wav2vec2-Y_

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: 1.5842
• Cer: 30.3584

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
37.2038	0.1289	200	5.0616	100.0
5.0402	0.2579	400	4.7466	100.0
4.87	0.3868	600	4.7230	100.0
4.8539	0.5158	800	4.7158	100.0
4.7772	0.6447	1000	4.6757	100.0
4.7606	0.7737	1200	4.6469	100.0
4.7489	0.9026	1400	4.6367	100.0
4.7074	1.0316	1600	4.6402	100.0
4.6732	1.1605	1800	4.5753	99.2127
4.53	1.2895	2000	4.4595	98.4665
4.3522	1.4184	2200	4.2283	99.0247
3.6439	1.5474	2400	3.2941	57.8907
2.8931	1.6763	2600	2.7877	51.0987
2.4895	1.8053	2800	2.4939	46.3984
2.2386	1.9342	3000	2.3376	42.0094
2.012	2.0632	3200	2.1472	38.8073
1.8082	2.1921	3400	1.9733	37.4207
1.6902	2.3211	3600	1.8465	34.8825
1.5764	2.4500	3800	1.7415	32.4971
1.5547	2.5790	4000	1.6877	31.8801
1.472	2.7079	4200	1.6360	30.9694
1.4054	2.8369	4400	1.6052	30.6757
1.3754	2.9658	4600	1.5842	30.3584

Framework versions

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