x)=41538621)
+1980 train 4.111691 (lr=2.2793e-03) (hash(x)=36490902)
+1990 train 4.140900 (lr=2.2691e-03) (hash(x)=42739315)
+2000 val loss 4.1559
+2000 val perplexity 63.8092
+2000 train 4.154397 (lr=2.2588e-03) (hash(x)=38600074)
+2010 train 4.173006 (lr=2.2484e-03) (hash(x)=34555488)
+2020 train 4.189758 (lr=2.2380e-03) (hash(x)=37227345)
+2030 train 4.141366 (lr=2.2275e-03) (hash(x)=38172386)
+2040 train 4.252366 (lr=2.2170e-03) (hash(x)=39340786)
+2050 train 4.096076 (lr=2.2065e-03) (hash(x)=37237298)
+2060 train 4.336809 (lr=2.1959e-03) (hash(x)=35894782)
+2070 train 4.156165 (lr=2.1853e-03) (hash(x)=44161861)
+2080 train 4.170679 (lr=2.1746e-03) (hash(x)=38062558)
+2090 train 4.026491 (lr=2.1639e-03) (hash(x)=38608301)
+2100 val loss 4.1336
+2100 val perplexity 62.4022
+2100 train 4.050180 (lr=2.1532e-03) (hash(x)=47611537)
+2110 train 4.056353 (lr=2.1424e-03) (hash(x)=38076651)
+2120 train 4.061294 (lr=2.1316e-03) (hash(x)=39870893)
+2130 train 4.214941 (lr=2.1208e-03) (hash(x)=48123321)
+2140 train 4.240106 (lr=2.1099e-03) (hash(x)=44124977)
+2150 train 4.140234 (lr=2.0990e-03) (hash(x)=42181374)
+2160 train 4.167806 (lr=2.0881e-03) (hash(x)=41837878)
+2170 train 4.184142 (lr=2.0771e-03) (hash(x)=42161375)
+2180 train 4.151351 (lr=2.0661e-03) (hash(x)=41792986)
+2190 train 4.124248 (lr=2.0550e-03) (hash(x)=41872416)
+2200 val loss 4.1133
+2200 val perplexity 61.1486
+2200 train 4.101959 (lr=2.0440e-03) (hash(x)=37395985)
+2210 train 4.149935 (lr=2.0329e-03) (hash(x)=40616096)
+2220 train 4.050853 (lr=2.0217e-03) (hash(x)=40118423)
+2230 train 3.964693 (lr=2.0106e-03) (hash(x)=39641586)
+2240 train 4.058895 (lr=1.9994e-03) (hash(x)=34720116)
+2250 train 3.997940 (lr=1.9882e-03) (hash(x)=43641508)
+2260 train 4.016213 (lr=1.9770e-03) (hash(x)=31179786)
+2270 train 3.987716 (lr=1.9657e-03) (hash(x)=42721932)
+2280 train 4.313298 (lr=1.9544e-03) (hash(x)=38474505)
+2290 train 4.138318 (lr=1.9431e-03) (hash(x)=41398545)
+2300 val loss 4.0968
+2300 val perplexity 60.1500
+2300 train 4.070576 (lr=1.9318e-03) (hash(x)=44131094)
+2310 train 4.149317 (lr=1.9205e-03) (hash(x)=41667948)
+2320 train 4.225599 (lr=1.9091e-03) (hash(x)=36339232)
+2330 train 4.105408 (lr=1.8977e-03) (hash(x)=41630644)
+2340 train 4.193828 (lr=1.8863e-03) (hash(x)=41278955)
+2350 train 4.065461 (lr=1.8749e-03) (hash(x)=39491976)
+2360 train 4.131691 (lr=1.8635e-03) (hash(x)=41771238)
+2370 train 4.226571 (lr=1.8520e-03) (hash(x)=45374570)
+2380 train 4.104238 (lr=1.8406e-03) (hash(x)=41419267)
+2390 train 4.135674 (lr=1.8291e-03) (hash(x)=38869169)
