saq_asr-scr_w2v2-base_001

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: 1.2549
• Per: 0.1327
• Pcc: 0.6578
• Ctc Loss: 0.4805
• Mse Loss: 1.0040

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: 16
• eval_batch_size: 1
• seed: 1111
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• lr_scheduler_warmup_steps: 742
• training_steps: 7420

Training results

Training Loss	Epoch	Step	Validation Loss	Per	Pcc	Ctc Loss	Mse Loss
11.7284	1.0	742	4.5186	0.9994	0.5696	3.7385	0.9749
3.1398	2.0	1484	2.4884	0.2042	0.6246	0.7601	1.6844
1.5121	3.0	2226	1.5395	0.1627	0.6359	0.5898	0.9117
1.0897	4.0	2968	1.4423	0.1551	0.6390	0.5386	0.8928
0.6968	5.0	3710	1.5142	0.1477	0.6443	0.5085	1.0010
0.3184	6.0	4452	1.8725	0.1411	0.6557	0.4879	1.2796
-0.0502	7.0	5194	1.4015	0.1387	0.6577	0.4808	1.0161
-0.3567	8.0	5936	1.3481	0.1345	0.6557	0.4852	1.0170
-0.5908	9.0	6678	1.2779	0.1340	0.6604	0.4810	1.0066
-0.7364	10.0	7420	1.2549	0.1327	0.6578	0.4805	1.0040

Framework versions

• Transformers 4.38.1
• Pytorch 2.0.1
• Datasets 2.16.1
• Tokenizers 0.15.2