Model_G_S_D_Wav2Vec2

This model was trained from scratch on the None dataset. It achieves the following results on the evaluation set:

• Loss: 0.0425
• Wer: 0.0310
• Cer: 0.0095

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: 0.0003
• train_batch_size: 16
• eval_batch_size: 8
• seed: 42
• gradient_accumulation_steps: 2
• total_train_batch_size: 32
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• lr_scheduler_warmup_steps: 500
• num_epochs: 30

Training results

Training Loss	Epoch	Step	Validation Loss	Wer	Cer
0.4434	0.85	400	0.0763	0.0984	0.0254
0.1737	1.71	800	0.0639	0.0781	0.0199
0.1293	2.56	1200	0.0522	0.0653	0.0167
0.0965	3.41	1600	0.0471	0.0659	0.0163
0.0874	4.26	2000	0.0464	0.0535	0.0139
0.0679	5.12	2400	0.0395	0.0490	0.0132
0.0618	5.97	2800	0.0424	0.0533	0.0143
0.056	6.82	3200	0.0471	0.0511	0.0132
0.0568	7.68	3600	0.0432	0.0468	0.0123
0.046	8.53	4000	0.0425	0.0472	0.0130
0.0459	9.38	4400	0.0502	0.0499	0.0134
0.0408	10.23	4800	0.0450	0.0488	0.0131
0.0436	11.09	5200	0.0431	0.0420	0.0119
0.0375	11.94	5600	0.0463	0.0484	0.0132
0.0327	12.79	6000	0.0412	0.0424	0.0116
0.0322	13.65	6400	0.0381	0.0382	0.0111
0.0316	14.5	6800	0.0441	0.0460	0.0128
0.0296	15.35	7200	0.0426	0.0415	0.0119
0.0274	16.2	7600	0.0421	0.0383	0.0106
0.0247	17.06	8000	0.0442	0.0391	0.0120
0.0235	17.91	8400	0.0449	0.0409	0.0116
0.0219	18.76	8800	0.0394	0.0353	0.0106
0.0174	19.62	9200	0.0489	0.0393	0.0117
0.0161	20.47	9600	0.0421	0.0347	0.0099
0.0158	21.32	10000	0.0425	0.0349	0.0108
0.0141	22.17	10400	0.0436	0.0397	0.0116
0.0156	23.03	10800	0.0432	0.0375	0.0114
0.0138	23.88	11200	0.0438	0.0364	0.0110
0.0116	24.73	11600	0.0420	0.0368	0.0108
0.0108	25.59	12000	0.0407	0.0341	0.0103
0.0073	26.44	12400	0.0428	0.0336	0.0101
0.0085	27.29	12800	0.0432	0.0328	0.0101
0.0078	28.14	13200	0.0416	0.0318	0.0096
0.0065	29.0	13600	0.0423	0.0310	0.0097
0.0062	29.85	14000	0.0425	0.0310	0.0095

Framework versions

• Transformers 4.31.0
• Pytorch 2.0.1+cu117
• Datasets 1.18.3
• Tokenizers 0.13.3