STT_Model_17

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: 0.1172
• Wer: 0.1190

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: 50

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
4.1934	2.1	500	3.7998	0.9999
1.14	4.2	1000	0.4083	0.3740
0.2217	6.3	1500	0.2515	0.2184
0.1276	8.4	2000	0.1623	0.1803
0.0914	10.5	2500	0.1586	0.1672
0.0731	12.61	3000	0.1648	0.1583
0.0572	14.71	3500	0.4059	0.1534
0.054	16.81	4000	0.1694	0.1391
0.043	18.91	4500	0.1390	0.1439
0.035	21.01	5000	0.1210	0.1362
0.0317	23.11	5500	0.1389	0.1285
0.031	25.21	6000	0.1340	0.1316
0.0266	27.31	6500	0.1312	0.1280
0.0209	29.41	7000	0.1484	0.1256
0.0184	31.51	7500	0.1345	0.1289
0.0201	33.61	8000	0.1350	0.1248
0.026	35.71	8500	0.1226	0.1235
0.016	37.82	9000	0.1235	0.1232
0.0115	39.92	9500	0.1223	0.1216
0.013	42.02	10000	0.1314	0.1206
0.0225	44.12	10500	0.1158	0.1211
0.011	46.22	11000	0.1181	0.1203
0.0106	48.32	11500	0.1172	0.1190

Framework versions

• Transformers 4.27.3
• Pytorch 1.13.1+cu116
• Datasets 2.10.1
• Tokenizers 0.13.2