sew-small-100k-timit

This model is a fine-tuned version of asapp/sew-small-100k on the TIMIT_ASR - NA dataset. It achieves the following results on the evaluation set:

• Loss: 0.4926
• Wer: 0.2988

Model description

More information needed

Intended uses & limitations

More information needed

Training and evaluation data

More information needed

Training procedure

Training hyperparameters

The following hyperparameters were used during training:

• learning_rate: 0.0001
• train_batch_size: 32
• eval_batch_size: 1
• seed: 42
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• lr_scheduler_warmup_steps: 1000
• num_epochs: 20.0
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
3.071	0.69	100	3.0262	1.0
2.9304	1.38	200	2.9297	1.0
2.8823	2.07	300	2.8367	1.0
1.5668	2.76	400	1.2310	0.8807
0.7422	3.45	500	0.7080	0.5957
0.4121	4.14	600	0.5829	0.5073
0.3981	4.83	700	0.5153	0.4461
0.5038	5.52	800	0.4908	0.4151
0.2899	6.21	900	0.5122	0.4111
0.2198	6.9	1000	0.4908	0.3803
0.2129	7.59	1100	0.4668	0.3789
0.3007	8.28	1200	0.4788	0.3562
0.2264	8.97	1300	0.5113	0.3635
0.1536	9.66	1400	0.4950	0.3441
0.1206	10.34	1500	0.5062	0.3421
0.2021	11.03	1600	0.4900	0.3283
0.1458	11.72	1700	0.5019	0.3307
0.1151	12.41	1800	0.4989	0.3270
0.0985	13.1	1900	0.4925	0.3173
0.1412	13.79	2000	0.4868	0.3125
0.1579	14.48	2100	0.4983	0.3147
0.1043	15.17	2200	0.4914	0.3091
0.0773	15.86	2300	0.4858	0.3102
0.1327	16.55	2400	0.5084	0.3064
0.1281	17.24	2500	0.5017	0.3025
0.0845	17.93	2600	0.5001	0.3012
0.0717	18.62	2700	0.4894	0.3004
0.0835	19.31	2800	0.4963	0.2998
0.1181	20.0	2900	0.4926	0.2988

Framework versions

• Transformers 4.12.0.dev0
• Pytorch 1.8.1
• Datasets 1.14.1.dev0
• Tokenizers 0.10.3