wav2vec2-E50_freq_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: 1.2004
• Cer: 25.1351

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
31.4584	0.1289	200	5.0025	100.0
4.882	0.2579	400	4.6922	100.0
4.7642	0.3868	600	4.7290	100.0
4.7287	0.5158	800	4.6828	100.0
4.6641	0.6447	1000	4.6322	100.0
4.6322	0.7737	1200	4.5289	100.0
4.5965	0.9026	1400	4.5190	98.8132
4.4551	1.0316	1600	4.3994	97.3678
3.9667	1.1605	1800	3.5939	65.2409
3.138	1.2895	2000	2.8840	52.5147
2.6722	1.4184	2200	2.4844	44.6122
2.3367	1.5474	2400	2.1465	39.5300
2.1071	1.6763	2600	1.9978	37.5206
1.9574	1.8053	2800	1.8497	35.2585
1.7583	1.9342	3000	1.6906	33.4195
1.6158	2.0632	3200	1.5764	31.8096
1.4885	2.1921	3400	1.4695	30.5582
1.3927	2.3211	3600	1.4137	29.6710
1.3595	2.4500	3800	1.3518	27.6146
1.2957	2.5790	4000	1.2965	26.9036
1.2472	2.7079	4200	1.2612	25.9929
1.1913	2.8369	4400	1.2208	25.4289
1.1683	2.9658	4600	1.2004	25.1351

Framework versions

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