Upload folder using huggingface_hub

attention_kindself_n_heads2_seed1338/args.json

@@ -1 +1 @@
-{"hellaswag": true, "attention_kind": "selective", "log_dir": "wider_is_better_4/attention_kindself_n_heads2_seed1338", "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": 1338, "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": true, "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_1338", "n_embd": 128}
+{"hellaswag": true, "attention_kind": "selective", "log_dir": "wider_is_better_4/attention_kindself_n_heads2_seed1338", "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": 1338, "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": true, "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_1338", "n_embd": 128}

attention_kindself_n_heads2_seed1338/log2.txt

@@ -1,303 +1,303 @@
 max_steps: 10000
 0 val loss 11.7416
 0 val perplexity 125697.3672
-0 train 11.753315 (lr=7.5000e-07) (hash(x)=150327452)
-100 val loss 9.6016
-100 val perplexity 14788.6875
-100 train 9.584531 (lr=7.5750e-05) (hash(x)=166441190)
-200 val loss 7.9391
-200 val perplexity 2804.8625
-200 train 8.136007 (lr=1.5000e-04) (hash(x)=166780046)
-300 val loss 7.6475
-300 val perplexity 2095.3887
-300 train 7.628962 (lr=1.4997e-04) (hash(x)=159835303)
-400 val loss 7.6130
-400 val perplexity 2024.3168
-400 train 7.384406 (lr=1.4986e-04) (hash(x)=155040610)
-500 val loss 7.5559
-500 val perplexity 1911.9650
-500 train 7.246648 (lr=1.4969e-04) (hash(x)=130190460)
-600 val loss 7.4062
-600 val perplexity 1646.1008
-600 train 7.434457 (lr=1.4945e-04) (hash(x)=155504036)
-700 val loss 7.2972
-700 val perplexity 1476.0964
-700 train 7.166144 (lr=1.4913e-04) (hash(x)=137347213)
-800 val loss 7.2131
-800 val perplexity 1357.0286
-800 train 7.062815 (lr=1.4876e-04) (hash(x)=143823248)
-900 val loss 7.1552
-900 val perplexity 1280.6869
-900 train 7.249660 (lr=1.4831e-04) (hash(x)=156260416)
-1000 val loss 7.1343
-1000 val perplexity 1254.2183
-1000 train 7.099644 (lr=1.4779e-04) (hash(x)=143734685)
-1100 val loss 7.1151
-1100 val perplexity 1230.4236
-1100 train 7.043428 (lr=1.4721e-04) (hash(x)=160013925)
-1200 val loss 7.1059
-1200 val perplexity 1219.0956
-1200 train 6.837766 (lr=1.4656e-04) (hash(x)=150678249)
-1300 val loss 7.0870
-1300 val perplexity 1196.3333
-1300 train 6.858798 (lr=1.4585e-04) (hash(x)=149073315)
-1400 val loss 7.0332
-1400 val perplexity 1133.6700
-1400 train 7.272099 (lr=1.4507e-04) (hash(x)=175802021)
-1500 val loss 6.9964
-1500 val perplexity 1092.7391
-1500 train 7.241509 (lr=1.4422e-04) (hash(x)=171034639)
-1600 val loss 6.9737
-1600 val perplexity 1068.1917
-1600 train 7.144388 (lr=1.4332e-04) (hash(x)=158681215)
-1700 val loss 6.9556
-1700 val perplexity 1048.9771
-1700 train 6.908618 (lr=1.4235e-04) (hash(x)=152116061)
-1800 val loss 6.9507
-1800 val perplexity 1043.9277
-1800 train 6.896075 (lr=1.4131e-04) (hash(x)=146108145)
-1900 val loss 6.9407
-1900 val perplexity 1033.4960
-1900 train 6.769259 (lr=1.4022e-04) (hash(x)=147598108)
-2000 val loss 6.9636
-2000 val perplexity 1057.4624
-2000 train 6.809649 (lr=1.3907e-04) (hash(x)=154996086)
-2100 val loss 6.9618
-2100 val perplexity 1055.5032
-2100 train 6.740406 (lr=1.3786e-04) (hash(x)=153396183)
-2200 val loss 6.9137
