wav2vec2-E50_speed

This model is a fine-tuned version of facebook/wav2vec2-xls-r-300m on an unknown dataset. It achieves the following results on the evaluation set:

• Loss: 1.7201
• Cer: 34.5747

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: 50
• num_epochs: 3
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Cer
38.2863	0.1289	200	4.9788	100.0
4.8884	0.2579	400	4.7635	100.0
4.7532	0.3868	600	4.6460	100.0
4.7285	0.5158	800	4.6380	100.0
4.6656	0.6447	1000	4.6877	100.0
4.6484	0.7737	1200	4.6586	100.0
4.6328	0.9026	1400	4.6110	100.0
4.5589	1.0316	1600	4.5007	100.0
4.4938	1.1605	1800	4.4103	98.0557
4.3191	1.2895	2000	4.2620	95.5768
3.9702	1.4184	2200	3.6438	68.0099
3.3814	1.5474	2400	3.1348	60.4323
2.9655	1.6763	2600	2.9093	59.5865
2.7274	1.8053	2800	2.5505	51.1396
2.5117	1.9342	3000	2.2604	46.0644
2.3308	2.0632	3200	2.0918	42.4871
2.1864	2.1921	3400	2.0284	41.0832
2.0692	2.3211	3600	1.9906	40.9774
2.0208	2.4500	3800	1.9112	38.6278
1.9439	2.5790	4000	1.8649	38.3870
1.8928	2.7079	4200	1.7703	35.7789
1.8225	2.8369	4400	1.7312	34.8508
1.8341	2.9658	4600	1.7201	34.5747

Framework versions

• Transformers 4.44.2
• Pytorch 2.4.1+cu121
• Datasets 3.0.1
• Tokenizers 0.19.1