+2400 val loss 4.0823
+2400 val perplexity 59.2833
+2400 train 4.035968 (lr=1.8176e-03) (hash(x)=38619293)
+2410 train 3.865609 (lr=1.8061e-03) (hash(x)=41600240)
+2420 train 3.989235 (lr=1.7946e-03) (hash(x)=40891045)
+2430 train 3.982191 (lr=1.7830e-03) (hash(x)=41023249)
+2440 train 3.952365 (lr=1.7715e-03) (hash(x)=33813452)
+2450 train 3.906143 (lr=1.7600e-03) (hash(x)=38464119)
+2460 train 3.798860 (lr=1.7484e-03) (hash(x)=40699982)
+2470 train 3.809197 (lr=1.7368e-03) (hash(x)=38254854)
+2480 train 4.096316 (lr=1.7253e-03) (hash(x)=42736069)
+2490 train 4.110186 (lr=1.7137e-03) (hash(x)=41928525)
+2500 val loss 4.0642
+2500 val perplexity 58.2198
+2500 train 4.060219 (lr=1.7021e-03) (hash(x)=45864011)
+2510 train 4.076487 (lr=1.6906e-03) (hash(x)=41380960)
+2520 train 4.071254 (lr=1.6790e-03) (hash(x)=38366545)
+2530 train 4.078263 (lr=1.6674e-03) (hash(x)=41045176)
+2540 train 4.105968 (lr=1.6558e-03) (hash(x)=43435705)
+2550 train 4.058458 (lr=1.6442e-03) (hash(x)=40693090)
+2560 train 3.975257 (lr=1.6326e-03) (hash(x)=40502478)
+2570 train 4.009761 (lr=1.6210e-03) (hash(x)=38610920)
+2580 train 3.932878 (lr=1.6094e-03) (hash(x)=38333499)
+2590 train 4.011162 (lr=1.5979e-03) (hash(x)=41982736)
+2600 val loss 4.0478
+2600 val perplexity 57.2725
+2600 train 3.989073 (lr=1.5863e-03) (hash(x)=37724702)
+2610 train 3.925298 (lr=1.5747e-03) (hash(x)=39564630)
+2620 train 3.873578 (lr=1.5632e-03) (hash(x)=38465803)
+2630 train 3.794012 (lr=1.5516e-03) (hash(x)=36955007)
+2640 train 3.834682 (lr=1.5400e-03) (hash(x)=32994568)
+2650 train 4.010465 (lr=1.5285e-03) (hash(x)=41165765)
+2660 train 4.020971 (lr=1.5170e-03) (hash(x)=42934000)
+2670 train 4.223475 (lr=1.5054e-03) (hash(x)=42726316)
+2680 train 4.103465 (lr=1.4939e-03) (hash(x)=35880318)
+2690 train 3.990810 (lr=1.4824e-03) (hash(x)=41747329)
+2700 val loss 4.0371
+2700 val perplexity 56.6628
+2700 train 4.117911 (lr=1.4709e-03) (hash(x)=40259630)
+2710 train 4.240554 (lr=1.4594e-03) (hash(x)=39895798)
+2720 train 3.971967 (lr=1.4480e-03) (hash(x)=36146683)
+2730 train 3.954703 (lr=1.4365e-03) (hash(x)=36181984)
+2740 train 3.928339 (lr=1.4251e-03) (hash(x)=43700349)
+2750 train 3.994356 (lr=1.4137e-03) (hash(x)=41610597)
+2760 train 3.985327 (lr=1.4023e-03) (hash(x)=31183639)
+2770 train 3.978098 (lr=1.3909e-03) (hash(x)=37722489)
+2780 train 3.921297 (lr=1.3795e-03) (hash(x)=47290688)
+2790 train 3.960868 (lr=1.3682e-03) (hash(x)=41205574)
+2800 val loss 4.0213
+2800 val perplexity 55.7758
+2800 train 3.810033 (lr=1.3569e-03) (hash(x)=40257962)