-2200 val perplexity 1005.9783
-2200 train 6.904079 (lr=1.3660e-04) (hash(x)=153885445)
-2300 val loss 6.9352
-2300 val perplexity 1027.8553
-2300 train 6.920026 (lr=1.3527e-04) (hash(x)=159666385)
-2400 val loss 6.8895
-2400 val perplexity 981.8969
-2400 train 6.877400 (lr=1.3390e-04) (hash(x)=142353087)
-2500 val loss 6.8939
-2500 val perplexity 986.2402
-2500 train 6.895769 (lr=1.3247e-04) (hash(x)=146491718)
-2600 val loss 6.8935
-2600 val perplexity 985.8123
-2600 train 6.819046 (lr=1.3099e-04) (hash(x)=150750353)
-2700 val loss 6.8923
-2700 val perplexity 984.6749
-2700 train 6.634944 (lr=1.2946e-04) (hash(x)=129849193)
-2800 val loss 6.8994
-2800 val perplexity 991.6816
-2800 train 6.771274 (lr=1.2788e-04) (hash(x)=152767913)
-2900 val loss 6.9001
-2900 val perplexity 992.3586
-2900 train 6.690370 (lr=1.2626e-04) (hash(x)=146531140)
-3000 val loss 6.8711
-3000 val perplexity 963.9977
-3000 train 6.941024 (lr=1.2459e-04) (hash(x)=151562048)
-3100 val loss 6.8716
-3100 val perplexity 964.5380
-3100 train 6.861748 (lr=1.2287e-04) (hash(x)=146001424)
-3200 val loss 6.8663
-3200 val perplexity 959.3974
-3200 train 6.882690 (lr=1.2112e-04) (hash(x)=166486165)
-3300 val loss 6.8767
-3300 val perplexity 969.3868
-3300 train 6.821402 (lr=1.1932e-04) (hash(x)=150866680)
-3400 val loss 6.9140
-3400 val perplexity 1006.2456
-3400 train 6.838375 (lr=1.1749e-04) (hash(x)=143900419)
-3500 val loss 6.8935
-3500 val perplexity 985.8800
-3500 train 6.744522 (lr=1.1562e-04) (hash(x)=148845794)
-3600 val loss 6.9044
-3600 val perplexity 996.6188
-3600 train 6.752177 (lr=1.1372e-04) (hash(x)=145667796)
-3700 val loss 6.9155
-3700 val perplexity 1007.7678
-3700 train 6.933109 (lr=1.1179e-04) (hash(x)=163563851)
-3800 val loss 6.9196
-3800 val perplexity 1011.9581
-3800 train 6.873115 (lr=1.0982e-04) (hash(x)=147488689)
-3900 val loss 6.8857
-3900 val perplexity 978.1486
-3900 train 6.875676 (lr=1.0783e-04) (hash(x)=148186608)
-4000 val loss 6.8883
-4000 val perplexity 980.6855
-4000 train 6.762918 (lr=1.0581e-04) (hash(x)=142970187)
-4100 val loss 6.8691
-4100 val perplexity 962.1278
-4100 train 6.836645 (lr=1.0377e-04) (hash(x)=141584883)
-4200 val loss 6.8615
-4200 val perplexity 954.7748
-4200 train 6.677073 (lr=1.0171e-04) (hash(x)=145664585)
-4300 val loss 6.8726
-4300 val perplexity 965.4155
-4300 train 6.729110 (lr=9.9622e-05) (hash(x)=143736499)
-4400 val loss 6.8957
-4400 val perplexity 987.9940
-4400 train 6.713492 (lr=9.7520e-05) (hash(x)=151883322)
-4500 val loss 6.8914
-4500 val perplexity 983.7997
-4500 train 6.816208 (lr=9.5403e-05) (hash(x)=153904871)
-4600 val loss 6.8958
-4600 val perplexity 988.0821
-4600 train 6.939135 (lr=9.3273e-05) (hash(x)=154893521)
-4700 val loss 6.8626
-4700 val perplexity 955.8658
-4700 train 6.908752 (lr=9.1132e-05) (hash(x)=152323949)
-4800 val loss 6.8600
-4800 val perplexity 953.3958
-4800 train 6.810313 (lr=8.8982e-05) (hash(x)=154104619)
-4900 val loss 6.8481
-4900 val perplexity 942.1285
-4900 train 6.895549 (lr=8.6825e-05) (hash(x)=146311426)
-5000 val loss 6.8550
-5000 val perplexity 948.5805
-5000 train 6.909916 (lr=8.4663e-05) (hash(x)=156741847)
-5100 val loss 6.8591
-5100 val perplexity 952.5506
-5100 train 6.666034 (lr=8.2500e-05) (hash(x)=142086346)
