wav2vec2-base-timit-demo-colab

This model is a fine-tuned version of facebook/wav2vec2-base on the None dataset. It achieves the following results on the evaluation set:

• Loss: 0.5532
• Wer: 0.3373
• Cer: 0.1112

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: 8
• eval_batch_size: 8
• 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: 30

Training results

Training Loss	Epoch	Step	Validation Loss	Wer	Cer
0.1293	1.0	500	0.3918	0.3677	0.1170
0.133	2.01	1000	0.4392	0.3797	0.1234
0.1473	3.01	1500	0.4959	0.3914	0.1267
0.1373	4.02	2000	0.4781	0.3851	0.1260
0.1259	5.02	2500	0.4473	0.3810	0.1237
0.1123	6.02	3000	0.5314	0.3774	0.1243
0.1086	7.03	3500	0.4231	0.3801	0.1228
0.0956	8.03	4000	0.5203	0.3734	0.1236
0.0839	9.04	4500	0.5310	0.3750	0.1227
0.0778	10.04	5000	0.5279	0.3793	0.1257
0.0772	11.04	5500	0.4969	0.3792	0.1265
0.072	12.05	6000	0.5489	0.3701	0.1239
0.0678	13.05	6500	0.5123	0.3669	0.1207
0.067	14.06	7000	0.4969	0.3663	0.1192
0.061	15.06	7500	0.4742	0.3664	0.1212
0.0575	16.06	8000	0.5304	0.3643	0.1194
0.0574	17.07	8500	0.4936	0.3729	0.1218
0.0474	18.07	9000	0.5363	0.3601	0.1185
0.0447	19.08	9500	0.5347	0.3552	0.1177
0.0372	20.08	10000	0.5372	0.3519	0.1157
0.0325	21.08	10500	0.5455	0.3525	0.1159
0.0309	22.09	11000	0.5193	0.3514	0.1146
0.0314	23.09	11500	0.5402	0.3494	0.1160
0.0272	24.1	12000	0.5309	0.3457	0.1129
0.0238	25.1	12500	0.5490	0.3447	0.1132
0.0217	26.1	13000	0.5702	0.3406	0.1117
0.0225	27.11	13500	0.5575	0.3414	0.1116
0.0189	28.11	14000	0.5572	0.3391	0.1115
0.0179	29.12	14500	0.5532	0.3373	0.1112

Framework versions

• Transformers 4.32.0.dev0
• Pytorch 2.0.1+cu118
• Datasets 1.18.3
• Tokenizers 0.13.3