k2e-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: 1.4632
• Per: 0.1621
• Pcc: 0.5775
• Ctc Loss: 0.5292
• Mse Loss: 0.9318

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: 2235
• training_steps: 22350
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Per	Pcc	Ctc Loss	Mse Loss
43.3938	1.0	745	18.4290	0.9890	0.1462	6.2276	12.2287
9.2178	2.0	1490	4.8023	0.9890	0.4671	3.8571	0.9719
4.753	3.0	2235	4.7789	0.9890	0.5587	3.7983	1.0460
4.5085	4.01	2980	4.4859	0.9890	0.5896	3.6917	0.8937
4.2649	5.01	3725	4.2852	0.9890	0.6067	3.6166	0.7981
4.0688	6.01	4470	4.2684	0.9632	0.6001	3.5490	0.8725
3.9012	7.01	5215	4.3288	0.9628	0.5973	3.5020	0.9987
3.6391	8.01	5960	4.2077	0.9430	0.5765	3.1575	1.2098
3.0582	9.01	6705	3.4214	0.7155	0.5656	2.3835	1.1480
2.3307	10.01	7450	2.4659	0.4524	0.5672	1.6378	0.8917
1.7509	11.01	8195	2.3288	0.3189	0.5824	1.1749	1.1589
1.3679	12.02	8940	2.0410	0.2541	0.5669	0.9275	1.0986
1.1463	13.02	9685	1.8387	0.2238	0.5714	0.8029	1.0158
1.013	14.02	10430	1.6558	0.2068	0.5635	0.7253	0.9141
0.912	15.02	11175	1.6381	0.1992	0.5530	0.6788	0.9381
0.8321	16.02	11920	1.8336	0.1908	0.5660	0.6407	1.1459
0.7676	17.02	12665	1.6002	0.1819	0.5851	0.6197	0.9574
0.7186	18.02	13410	1.6110	0.1807	0.5562	0.5981	0.9870
0.6707	19.03	14155	1.6138	0.1748	0.5735	0.5827	1.0041
0.6362	20.03	14900	1.5090	0.1729	0.5706	0.5688	0.9264
0.6016	21.03	15645	1.5540	0.1698	0.5742	0.5619	0.9732
0.5724	22.03	16390	1.5076	0.1686	0.5846	0.5517	0.9435
0.5434	23.03	17135	1.4442	0.1676	0.5753	0.5443	0.8970
0.5272	24.03	17880	1.4617	0.1656	0.5699	0.5409	0.9165
0.5119	25.03	18625	1.4886	0.1642	0.5654	0.5400	0.9414
0.4963	26.03	19370	1.4959	0.1644	0.5751	0.5342	0.9534
0.4882	27.04	20115	1.4686	0.1634	0.5711	0.5320	0.9329
0.4697	28.04	20860	1.4663	0.1627	0.5730	0.5302	0.9330
0.4604	29.04	21605	1.4417	0.1623	0.5782	0.5293	0.9134
0.4536	30.04	22350	1.4632	0.1621	0.5775	0.5292	0.9318

Framework versions

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