-5200 val loss 6.8607
-5200 val perplexity 954.0717
-5200 train 6.674594 (lr=8.0337e-05) (hash(x)=150265428)
-5300 val loss 6.8700
-5300 val perplexity 962.9457
-5300 train 6.782301 (lr=7.8175e-05) (hash(x)=151339108)
-5400 val loss 6.8699
-5400 val perplexity 962.8424
-5400 train 6.863763 (lr=7.6018e-05) (hash(x)=154654372)
-5500 val loss 6.8525
-5500 val perplexity 946.2426
-5500 train 6.858989 (lr=7.3868e-05) (hash(x)=150575051)
-5600 val loss 6.8406
-5600 val perplexity 935.0728
-5600 train 6.718468 (lr=7.1727e-05) (hash(x)=140396423)
-5700 val loss 6.8446
-5700 val perplexity 938.7697
-5700 train 6.739853 (lr=6.9597e-05) (hash(x)=144678758)
-5800 val loss 6.8472
-5800 val perplexity 941.2125
-5800 train 6.979609 (lr=6.7480e-05) (hash(x)=151992743)
-5900 val loss 6.8624
-5900 val perplexity 955.6702
-5900 train 6.669429 (lr=6.5378e-05) (hash(x)=144396927)
-6000 val loss 6.8714
-6000 val perplexity 964.2631
-6000 train 6.873394 (lr=6.3294e-05) (hash(x)=165478625)
-6100 val loss 6.8850
-6100 val perplexity 977.4706
-6100 train 6.676266 (lr=6.1230e-05) (hash(x)=147088621)
-6200 val loss 6.8726
-6200 val perplexity 965.4542
-6200 train 6.689113 (lr=5.9188e-05) (hash(x)=140794994)
-6300 val loss 6.8632
-6300 val perplexity 956.4193
-6300 train 6.778849 (lr=5.7169e-05) (hash(x)=134780906)
-6400 val loss 6.8792
-6400 val perplexity 971.8004
-6400 train 6.884351 (lr=5.5177e-05) (hash(x)=149023655)
-6500 val loss 6.8673
-6500 val perplexity 960.3407
-6500 train 6.772391 (lr=5.3213e-05) (hash(x)=147497796)
-6600 val loss 6.8737
-6600 val perplexity 966.5330
-6600 train 6.736279 (lr=5.1279e-05) (hash(x)=152902689)
-6700 val loss 6.8804
-6700 val perplexity 973.0472
-6700 train 6.927027 (lr=4.9377e-05) (hash(x)=153846046)
-6800 val loss 6.8637
-6800 val perplexity 956.9379
-6800 train 6.987075 (lr=4.7509e-05) (hash(x)=158512738)
-6900 val loss 6.8529
-6900 val perplexity 946.5829
-6900 train 7.156582 (lr=4.5676e-05) (hash(x)=156849968)
-7000 val loss 6.8527
-7000 val perplexity 946.4376
-7000 train 6.677052 (lr=4.3882e-05) (hash(x)=142395855)
-7100 val loss 6.8548
-7100 val perplexity 948.3797
-7100 train 6.739220 (lr=4.2128e-05) (hash(x)=147114884)
-7200 val loss 6.8482
-7200 val perplexity 942.1964
-7200 train 6.868762 (lr=4.0414e-05) (hash(x)=156979839)
-7300 val loss 6.8510
-7300 val perplexity 944.8517
-7300 train 6.640210 (lr=3.8745e-05) (hash(x)=145584373)
-7400 val loss 6.8517
-7400 val perplexity 945.4772
-7400 train 6.622959 (lr=3.7120e-05) (hash(x)=141508204)
-7500 val loss 6.8480
-7500 val perplexity 941.9704
-7500 train 6.891849 (lr=3.5541e-05) (hash(x)=148803965)
-7600 val loss 6.8441
-7600 val perplexity 938.2962
-7600 train 6.876089 (lr=3.4011e-05) (hash(x)=151019676)
-7700 val loss 6.8384
-7700 val perplexity 933.0297
-7700 train 6.809849 (lr=3.2531e-05) (hash(x)=143155750)
-7800 val loss 6.8369
-7800 val perplexity 931.5507
-7800 train 6.835241 (lr=3.1102e-05) (hash(x)=152569653)
-7900 val loss 6.8397
-7900 val perplexity 934.1885
-7900 train 6.704440 (lr=2.9726e-05) (hash(x)=143519455)
-8000 val loss 6.8407
-8000 val perplexity 935.1360
-8000 train 7.011894 (lr=2.8405e-05) (hash(x)=161180944)
-8100 val loss 6.8387
-8100 val perplexity 933.2611
-8100 train 6.850706 (lr=2.7138e-05) (hash(x)=154107345)