+2810 train 3.791974 (lr=1.3456e-03) (hash(x)=39529014)
+2820 train 3.809175 (lr=1.3343e-03) (hash(x)=42244749)
+2830 train 3.861666 (lr=1.3230e-03) (hash(x)=32820090)
+2840 train 3.823705 (lr=1.3118e-03) (hash(x)=40315769)
+2850 train 4.058960 (lr=1.3006e-03) (hash(x)=41524462)
+2860 train 4.052763 (lr=1.2894e-03) (hash(x)=38365734)
+2870 train 4.179628 (lr=1.2783e-03) (hash(x)=37682602)
+2880 train 4.035479 (lr=1.2671e-03) (hash(x)=39162991)
+2890 train 3.894207 (lr=1.2560e-03) (hash(x)=33316384)
+2900 val loss 3.9967
+2900 val perplexity 54.4161
+2900 train 3.946078 (lr=1.2450e-03) (hash(x)=37271132)
+2910 train 4.133524 (lr=1.2339e-03) (hash(x)=35586242)
+2920 train 4.027472 (lr=1.2229e-03) (hash(x)=33320586)
+2930 train 4.037361 (lr=1.2119e-03) (hash(x)=43531361)
+2940 train 3.922794 (lr=1.2010e-03) (hash(x)=37368286)
+2950 train 3.955195 (lr=1.1901e-03) (hash(x)=40363394)
+2960 train 3.927172 (lr=1.1792e-03) (hash(x)=42253792)
+2970 train 3.866165 (lr=1.1684e-03) (hash(x)=38072598)
+2980 train 3.941077 (lr=1.1576e-03) (hash(x)=41470557)
+2990 train 3.867032 (lr=1.1468e-03) (hash(x)=42600033)
+3000 val loss 3.9980
+3000 val perplexity 54.4914
+3000 train 3.819512 (lr=1.1361e-03) (hash(x)=46890983)
+3010 train 3.687428 (lr=1.1254e-03) (hash(x)=37986759)
+3020 train 3.810691 (lr=1.1147e-03) (hash(x)=36270703)
+3030 train 4.076555 (lr=1.1041e-03) (hash(x)=38228599)
+3040 train 3.931058 (lr=1.0935e-03) (hash(x)=44344296)
+3050 train 4.014858 (lr=1.0830e-03) (hash(x)=39962297)
+3060 train 4.003596 (lr=1.0725e-03) (hash(x)=39817394)
+3070 train 3.996688 (lr=1.0620e-03) (hash(x)=41763868)
+3080 train 4.076163 (lr=1.0516e-03) (hash(x)=42343051)
+3090 train 4.070370 (lr=1.0412e-03) (hash(x)=39204893)
+3100 val loss 3.9772
+3100 val perplexity 53.3671
+3100 train 3.976465 (lr=1.0309e-03) (hash(x)=39353599)
+3110 train 4.044365 (lr=1.0207e-03) (hash(x)=36118073)
+3120 train 3.746538 (lr=1.0104e-03) (hash(x)=44070271)
+3130 train 3.773758 (lr=1.0002e-03) (hash(x)=43368151)
+3140 train 3.848818 (lr=9.9011e-04) (hash(x)=40612203)
+3150 train 3.991338 (lr=9.8002e-04) (hash(x)=39598447)
+3160 train 4.034787 (lr=9.6999e-04) (hash(x)=43897682)
+3170 train 4.019437 (lr=9.6000e-04) (hash(x)=39029243)
+3180 train 4.027864 (lr=9.5007e-04) (hash(x)=38586684)
+3190 train 3.981925 (lr=9.4019e-04) (hash(x)=39299115)
+3200 val loss 3.9579
+3200 val perplexity 52.3484
+3200 train 4.006319 (lr=9.3036e-04) (hash(x)=41064949)
+3210 train 3.993791 (lr=9.2058e-04) (hash(x)=41752822)
+3220 train 3.999174 (lr=9.1085e-04) (hash(x)=37484805)
