saq-20s_asr-scr_w2v2-base_002

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: 1.5036
• Per: 0.1541
• Pcc: 0.6677
• Ctc Loss: 0.5422
• Mse Loss: 0.9427

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: 1e-05
• train_batch_size: 16
• eval_batch_size: 1
• seed: 2222
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• lr_scheduler_warmup_steps: 2226
• training_steps: 22260
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Per	Pcc	Ctc Loss	Mse Loss
16.9349	3.0	2226	4.6257	0.9983	0.6397	3.7753	0.9152
4.358	6.0	4452	4.3728	0.9983	0.6743	3.7449	0.7973
3.976	9.0	6678	4.2399	0.9983	0.6928	3.6699	0.8195
2.9839	12.0	8904	2.3433	0.3730	0.6740	1.5100	0.8973
1.2641	15.0	11130	1.7650	0.2095	0.6732	0.7985	0.9498
0.8466	18.0	13356	1.5664	0.1818	0.6642	0.6611	0.8872
0.6752	21.0	15582	1.5958	0.1708	0.6690	0.6012	0.9664
0.5802	24.0	17808	1.7719	0.1651	0.6737	0.5668	1.1474
0.5266	27.0	20034	1.6479	0.1577	0.6707	0.5482	1.0587
0.4851	30.0	22260	1.5036	0.1541	0.6677	0.5422	0.9427

Framework versions

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