-8200 val loss 6.8373
-8200 val perplexity 931.9835
-8200 train 6.840284 (lr=2.5929e-05) (hash(x)=152486517)
-8300 val loss 6.8284
-8300 val perplexity 923.7352
-8300 train 6.881617 (lr=2.4778e-05) (hash(x)=156167749)
-8400 val loss 6.8222
-8400 val perplexity 918.0245
-8400 train 6.878985 (lr=2.3686e-05) (hash(x)=149155006)
-8500 val loss 6.8197
-8500 val perplexity 915.7223
-8500 train 6.796363 (lr=2.2655e-05) (hash(x)=147844390)
-8600 val loss 6.8211
-8600 val perplexity 916.9929
-8600 train 7.058694 (lr=2.1685e-05) (hash(x)=165753320)
-8700 val loss 6.8217
-8700 val perplexity 917.5388
-8700 train 6.619232 (lr=2.0777e-05) (hash(x)=146079979)
-8800 val loss 6.8215
-8800 val perplexity 917.3826
-8800 train 7.030809 (lr=1.9933e-05) (hash(x)=172259509)
-8900 val loss 6.8250
-8900 val perplexity 920.5526
-8900 train 6.548982 (lr=1.9153e-05) (hash(x)=145148314)
-9000 val loss 6.8254
-9000 val perplexity 920.9134
-9000 train 6.753653 (lr=1.8439e-05) (hash(x)=144250633)
-9100 val loss 6.8199
-9100 val perplexity 915.8677
-9100 train 6.908060 (lr=1.7790e-05) (hash(x)=157219797)
-9200 val loss 6.8155
-9200 val perplexity 911.9052
-9200 train 6.769673 (lr=1.7208e-05) (hash(x)=142743778)
-9300 val loss 6.8131
-9300 val perplexity 909.6530
-9300 train 6.692047 (lr=1.6692e-05) (hash(x)=139669771)
-9400 val loss 6.8112
-9400 val perplexity 907.9672
-9400 train 6.762713 (lr=1.6245e-05) (hash(x)=145916843)
-9500 val loss 6.8124
-9500 val perplexity 909.0519
-9500 train 6.740504 (lr=1.5865e-05) (hash(x)=150196125)
-9600 val loss 6.8116
-9600 val perplexity 908.3495
-9600 train 7.057937 (lr=1.5554e-05) (hash(x)=160041419)
-9700 val loss 6.8098
-9700 val perplexity 906.6606
-9700 train 6.630347 (lr=1.5312e-05) (hash(x)=139931627)
-9800 val loss 6.8081
-9800 val perplexity 905.1487
-9800 train 6.762443 (lr=1.5139e-05) (hash(x)=150370792)
-9900 val loss 6.8101
-9900 val perplexity 906.9745
-9900 train 6.705820 (lr=1.5035e-05) (hash(x)=153014886)
-9999 val loss 6.8126
-9999 val perplexity 909.2761
+0 train 11.753315 (lr=2.5000e-07) (hash(x)=150327452)
+100 val loss 10.3420
+100 val perplexity 31008.8418
+100 train 10.328853 (lr=2.5250e-05) (hash(x)=166441190)
+200 val loss 9.3460
+200 val perplexity 11452.9824
+200 train 9.453581 (lr=5.0000e-05) (hash(x)=166780046)
+300 val loss 8.4134
+300 val perplexity 4507.2549
+300 train 8.403763 (lr=4.9988e-05) (hash(x)=159835303)
+400 val loss 7.9846
+400 val perplexity 2935.4409
+400 train 7.832768 (lr=4.9954e-05) (hash(x)=155040610)
+500 val loss 7.7774
+500 val perplexity 2386.0586
+500 train 7.562558 (lr=4.9896e-05) (hash(x)=130190460)
+600 val loss 7.6696
+600 val perplexity 2142.2061
+600 train 7.707840 (lr=4.9815e-05) (hash(x)=155504036)
+700 val loss 7.6034
+700 val perplexity 2005.0209
+700 train 7.450253 (lr=4.9712e-05) (hash(x)=137347213)
+800 val loss 7.5625
+800 val perplexity 1924.6025
+800 train 7.433071 (lr=4.9585e-05) (hash(x)=143823248)
+900 val loss 7.5305
+900 val perplexity 1864.0834
+900 train 7.644044 (lr=4.9436e-05) (hash(x)=156260416)
+1000 val loss 7.5082
+1000 val perplexity 1822.9496
+1000 train 7.477904 (lr=4.9264e-05) (hash(x)=143734685)
+1100 val loss 7.4909
+1100 val perplexity 1791.7383