+3230 train 3.965427 (lr=9.0118e-04) (hash(x)=40514919)
+3240 train 4.017501 (lr=8.9157e-04) (hash(x)=42440343)
+3250 train 3.921484 (lr=8.8201e-04) (hash(x)=32464136)
+3260 train 3.941777 (lr=8.7251e-04) (hash(x)=47270558)
+3270 train 3.877330 (lr=8.6307e-04) (hash(x)=39307546)
+3280 train 3.914576 (lr=8.5368e-04) (hash(x)=40410299)
+3290 train 3.908976 (lr=8.4435e-04) (hash(x)=40541497)
+3300 val loss 3.9534
+3300 val perplexity 52.1106
+3300 train 3.812737 (lr=8.3508e-04) (hash(x)=38089139)
+3310 train 3.845575 (lr=8.2588e-04) (hash(x)=42006576)
+3320 train 3.935588 (lr=8.1673e-04) (hash(x)=40422556)
+3330 train 3.729753 (lr=8.0764e-04) (hash(x)=39021643)
+3340 train 3.995386 (lr=7.9862e-04) (hash(x)=40434305)
+3350 train 4.014440 (lr=7.8966e-04) (hash(x)=40833559)
+3360 train 4.015814 (lr=7.8076e-04) (hash(x)=40045231)
+3370 train 3.978151 (lr=7.7192e-04) (hash(x)=34668317)
+3380 train 3.977051 (lr=7.6315e-04) (hash(x)=42738568)
+3390 train 4.075114 (lr=7.5445e-04) (hash(x)=38980166)
+3400 val loss 3.9321
+3400 val perplexity 51.0156
+3400 train 3.982142 (lr=7.4581e-04) (hash(x)=43878176)
+3410 train 4.026713 (lr=7.3724e-04) (hash(x)=38201991)
+3420 train 4.000719 (lr=7.2874e-04) (hash(x)=34905889)
+3430 train 4.004856 (lr=7.2030e-04) (hash(x)=41680405)
+3440 train 3.964763 (lr=7.1193e-04) (hash(x)=39948184)
+3450 train 4.013325 (lr=7.0363e-04) (hash(x)=33623728)
+3460 train 3.882361 (lr=6.9541e-04) (hash(x)=41027647)
+3470 train 3.914577 (lr=6.8725e-04) (hash(x)=42779482)
+3480 train 3.936655 (lr=6.7916e-04) (hash(x)=41579288)
+3490 train 3.787006 (lr=6.7114e-04) (hash(x)=37174846)
+3500 val loss 3.9291
+3500 val perplexity 50.8637
+3500 train 3.666730 (lr=6.6320e-04) (hash(x)=37082090)
+3510 train 3.868412 (lr=6.5533e-04) (hash(x)=43131340)
+3520 train 3.781068 (lr=6.4753e-04) (hash(x)=42457250)
+3530 train 4.007407 (lr=6.3981e-04) (hash(x)=39504492)
+3540 train 3.956300 (lr=6.3216e-04) (hash(x)=36225997)
+3550 train 4.083768 (lr=6.2458e-04) (hash(x)=40912087)
+3560 train 3.919294 (lr=6.1708e-04) (hash(x)=41545925)
+3570 train 4.009060 (lr=6.0966e-04) (hash(x)=32553193)
+3580 train 3.855056 (lr=6.0231e-04) (hash(x)=38799856)
+3590 train 3.902007 (lr=5.9504e-04) (hash(x)=40772075)
+3600 val loss 3.9103
+3600 val perplexity 49.9139
+3600 train 3.941343 (lr=5.8785e-04) (hash(x)=39299903)
+3610 train 3.984620 (lr=5.8074e-04) (hash(x)=39503247)
+3620 train 3.884426 (lr=5.7370e-04) (hash(x)=38730668)
+3630 train 3.903308 (lr=5.6675e-04) (hash(x)=40503276)
+3640 train 3.913339 (lr=5.5987e-04) (hash(x)=36504573)
