wav2vec-large-en

This model is a fine-tuned version of facebook/wav2vec2-large on an unknown dataset. It achieves the following results on the evaluation set:

• Loss: 2.9034
• Wer: 1.0

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: 5e-06
• train_batch_size: 32
• 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
• num_epochs: 30
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
No log	0.3521	50	3.5027	1.0
5.2013	0.7042	100	2.9715	1.0
5.2013	1.0563	150	3.0431	1.0
3.1541	1.4085	200	2.8890	1.0
3.1541	1.7606	250	2.9611	1.0
2.8986	2.1127	300	2.8821	1.0
2.8986	2.4648	350	2.8843	1.0
2.9209	2.8169	400	2.9202	1.0
2.9209	3.1690	450	2.9015	1.0
2.9283	3.5211	500	2.8956	1.0
2.9283	3.8732	550	2.8913	1.0
2.8832	4.2254	600	2.9140	1.0
2.8832	4.5775	650	2.8749	1.0
2.8789	4.9296	700	2.8970	1.0
2.8789	5.2817	750	2.9359	1.0
2.8753	5.6338	800	2.8821	1.0
2.8753	5.9859	850	2.8868	1.0
2.8859	6.3380	900	2.9063	1.0
2.8859	6.6901	950	2.8784	1.0
2.8693	7.0423	1000	2.9513	1.0
2.8693	7.3944	1050	2.8936	1.0
2.8911	7.7465	1100	2.8821	1.0
2.8911	8.0986	1150	2.9227	1.0
2.8625	8.4507	1200	2.8949	1.0
2.8625	8.8028	1250	2.9313	1.0
2.8641	9.1549	1300	2.8978	1.0
2.8641	9.5070	1350	2.9034	1.0

Framework versions

• Transformers 4.46.2
• Pytorch 2.5.1+cu121
• Tokenizers 0.20.3