+1100 train 7.440104 (lr=4.9070e-05) (hash(x)=160013925)
+1200 val loss 7.4746
+1200 val perplexity 1762.7366
+1200 train 7.243001 (lr=4.8854e-05) (hash(x)=150678249)
+1300 val loss 7.4591
+1300 val perplexity 1735.6089
+1300 train 7.257017 (lr=4.8616e-05) (hash(x)=149073315)
+1400 val loss 7.4199
+1400 val perplexity 1668.8872
+1400 train 7.819283 (lr=4.8356e-05) (hash(x)=175802021)
+1500 val loss 7.3796
+1500 val perplexity 1603.0197
+1500 train 7.631110 (lr=4.8074e-05) (hash(x)=171034639)
+1600 val loss 7.3561
+1600 val perplexity 1565.6826
+1600 train 7.582740 (lr=4.7772e-05) (hash(x)=158681215)
+1700 val loss 7.3446
+1700 val perplexity 1547.8124
+1700 train 7.297821 (lr=4.7448e-05) (hash(x)=152116061)
+1800 val loss 7.3153
+1800 val perplexity 1503.1793
+1800 train 7.247635 (lr=4.7105e-05) (hash(x)=146108145)
+1900 val loss 7.3141
+1900 val perplexity 1501.3834
+1900 train 7.144433 (lr=4.6741e-05) (hash(x)=147598108)
+2000 val loss 7.2828
+2000 val perplexity 1455.0540
+2000 train 7.159031 (lr=4.6357e-05) (hash(x)=154996086)
+2100 val loss 7.2758
+2100 val perplexity 1444.8994
+2100 train 7.088547 (lr=4.5954e-05) (hash(x)=153396183)
+2200 val loss 7.2541
+2200 val perplexity 1413.9039
+2200 train 7.251993 (lr=4.5532e-05) (hash(x)=153885445)
+2300 val loss 7.2741
+2300 val perplexity 1442.4404
+2300 train 7.256250 (lr=4.5091e-05) (hash(x)=159666385)
+2400 val loss 7.2112
+2400 val perplexity 1354.4802
+2400 train 7.179773 (lr=4.4633e-05) (hash(x)=142353087)
+2500 val loss 7.2195
+2500 val perplexity 1365.7396
+2500 train 7.236165 (lr=4.4156e-05) (hash(x)=146491718)
+2600 val loss 7.1954
+2600 val perplexity 1333.2642
+2600 train 7.147064 (lr=4.3663e-05) (hash(x)=150750353)
+2700 val loss 7.1977
+2700 val perplexity 1336.3307
+2700 train 6.934301 (lr=4.3153e-05) (hash(x)=129849193)
+2800 val loss 7.1808
+2800 val perplexity 1313.9878
+2800 train 7.054011 (lr=4.2627e-05) (hash(x)=152767913)
+2900 val loss 7.1768
+2900 val perplexity 1308.7028
+2900 train 6.983736 (lr=4.2085e-05) (hash(x)=146531140)
+3000 val loss 7.1609
+3000 val perplexity 1288.1282
+3000 train 7.214781 (lr=4.1529e-05) (hash(x)=151562048)
+3100 val loss 7.1527
+3100 val perplexity 1277.4952
+3100 train 7.144102 (lr=4.0957e-05) (hash(x)=146001424)
+3200 val loss 7.1358
+3200 val perplexity 1256.1335
+3200 train 7.131488 (lr=4.0373e-05) (hash(x)=166486165)
+3300 val loss 7.1202
+3300 val perplexity 1236.6389
+3300 train 7.046004 (lr=3.9775e-05) (hash(x)=150866680)
+3400 val loss 7.1299
+3400 val perplexity 1248.7705
+3400 train 7.053288 (lr=3.9164e-05) (hash(x)=143900419)
+3500 val loss 7.1033
+3500 val perplexity 1215.9182
+3500 train 6.957945 (lr=3.8541e-05) (hash(x)=148845794)
+3600 val loss 7.1090
+3600 val perplexity 1222.8630
+3600 train 6.980226 (lr=3.7907e-05) (hash(x)=145667796)
+3700 val loss 7.1050
+3700 val perplexity 1218.0392
+3700 train 7.114458 (lr=3.7262e-05) (hash(x)=163563851)
+3800 val loss 7.1221
+3800 val perplexity 1239.0602
+3800 train 7.084368 (lr=3.6608e-05) (hash(x)=147488689)
+3900 val loss 7.1000
+3900 val perplexity 1211.9652
+3900 train 7.077944 (lr=3.5944e-05) (hash(x)=148186608)
+4000 val loss 7.1020
+4000 val perplexity 1214.4360
+4000 train 6.946519 (lr=3.5271e-05) (hash(x)=142970187)
+4100 val loss 7.0786
