las-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: 1.9105
• Per: 0.1627
• Pcc: 0.6615
• Ctc Loss: 0.5322
• Mse Loss: 1.2951

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
39.6857	1.0	742	18.9148	0.9897	0.3162	5.7239	13.2193
9.7583	2.0	1484	5.1408	0.9897	0.5908	3.8631	1.3041
4.8781	3.0	2226	5.1735	0.9897	0.6553	3.7845	1.4538
4.6095	4.0	2968	4.6937	0.9897	0.6826	3.7654	1.0306
4.4289	5.0	3710	4.6305	0.9897	0.6823	3.7502	1.0155
4.2524	6.0	4452	4.8477	0.9897	0.6824	3.6937	1.3131
4.0682	7.0	5194	4.6683	0.9745	0.6899	3.6168	1.2312
3.8969	8.0	5936	4.4112	0.9719	0.6860	3.5709	1.0439
3.7246	9.0	6678	4.3463	0.9717	0.6836	3.4542	1.1059
3.4041	10.0	7420	4.0102	0.9714	0.6710	3.0131	1.1831
2.8571	11.0	8162	3.2679	0.7278	0.6639	2.2983	1.0980
2.1949	12.0	8904	3.0018	0.4811	0.6546	1.6107	1.4258
1.6457	13.0	9646	2.4409	0.3084	0.6579	1.1488	1.2846
1.3213	14.0	10388	2.0040	0.2490	0.6628	0.9346	1.0566
1.1321	15.0	11130	2.0496	0.2280	0.6471	0.8158	1.1931
1.0106	16.0	11872	2.0461	0.2133	0.6625	0.7316	1.2561
0.9147	17.0	12614	1.9235	0.1973	0.6696	0.6863	1.1818
0.8449	18.0	13356	1.9574	0.1876	0.6619	0.6482	1.2421
0.7954	19.0	14098	2.1967	0.1836	0.6674	0.6185	1.4764
0.7419	20.0	14840	1.8551	0.1775	0.6709	0.5962	1.1941
0.7084	21.0	15582	1.9213	0.1746	0.6592	0.5807	1.2646
0.6668	22.0	16324	1.9464	0.1721	0.6689	0.5661	1.2983
0.6403	23.0	17066	1.8137	0.1685	0.6666	0.5603	1.1886
0.616	24.0	17808	1.7154	0.1664	0.6654	0.5502	1.1127
0.6046	25.0	18550	1.6261	0.1657	0.6642	0.5434	1.0422
0.5996	26.0	19292	1.6967	0.1648	0.6636	0.5424	1.1041
0.58	27.0	20034	1.8841	0.1640	0.6625	0.5367	1.2689
0.5592	28.0	20776	1.8360	0.1631	0.6626	0.5345	1.2297
0.5502	29.0	21518	1.9516	0.1630	0.6581	0.5339	1.3286
0.5398	30.0	22260	1.9105	0.1627	0.6615	0.5322	1.2951

Framework versions

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