wav2vec2-E30_ps

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: 1.3344
• Cer: 26.2632

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
32.5841	0.1289	200	5.0826	100.0
5.3998	0.2579	400	4.6908	100.0
4.882	0.3868	600	4.7141	100.0
4.8041	0.5158	800	4.6746	100.0
4.7502	0.6447	1000	4.6224	100.0
4.7187	0.7737	1200	4.6150	100.0
4.6627	0.9026	1400	4.5518	99.8825
4.5966	1.0316	1600	4.5186	98.5488
4.5155	1.1605	1800	4.4687	98.7368
4.2292	1.2895	2000	4.3169	95.4642
3.3494	1.4184	2200	2.9891	55.7638
2.7084	1.5474	2400	2.6772	48.9718
2.3971	1.6763	2600	2.3131	43.8719
2.1476	1.8053	2800	2.0864	39.6240
1.9298	1.9342	3000	1.8900	35.7814
1.7723	2.0632	3200	1.7887	34.0952
1.657	2.1921	3400	1.6924	32.7203
1.4995	2.3211	3600	1.5108	29.1363
1.4486	2.4500	3800	1.4734	29.3420
1.3692	2.5790	4000	1.3831	27.1445
1.3174	2.7079	4200	1.3562	26.6216
1.2546	2.8369	4400	1.3616	26.7979
1.2237	2.9658	4600	1.3344	26.2632

Framework versions

• Transformers 4.46.3
• Pytorch 2.5.1+cu121
• Datasets 3.1.0
• Tokenizers 0.20.3