+4100 val perplexity 1186.2617
+4100 train 7.081099 (lr=3.4590e-05) (hash(x)=141584883)
+4200 val loss 7.0843
+4200 val perplexity 1193.0679
+4200 train 6.903859 (lr=3.3902e-05) (hash(x)=145664585)
+4300 val loss 7.0775
+4300 val perplexity 1184.9812
+4300 train 6.921177 (lr=3.3207e-05) (hash(x)=143736499)
+4400 val loss 7.0813
+4400 val perplexity 1189.5664
+4400 train 6.910584 (lr=3.2507e-05) (hash(x)=151883322)
+4500 val loss 7.0931
+4500 val perplexity 1203.6604
+4500 train 7.027113 (lr=3.1801e-05) (hash(x)=153904871)
+4600 val loss 7.0660
+4600 val perplexity 1171.4612
+4600 train 7.100923 (lr=3.1091e-05) (hash(x)=154893521)
+4700 val loss 7.0486
+4700 val perplexity 1151.2043
+4700 train 7.120266 (lr=3.0377e-05) (hash(x)=152323949)
+4800 val loss 7.0417
+4800 val perplexity 1143.3562
+4800 train 6.989632 (lr=2.9661e-05) (hash(x)=154104619)
+4900 val loss 7.0359
+4900 val perplexity 1136.7014
+4900 train 7.077444 (lr=2.8942e-05) (hash(x)=146311426)
+5000 val loss 7.0312
+5000 val perplexity 1131.3464
+5000 train 7.074374 (lr=2.8221e-05) (hash(x)=156741847)
+5100 val loss 7.0339
+5100 val perplexity 1134.4683
+5100 train 6.836224 (lr=2.7500e-05) (hash(x)=142086346)
+5200 val loss 7.0282
+5200 val perplexity 1127.9674
+5200 train 6.841303 (lr=2.6779e-05) (hash(x)=150265428)
+5300 val loss 7.0328
+5300 val perplexity 1133.2178
+5300 train 6.949815 (lr=2.6058e-05) (hash(x)=151339108)
+5400 val loss 7.0233
+5400 val perplexity 1122.4384
+5400 train 7.012791 (lr=2.5339e-05) (hash(x)=154654372)
+5500 val loss 7.0070
+5500 val perplexity 1104.3318
+5500 train 6.992548 (lr=2.4623e-05) (hash(x)=150575051)
+5600 val loss 7.0122
+5600 val perplexity 1110.1151
+5600 train 6.907769 (lr=2.3909e-05) (hash(x)=140396423)
+5700 val loss 6.9910
+5700 val perplexity 1086.8059
+5700 train 6.885193 (lr=2.3199e-05) (hash(x)=144678758)
+5800 val loss 6.9951
+5800 val perplexity 1091.2894
+5800 train 7.114344 (lr=2.2493e-05) (hash(x)=151992743)
+5900 val loss 6.9942
+5900 val perplexity 1090.2992
+5900 train 6.799440 (lr=2.1793e-05) (hash(x)=144396927)
+6000 val loss 6.9799
+6000 val perplexity 1074.8026
+6000 train 6.948314 (lr=2.1098e-05) (hash(x)=165478625)
+6100 val loss 6.9931
+6100 val perplexity 1089.1113
+6100 train 6.789269 (lr=2.0410e-05) (hash(x)=147088621)
+6200 val loss 6.9827
+6200 val perplexity 1077.7745
+6200 train 6.811790 (lr=1.9729e-05) (hash(x)=140794994)
+6300 val loss 6.9733
+6300 val perplexity 1067.7700
+6300 train 6.887860 (lr=1.9056e-05) (hash(x)=134780906)
+6400 val loss 6.9637
+6400 val perplexity 1057.5562
+6400 train 6.974144 (lr=1.8392e-05) (hash(x)=149023655)
+6500 val loss 6.9550
+6500 val perplexity 1048.3354
+6500 train 6.833264 (lr=1.7738e-05) (hash(x)=147497796)
+6600 val loss 6.9554
+6600 val perplexity 1048.7729
+6600 train 6.844741 (lr=1.7093e-05) (hash(x)=152902689)
+6700 val loss 6.9507
+6700 val perplexity 1043.9287
+6700 train 7.012224 (lr=1.6459e-05) (hash(x)=153846046)
+6800 val loss 6.9446
+6800 val perplexity 1037.5162
+6800 train 7.072952 (lr=1.5836e-05) (hash(x)=158512738)
+6900 val loss 6.9373
+6900 val perplexity 1029.9656
+6900 train 7.254258 (lr=1.5225e-05) (hash(x)=156849968)
+7000 val loss 6.9460
+7000 val perplexity 1038.9885
+7000 train 6.769375 (lr=1.4627e-05) (hash(x)=142395855)
