wav2vec2-E50_speed_pause3

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: 1.1482
• Cer: 22.9260

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
30.2016	0.1289	200	4.9200	100.0
4.8946	0.2579	400	4.6516	100.0
4.7791	0.3868	600	4.7626	98.3020
4.6864	0.5158	800	4.5883	97.9377
4.6296	0.6447	1000	4.6020	97.9318
4.6182	0.7737	1200	4.5626	97.7850
4.5557	0.9026	1400	4.4614	97.2033
4.4045	1.0316	1600	4.3001	95.7051
3.8003	1.1605	1800	3.5020	62.4089
2.9909	1.2895	2000	2.6455	49.1481
2.5693	1.4184	2200	2.4584	44.4418
2.2791	1.5474	2400	2.1281	39.1539
2.0778	1.6763	2600	1.8969	35.5875
1.8906	1.8053	2800	1.7522	32.5617
1.7518	1.9342	3000	1.6432	31.2397
1.6004	2.0632	3200	1.5258	29.3890
1.4787	2.1921	3400	1.4629	28.5664
1.3618	2.3211	3600	1.3416	25.9577
1.2967	2.4500	3800	1.3214	26.0752
1.2639	2.5790	4000	1.2388	24.6886
1.2042	2.7079	4200	1.2169	24.2656
1.1546	2.8369	4400	1.1823	23.6486
1.1147	2.9658	4600	1.1482	22.9260

Framework versions

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