learning_rate: 0.0003
train_batch_size: 16
eval_batch_size: 8
seed: 42
gradient_accumulation_steps: 2
total_train_batch_size: 32
optimizer: Use OptimizerNames.ADAMW_TORCH with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
lr_scheduler_type: linear
lr_scheduler_warmup_steps: 100
num_epochs: 50
mixed_precision_training: Native AMP

Training Loss	Epoch	Step	Validation Loss	Cer
11.1684	1.1117	100	3.7742	0.9927
4.1354	2.2235	200	3.6767	0.9925
3.891	3.3352	300	3.6737	0.9927
3.8852	4.4469	400	3.7518	0.9929
3.8742	5.5587	500	3.7371	0.9926
3.873	6.6704	600	3.6927	0.9927
3.8641	7.7821	700	3.6704	0.9928
3.7922	8.8939	800	3.7444	0.9929
3.7484	10.0	900	3.6704	0.9928
3.6984	11.1117	1000	3.6351	0.9928
3.6613	12.2235	1100	3.6677	0.9928
3.6179	13.3352	1200	3.5538	0.9929
3.5718	14.4469	1300	3.5695	0.9926
3.5143	15.5587	1400	3.5178	0.9929
3.4629	16.6704	1500	3.5162	0.9928
3.3039	17.7821	1600	3.3854	0.9923
3.177	18.8939	1700	3.5062	0.9184
2.8344	20.0	1800	3.1924	0.8979
2.5845	21.1117	1900	3.0267	0.8324
2.3069	22.2235	2000	3.1840	0.7795
2.1629	23.3352	2100	3.0168	0.7724
2.024	24.4469	2200	3.1201	0.7774
1.8387	25.5587	2300	3.0534	0.7482
1.7332	26.6704	2400	3.0840	0.7492
1.6071	27.7821	2500	3.2198	0.7631
1.5211	28.8939	2600	3.2946	0.7418
1.4258	30.0	2700	3.4587	0.7664
1.3234	31.1117	2800	3.5575	0.7677
1.2455	32.2235	2900	3.6007	0.7652
1.1882	33.3352	3000	3.4905	0.7456
1.0905	34.4469	3100	3.7095	0.7707
1.0538	35.5587	3200	3.7194	0.7608
0.9739	36.6704	3300	3.8311	0.7635
0.9413	37.7821	3400	3.8703	0.7707
0.849	38.8939	3500	4.2106	0.8267
0.8285	40.0	3600	4.1083	0.7801
0.7592	41.1117	3700	4.1399	0.7880
0.7467	42.2235	3800	4.2524	0.8161
0.6883	43.3352	3900	4.3478	0.8065
0.6512	44.4469	4000	4.4906	0.8085
0.6346	45.5587	4100	4.5412	0.8093
0.6123	46.6704	4200	4.7010	0.8331
0.5965	47.7821	4300	4.6217	0.8164
0.573	48.8939	4400	4.7132	0.8317

ctaguchi/ikema-asr-ph

ikema-asr-ph