+7100 val loss 6.9355
+7100 val perplexity 1028.1099
+7100 train 6.835872 (lr=1.4043e-05) (hash(x)=147114884)
+7200 val loss 6.9281
+7200 val perplexity 1020.5371
+7200 train 6.944231 (lr=1.3471e-05) (hash(x)=156979839)
+7300 val loss 6.9315
+7300 val perplexity 1023.9854
+7300 train 6.715333 (lr=1.2915e-05) (hash(x)=145584373)
+7400 val loss 6.9387
+7400 val perplexity 1031.4105
+7400 train 6.686437 (lr=1.2373e-05) (hash(x)=141508204)
+7500 val loss 6.9232
+7500 val perplexity 1015.5560
+7500 train 6.973086 (lr=1.1847e-05) (hash(x)=148803965)
+7600 val loss 6.9176
+7600 val perplexity 1009.8733
+7600 train 6.958951 (lr=1.1337e-05) (hash(x)=151019676)
+7700 val loss 6.9107
+7700 val perplexity 1002.9551
+7700 train 6.883573 (lr=1.0844e-05) (hash(x)=143155750)
+7800 val loss 6.9094
+7800 val perplexity 1001.6260
+7800 train 6.919430 (lr=1.0367e-05) (hash(x)=152569653)
+7900 val loss 6.9066
+7900 val perplexity 998.8763
+7900 train 6.761757 (lr=9.9088e-06) (hash(x)=143519455)
+8000 val loss 6.9058
+8000 val perplexity 998.0589
+8000 train 7.089988 (lr=9.4682e-06) (hash(x)=161180944)
+8100 val loss 6.9017
+8100 val perplexity 993.9507
+8100 train 6.942193 (lr=9.0461e-06) (hash(x)=154107345)
+8200 val loss 6.9020
+8200 val perplexity 994.2878
+8200 train 6.913141 (lr=8.6430e-06) (hash(x)=152486517)
+8300 val loss 6.8937
+8300 val perplexity 986.0807
+8300 train 6.929050 (lr=8.2593e-06) (hash(x)=156167749)
+8400 val loss 6.8926
+8400 val perplexity 984.9684
+8400 train 6.948983 (lr=7.8953e-06) (hash(x)=149155006)
+8500 val loss 6.8884
+8500 val perplexity 980.8519
+8500 train 6.867157 (lr=7.5515e-06) (hash(x)=147844390)
+8600 val loss 6.8855
+8600 val perplexity 978.0273
+8600 train 7.149722 (lr=7.2282e-06) (hash(x)=165753320)
+8700 val loss 6.8860
+8700 val perplexity 978.4308
+8700 train 6.695945 (lr=6.9257e-06) (hash(x)=146079979)
+8800 val loss 6.8852
+8800 val perplexity 977.6963
+8800 train 7.105085 (lr=6.6444e-06) (hash(x)=172259509)
+8900 val loss 6.8825
+8900 val perplexity 975.0267
+8900 train 6.578733 (lr=6.3845e-06) (hash(x)=145148314)
+9000 val loss 6.8882
+9000 val perplexity 980.6425
+9000 train 6.826063 (lr=6.1462e-06) (hash(x)=144250633)
+9100 val loss 6.8776
+9100 val perplexity 970.2793
+9100 train 6.986150 (lr=5.9300e-06) (hash(x)=157219797)
+9200 val loss 6.8739
+9200 val perplexity 966.7206
+9200 train 6.840438 (lr=5.7359e-06) (hash(x)=142743778)
+9300 val loss 6.8724
+9300 val perplexity 965.2374
+9300 train 6.752581 (lr=5.5641e-06) (hash(x)=139669771)
+9400 val loss 6.8700
+9400 val perplexity 962.9398
+9400 train 6.804702 (lr=5.4149e-06) (hash(x)=145916843)
+9500 val loss 6.8689
+9500 val perplexity 961.9329
+9500 train 6.787594 (lr=5.2884e-06) (hash(x)=150196125)
+9600 val loss 6.8722
+9600 val perplexity 965.0754
+9600 train 7.153524 (lr=5.1847e-06) (hash(x)=160041419)
+9700 val loss 6.8670
+9700 val perplexity 960.0651
+9700 train 6.683146 (lr=5.1040e-06) (hash(x)=139931627)
+9800 val loss 6.8644
+9800 val perplexity 957.5345
+9800 train 6.816484 (lr=5.0462e-06) (hash(x)=150370792)
+9900 val loss 6.8633
+9900 val perplexity 956.5634
+9900 train 6.752453 (lr=5.0116e-06) (hash(x)=153014886)
+9999 val loss 6.8622
+9999 val perplexity 955.4310

