excalibur12/pic-20s_asr-scr_w2v2-base_003

pic-20s_asr-scr_w2v2-base_004

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: 1.4351
• Per: 0.1529
• Pcc: 0.6741
• Ctc Loss: 0.5362
• Mse Loss: 0.8881

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: 1234
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• lr_scheduler_warmup_steps: 2247
• training_steps: 22470
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Per	Pcc	Ctc Loss	Mse Loss
17.1513	3.0	2247	4.7158	0.9979	0.6302	3.7674	1.0133
4.2201	6.0	4494	4.2724	0.9979	0.7096	3.7172	0.7331
3.7967	9.0	6741	4.3058	0.9975	0.6875	3.6713	0.8900
3.0022	12.0	8988	2.4439	0.5310	0.6794	1.8297	0.7356
1.3508	15.0	11235	1.8496	0.2221	0.6786	0.8430	0.9894
0.825	18.0	13482	1.5793	0.1815	0.6751	0.6577	0.8999
0.6436	21.0	15729	1.7474	0.1690	0.6748	0.5867	1.1030
0.5476	24.0	17976	1.4824	0.1589	0.6730	0.5570	0.9069
0.4889	27.0	20223	1.3293	0.1544	0.6701	0.5413	0.7952
0.4593	30.0	22470	1.4351	0.1529	0.6741	0.5362	0.8881

Framework versions

• Transformers 4.38.1
• Pytorch 2.0.1
• Datasets 2.18.0
• Tokenizers 0.15.2