wav2vec2-E30_speed2

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: 1.2995
• Cer: 25.2938

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
41.4516	0.1289	200	5.3911	100.0
4.9936	0.2579	400	4.7064	100.0
4.8068	0.3868	600	4.6491	100.0
4.7989	0.5158	800	4.7160	100.0
4.7287	0.6447	1000	4.5872	100.0
4.7102	0.7737	1200	4.5938	100.0
4.6927	0.9026	1400	4.6049	100.0
4.6106	1.0316	1600	4.5710	100.0
4.5281	1.1605	1800	4.4045	100.0
4.2276	1.2895	2000	3.8233	74.6240
3.2563	1.4184	2200	2.8222	51.8508
2.5657	1.5474	2400	2.3714	42.7850
2.2757	1.6763	2600	2.1794	41.1163
2.0428	1.8053	2800	1.9496	36.1222
1.8705	1.9342	3000	1.8052	33.6310
1.6822	2.0632	3200	1.6552	31.3514
1.574	2.1921	3400	1.5774	30.2115
1.4683	2.3211	3600	1.4999	28.9424
1.4039	2.4500	3800	1.4358	28.1786
1.3323	2.5790	4000	1.3441	26.1868
1.3055	2.7079	4200	1.3460	25.8813
1.2428	2.8369	4400	1.3022	25.5170
1.2121	2.9658	4600	1.2995	25.2938

Framework versions

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