wav2vec2-speech-project-ownData

This model is a fine-tuned version of facebook/wav2vec2-base on the None dataset. It achieves the following results on the evaluation set:

• Loss: 0.1265
• Wer: 0.2154

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: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• lr_scheduler_warmup_steps: 1000
• num_epochs: 30
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
19.2051	1.67	50	17.2729	1.0
7.801	3.33	100	3.3743	1.0
3.2115	5.0	150	3.1156	1.0
3.0852	6.67	200	3.0413	1.0
3.025	8.33	250	3.0108	1.0
2.9964	10.0	300	3.0003	1.0
2.9638	11.67	350	2.9466	1.0
2.8766	13.33	400	2.7914	1.0
2.5502	15.0	450	2.1839	0.9969
1.9802	16.67	500	1.5820	0.7569
1.324	18.33	550	0.8405	0.6
0.67	20.0	600	0.4139	0.3323
0.3792	21.67	650	0.2685	0.2831
0.2518	23.33	700	0.1775	0.2523
0.208	25.0	750	0.1274	0.2308
0.1619	26.67	800	0.2169	0.2462
0.1603	28.33	850	0.1142	0.1846
0.1246	30.0	900	0.1265	0.2154

Framework versions

• Transformers 4.28.1
• Pytorch 2.0.0+cu118
• Datasets 1.18.3
• Tokenizers 0.13.3