+3650 train 3.936536 (lr=5.5308e-04) (hash(x)=44903075)
+3660 train 3.839731 (lr=5.4636e-04) (hash(x)=33800148)
+3670 train 3.752671 (lr=5.3973e-04) (hash(x)=39404541)
+3680 train 3.696729 (lr=5.3318e-04) (hash(x)=44777432)
+3690 train 3.793703 (lr=5.2671e-04) (hash(x)=45649913)
+3700 val loss 3.9128
+3700 val perplexity 50.0381
+3700 train 3.912190 (lr=5.2033e-04) (hash(x)=33812537)
+3710 train 4.065012 (lr=5.1402e-04) (hash(x)=36161222)
+3720 train 3.945955 (lr=5.0780e-04) (hash(x)=43690616)
+3730 train 3.949650 (lr=5.0167e-04) (hash(x)=35462313)
+3740 train 4.245599 (lr=4.9562e-04) (hash(x)=44674254)
+3750 train 3.925692 (lr=4.8965e-04) (hash(x)=44276297)
+3760 train 3.973274 (lr=4.8377e-04) (hash(x)=38492178)
+3770 train 3.980526 (lr=4.7798e-04) (hash(x)=44347531)
+3780 train 3.888068 (lr=4.7227e-04) (hash(x)=33792059)
+3790 train 3.986055 (lr=4.6665e-04) (hash(x)=38549318)
+3800 val loss 3.8952
+3800 val perplexity 49.1641
+3800 train 3.866667 (lr=4.6112e-04) (hash(x)=41437448)
+3810 train 3.767452 (lr=4.5567e-04) (hash(x)=36656768)
+3820 train 3.930921 (lr=4.5031e-04) (hash(x)=46387893)
+3830 train 3.928599 (lr=4.4504e-04) (hash(x)=44312129)
+3840 train 3.848206 (lr=4.3986e-04) (hash(x)=40341896)
+3850 train 3.900620 (lr=4.3477e-04) (hash(x)=43589160)
+3860 train 3.883543 (lr=4.2977e-04) (hash(x)=40388945)
+3870 train 3.781083 (lr=4.2486e-04) (hash(x)=33336679)
+3880 train 3.875562 (lr=4.2004e-04) (hash(x)=37643311)
+3890 train 3.899103 (lr=4.1530e-04) (hash(x)=38272927)
+3900 val loss 3.8884
+3900 val perplexity 48.8312
+3900 train 3.843403 (lr=4.1066e-04) (hash(x)=41925748)
+3910 train 3.933294 (lr=4.0611e-04) (hash(x)=38377253)
+3920 train 3.937147 (lr=4.0166e-04) (hash(x)=41616611)
+3930 train 3.862437 (lr=3.9729e-04) (hash(x)=32301827)
+3940 train 3.918568 (lr=3.9302e-04) (hash(x)=47697363)
+3950 train 3.869142 (lr=3.8884e-04) (hash(x)=37867767)
+3960 train 3.869044 (lr=3.8475e-04) (hash(x)=37187295)
+3970 train 3.930926 (lr=3.8076e-04) (hash(x)=41952752)
+3980 train 3.895832 (lr=3.7685e-04) (hash(x)=38358660)
+3990 train 3.762591 (lr=3.7305e-04) (hash(x)=40207878)
+4000 val loss 3.8828
+4000 val perplexity 48.5594
+4000 train 3.854169 (lr=3.6933e-04) (hash(x)=39134015)
+4010 train 3.922479 (lr=3.6572e-04) (hash(x)=38313135)
+4020 train 3.850610 (lr=3.6219e-04) (hash(x)=40710513)
+4030 train 3.977275 (lr=3.5876e-04) (hash(x)=35530165)
+4040 train 3.903843 (lr=3.5543e-04) (hash(x)=39494066)
+4050 train 3.894374 (lr=3.5219e-04) (hash(x)=40861777)
+4060 train 3.870666 (lr=3.4905e-04) (hash(x)=36151939)