attention_kindself_n_heads2_seed1338/model_02500.pt

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:3fade1208a41cb2b175867d950eb365783f29bdd54086b600fe45a04d74a5ebb
+oid sha256:162ddaf18e89e228ba6652b011f18b3fbe6ea4dbd15b1927a464f0b1261d1618
 size 38587970

attention_kindself_n_heads2_seed1338/model_05000.pt

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:8e77e90db3bc2846ba154f67397fac6656d389cc08d1eeb5e114c5da5ce4ccff
+oid sha256:081e193042470f48a4364138916cd15f9e7a8a14ac5cd03770c2edee904f9123
 size 38587970

attention_kindself_n_heads2_seed1338/model_07500.pt

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:0198ef51a7e199c994d41af77067eae6e1ebe6384403210ee7850d41577a9be2
+oid sha256:7e2cf7e7d8da54c074e0e2d65d09569394851c94956d6a02075c9763d6c8e36a
 size 38587970

attention_kindself_n_heads2_seed1338/model_09999.pt

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:29d98bfcc8d40ac56e7ec5e3b63b1a42512138762f80f43f366984036a6cef08
+oid sha256:1492e55ab8c618c621a1011bdaab52c625084b440cd4698cbc66d65a782b10af
 size 38587970

attention_kindself_n_heads2_seed1338/optimizer_02500.pt

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:11c167eec2983197392a41508b01fb54569138099600788788c7ef6c96ce96f2
+oid sha256:b138f1a8c51d246cb28f518fe899ca98d36455537a633967c69559581e6a74e5
 size 70895430

attention_kindself_n_heads2_seed1338/optimizer_05000.pt

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:c0d50a8a8b3f8862c1803d40215628e2118784c190271c19e20111058c62e52f
+oid sha256:bd574b2deede7ffd710dde6911c704c358d5dcb400314372a424fcf3cc9a56fd
 size 70895430

attention_kindself_n_heads2_seed1338/optimizer_07500.pt

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:913fcb44865d48cb68ed2293be1d270c126e4352cdc7b5ade6d837d4e014cde6
+oid sha256:f5ffea7152403ad721e061dfae3afd353afaa0a69ab1a376e22731e6c936c489
 size 70895430

attention_kindself_n_heads2_seed1338/optimizer_09999.pt

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:1b5294eee50fc4a2a3b88d850421218d33cb8007820b50accef37a1ae224a709
+oid sha256:1db75df11392c74683e75465555e72bb02c2978e80a3a78d67f0a44440e1b501
 size 70895430