wav2vec2-E50_speed_pause

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: 1.4049
• Cer: 29.7227

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
32.3933	0.1289	200	4.9500	100.0
4.8782	0.2579	400	4.6402	100.0
4.7485	0.3868	600	4.6460	100.0
4.7179	0.5158	800	4.5728	100.0
4.644	0.6447	1000	4.6080	99.0132
4.61	0.7737	1200	4.5600	98.2613
4.5722	0.9026	1400	4.5529	99.4537
4.4489	1.0316	1600	4.5026	98.1144
4.2793	1.1605	1800	4.1438	92.5928
3.6845	1.2895	2000	3.4651	61.3252
3.0089	1.4184	2200	2.6961	50.7049
2.6617	1.5474	2400	2.3715	46.2523
2.4745	1.6763	2600	2.2327	43.4739
2.2853	1.8053	2800	2.0575	41.7704
2.1079	1.9342	3000	1.9056	38.0639
1.9655	2.0632	3200	1.8005	35.8846
1.8115	2.1921	3400	1.6990	35.4088
1.7347	2.3211	3600	1.6111	33.3470
1.6653	2.4500	3800	1.5471	32.6833
1.5837	2.5790	4000	1.5360	31.9608
1.5514	2.7079	4200	1.4449	30.1398
1.4909	2.8369	4400	1.4166	29.7345
1.4908	2.9658	4600	1.4049	29.7227

Framework versions

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