Update README.md

README.md

@@ -12,12 +12,13 @@ rinna/japanese-hubert-largeにて特徴抽出を行い、日本語により適
 
 ## 直近の学習ログは以下の通りです。
 ```
-    [2024-10-25 21:21:56,105][valid][INFO] - epoch 027 | valid on 'valid' subset | loss 12.907 | count_m_0 1440.82 | count_u_0 1260.37 | loss_m_0 12.875 | loss_u_0 11.519 | loss_mlm 12.875 | loss_features_pen 0.011 | ctr_loss 0.024 | ctr_weighted_loss 0.023 | correct_m_0 0.00296573 | correct_u_0 0.00264994 | ppl 7680.84 | wps 5041 | wpb 1440.8 | bsz 3.7 | num_updates 97000 | best_loss 9.637
-    [2024-10-25 21:21:56,106][fairseq.checkpoint_utils][INFO] - Preparing to save checkpoint for epoch 27 @ 97000 updates
-    [2024-10-25 21:21:56,109][fairseq.trainer][INFO] - Saving checkpoint to /content/contentvec/tmp/checkpoints/checkpoint_27_97000.pt
-    [2024-10-25 21:21:59,508][fairseq.trainer][INFO] - Finished saving checkpoint to /content/contentvec/tmp/checkpoints/checkpoint_27_97000.pt
-    [2024-10-25 21:22:03,077][fairseq.checkpoint_utils][INFO] - Saved checkpoint checkpoints/checkpoint_27_97000.pt (epoch 27 @ 97000 updates, score 12.907) (writing took 6.970946933986852 seconds)
-    epoch 027:  61% 2255/3683 [23:53<13:23,  1.78it/s, loss=2.65, count_m_0=1440.3, count_u_0=1266.4, loss_m_0=2.584, loss_u_0=1.856, loss_mlm=2.584, loss_features_pen=0.011, ctr_loss=0.063, ctr_weighted_loss=0.062, correct_m_0=0.488006, correct_u_0=0.59136, ppl=6.28, wps=2588.6, ups=1.8, wpb=1440.3, bsz=3.6, num_updates=97900, lr=0.000297301, gnorm=2.706, clip=0, loss_scale=0.0312, train_wall=47, gb_free=12.2, wall=0][2024-10-25 21:31:19,576][fairseq_cli.train][INFO] - begin validation on "valid" subset
+[2024-11-04 21:36:53,622][valid][INFO] - epoch 026 | valid on 'valid' subset | loss 13.295 | count_m_0 1430.89 | count_u_0 1265.14 | loss_m_0 13.264 | loss_u_0 11.904 | loss_mlm 13.264 | loss_features_pen 0.011 | ctr_loss 0.025 | ctr_weighted_loss 0.023 | correct_m_0 0.0013102 | correct_u_0 0.00117504 | ppl 10052.6 | wps 5016.4 | wpb 1430.9 | bsz 3.7 | num_updates 93000 | best_loss 9.809
+[2024-11-04 21:36:53,623][fairseq.checkpoint_utils][INFO] - Preparing to save checkpoint for epoch 26 @ 93000 updates
+[2024-11-04 21:36:53,625][fairseq.trainer][INFO] - Saving checkpoint to /content/contentvec/tmp/checkpoints/checkpoint_26_93000.pt
+[2024-11-04 21:36:56,694][fairseq.trainer][INFO] - Finished saving checkpoint to /content/contentvec/tmp/checkpoints/checkpoint_26_93000.pt
+[2024-11-04 21:37:00,612][fairseq.checkpoint_utils][INFO] - Saved checkpoint checkpoints/checkpoint_26_93000.pt (epoch 26 @ 93000 updates, score 13.295) (writing took 6.988974927997333 seconds)
+                                                    epoch 026:  55% 2012/3680 [26:19<18:36,  1.49it/s, loss=2.806, count_m_0=1420.22, count_u_0=1272.58, loss_m_0=2.746, loss_u_0=2.059, loss_mlm=2.746, loss_features_pen=0.01, ctr_loss=0.057, ctr_weighted_loss=0.054, correct_m_0=0.462972, correct_u_0=0.55299, ppl=7, wps=2053, ups=1.45, wpb=1420.2, bsz=3.8, num_updates=93900, lr=0.000297421, gnorm=2.509, clip=0, loss_scale=0.125, train_wall=64, gb_free=13.4, wall=74426][2024-11-04 21:48:39,649][fairseq_cli.train][INFO] - begin validation on "valid" subset
+[2024-11-04 21:48:39,650][fairseq.tasks.fairseq_task][INFO] - can_reuse_epoch_itr = True
 ```
 ## 所感
 ボイスチェンジャーやTTS用途で試しているのでその分野で使用時の所感です。<br>