+4070 train 3.838002 (lr=3.4600e-04) (hash(x)=39008241)
+4080 train 3.923163 (lr=3.4305e-04) (hash(x)=40395746)
+4090 train 3.960686 (lr=3.4019e-04) (hash(x)=38855480)
+4100 val loss 3.8702
+4100 val perplexity 47.9507
+4100 train 3.892154 (lr=3.3744e-04) (hash(x)=45013254)
+4110 train 3.902840 (lr=3.3477e-04) (hash(x)=39934818)
+4120 train 3.908931 (lr=3.3221e-04) (hash(x)=33417705)
+4130 train 3.844224 (lr=3.2974e-04) (hash(x)=41267978)
+4140 train 3.841904 (lr=3.2737e-04) (hash(x)=38961609)
+4150 train 3.838660 (lr=3.2510e-04) (hash(x)=42173878)
+4160 train 3.804020 (lr=3.2292e-04) (hash(x)=36867405)
+4170 train 3.786273 (lr=3.2085e-04) (hash(x)=38991954)
+4180 train 3.842915 (lr=3.1887e-04) (hash(x)=39032592)
+4190 train 3.820993 (lr=3.1699e-04) (hash(x)=61329402)
+4200 val loss 3.8693
+4200 val perplexity 47.9103
+4200 train 3.836763 (lr=3.1520e-04) (hash(x)=40455200)
+4210 train 3.864570 (lr=3.1352e-04) (hash(x)=40913417)
+4220 train 3.893960 (lr=3.1193e-04) (hash(x)=38809704)
+4230 train 3.879956 (lr=3.1044e-04) (hash(x)=39560204)
+4240 train 3.882278 (lr=3.0905e-04) (hash(x)=41496549)
+4250 train 3.919138 (lr=3.0776e-04) (hash(x)=33994768)
+4260 train 3.905617 (lr=3.0657e-04) (hash(x)=39658687)
+4270 train 3.783062 (lr=3.0548e-04) (hash(x)=42583064)
+4280 train 3.848428 (lr=3.0449e-04) (hash(x)=40635811)
+4290 train 3.748234 (lr=3.0359e-04) (hash(x)=35313841)
+4300 val loss 3.8669
+4300 val perplexity 47.7929
+4300 train 3.927847 (lr=3.0280e-04) (hash(x)=39332245)
+4310 train 3.829502 (lr=3.0210e-04) (hash(x)=43732999)
+4320 train 3.832353 (lr=3.0150e-04) (hash(x)=40831835)
+4330 train 3.969457 (lr=3.0101e-04) (hash(x)=41813368)
+4340 train 3.870416 (lr=3.0061e-04) (hash(x)=36952586)
+4350 train 3.895773 (lr=3.0031e-04) (hash(x)=42470745)
+4360 train 3.817516 (lr=3.0011e-04) (hash(x)=39581495)
+4370 train 4.013284 (lr=3.0001e-04) (hash(x)=35914538)
+4374 val loss 3.8591
+4374 val perplexity 47.4219

att_conv_playground/i_small_heads_wd_2_latent_masks_seed_1339/model_04374.pt

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:12d88164dc9cbf758bd394678575652bb7bb723498418a3b57619e85588f09e1
-size 97746890
+oid sha256:c0e0326138ae07abb8d2f0127a2280177e5b63c2343bbd1eee618a840ac5ff34
+size 97469578

att_conv_playground/i_small_heads_wd_2_latent_masks_seed_1339/optimizer_04374.pt

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:8fd1657909922b0fd6b6da41d834ddfd349e7eb7223cbb7cf3e287166f859f3e
-size 189216718
+oid sha256:bd9c57fb10cd826b2addb1095c346a9a777b6bd50acc8cef5892deff46ff0335
+size 188662222