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_FUSED with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
lr_scheduler_type: linear
lr_scheduler_warmup_ratio: 0.1
num_epochs: 45
mixed_precision_training: Native AMP

Training Loss	Epoch	Step	Validation Loss	Wer	Cer
1.238	3.2326	1600	0.5900	0.6127	0.1549
1.6997	6.4651	3200	1.1790	0.8736	0.2882
6.0333	9.6977	4800	5.9760	1.0	0.9996
9.4521	12.9302	6400	10.7483	1.0	0.9800
17.3421	16.1618	8000	18.2156	1.0273	0.8260
17.2813	19.3943	9600	18.2161	1.0267	0.8254
17.3381	22.6269	11200	18.2158	1.0278	0.8258
17.3299	25.8595	12800	18.2165	1.0266	0.8258
17.2731	29.0910	14400	18.2165	1.0275	0.8259
17.3433	32.3236	16000	18.2161	1.0278	0.8259
17.287	35.5561	17600	18.2160	1.0273	0.8260
17.3217	38.7887	19200	18.2158	1.0275	0.8260
17.2962	42.0202	20800	18.2169	1.0277	0.8260

anvitamanne/wav2vec2-train

wav2vec2-train