wav2vec2-Y_speed_pause2

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: 1.6467
• Cer: 38.8660

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: Use adamw_torch with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
• 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
35.4246	0.1290	200	5.3966	100.0
5.0597	0.2581	400	4.7304	100.0
4.9257	0.3871	600	4.7995	100.0
4.8261	0.5161	800	4.6200	100.0
4.7176	0.6452	1000	4.5602	98.3079
4.5324	0.7742	1200	4.4388	97.9671
3.7536	0.9032	1400	3.6145	73.7427
2.8505	1.0323	1600	3.0887	59.1128
2.4185	1.1613	1800	2.6231	53.6780
2.1072	1.2903	2000	2.4919	52.2679
1.8961	1.4194	2200	2.3102	50.1469
1.7491	1.5484	2400	2.3485	49.7415
1.5897	1.6774	2600	2.0813	45.6580
1.4517	1.8065	2800	1.9086	44.1128
1.3285	1.9355	3000	1.9665	43.8425
1.2685	2.0645	3200	1.8661	43.1845
1.1435	2.1935	3400	1.8922	43.9600
1.0695	2.3226	3600	1.7025	40.4877
1.0431	2.4516	3800	1.6870	39.6357
0.9992	2.5806	4000	1.7723	41.7450
0.9622	2.7097	4200	1.6614	39.0834
0.9273	2.8387	4400	1.6228	38.4724
0.9176	2.9677	4600	1.6467	38.8660

Framework versions

• Transformers 4.46.2
• Pytorch 2.5.1+cu121
• Datasets 3.1.0
• Tokenizers 0.20.3