metadata

license: apache-2.0
tags:
  - automatic-speech-recognition
  - timit_asr
  - generated_from_trainer
datasets:
  - timit_asr
model-index:
  - name: sew-d-small-100k-timit
    results: []

sew-d-small-100k-timit

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

Loss: 1.7561
Wer: 0.7971

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
4.2068	0.69	100	4.0802	1.0
2.9806	1.38	200	2.9792	1.0
2.9781	2.07	300	2.9408	1.0
2.9655	2.76	400	2.9143	1.0
2.8953	3.45	500	2.8774	1.0
2.7712	4.14	600	2.7769	0.9999
2.6662	4.83	700	2.6425	0.9789
2.632	5.52	800	2.5142	1.0318
2.3794	6.21	900	2.4360	1.1475
2.1406	6.9	1000	2.2932	0.9962
2.223	7.59	1100	2.1590	0.9281
2.3607	8.28	1200	2.0553	0.8682
2.1058	8.97	1300	2.0443	0.8902
1.8191	9.66	1400	1.9586	0.8237
1.7013	10.34	1500	1.9586	0.8689
2.2289	11.03	1600	1.9082	0.8611
1.9125	11.72	1700	1.8772	0.8150
1.6424	12.41	1800	1.8671	0.7871
1.6553	13.1	1900	1.8192	0.8121
2.0382	13.79	2000	1.8146	0.8440
1.8785	14.48	2100	1.8094	0.8202
1.6148	15.17	2200	1.8131	0.8234
1.4948	15.86	2300	1.7969	0.8256
1.8844	16.55	2400	1.7790	0.8067
1.8099	17.24	2500	1.7783	0.8073
1.5488	17.93	2600	1.7668	0.7971
1.458	18.62	2700	1.7623	0.7973
1.7656	19.31	2800	1.7574	0.8013
1.7583	20.0	2900	1.7561	0.7971

Framework versions

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