learning_rate: 2e-05
train_batch_size: 4
eval_batch_size: 4
seed: 42
gradient_accumulation_steps: 8
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
No log	1.1364	100	0.7195
0.8244	2.2727	200	0.5949
0.8244	3.4091	300	0.5811
0.6242	4.5455	400	0.5692
0.6242	5.6818	500	0.5618
0.6035	6.8182	600	0.5568
0.6035	7.9545	700	0.5536
0.5895	9.0909	800	0.5506
0.5895	10.2273	900	0.5471
0.5814	11.3636	1000	0.5443
0.5814	12.5	1100	0.5421
0.5744	13.6364	1200	0.5387
0.5744	14.7727	1300	0.5398
0.5672	15.9091	1400	0.5361
0.5672	17.0455	1500	0.5351
0.5655	18.1818	1600	0.5332
0.5655	19.3182	1700	0.5345
0.5594	20.4545	1800	0.5305
0.5594	21.5909	1900	0.5297
0.5548	22.7273	2000	0.5304
0.5548	23.8636	2100	0.5281
0.5538	25.0	2200	0.5285
0.5538	26.1364	2300	0.5295
0.5518	27.2727	2400	0.5281
0.5518	28.4091	2500	0.5278
0.5497	29.5455	2600	0.5257
0.5497	30.6818	2700	0.5270
0.5463	31.8182	2800	0.5272
0.5463	32.9545	2900	0.5261
0.5451	34.0909	3000	0.5268
0.5451	35.2273	3100	0.5225
0.5428	36.3636	3200	0.5254
0.5428	37.5	3300	0.5267
0.5418	38.6364	3400	0.5244
0.5418	39.7727	3500	0.5252
0.5395	40.9091	3600	0.5265
0.5395	42.0455	3700	0.5229
0.5411	43.1818	3800	0.5221
0.5411	44.3182	3900	0.5220
0.538	45.4545	4000	0.5251
0.538	46.5909	4100	0.5256
0.5358	47.7273	4200	0.5256
0.5358	48.8636	4300	0.5242
0.5371	50.0	4400	0.5243

schalor/speecht5_finetuned

speecht5_finetuned