k2e_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: 1.5624
• Per: 0.1656
• Pcc: 0.5701
• Ctc Loss: 0.5415
• Mse Loss: 0.9981

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
45.3867	1.0	745	21.3997	0.9890	0.1720	5.7842	15.6506
10.6834	2.0	1490	4.9038	0.9890	0.4667	3.8839	1.0451
4.7613	3.0	2235	4.5932	0.9890	0.5577	3.8113	0.8478
4.5315	4.01	2980	4.4401	0.9890	0.6034	3.7726	0.7725
4.3577	5.01	3725	4.3030	0.9890	0.6066	3.7177	0.7229
4.1373	6.01	4470	4.2293	0.9890	0.6031	3.6366	0.7552
3.9539	7.01	5215	4.3927	0.9890	0.5807	3.6053	0.9706
3.7741	8.01	5960	4.1465	0.9886	0.5877	3.4982	0.8521
3.4397	9.01	6705	3.7672	0.9637	0.5536	2.9927	0.9519
2.8431	10.01	7450	3.0932	0.7901	0.5626	2.2607	0.9546
2.1711	11.01	8195	2.5671	0.5128	0.5584	1.6191	1.0089
1.6661	12.02	8940	2.1670	0.3754	0.5618	1.2111	0.9792
1.3242	13.02	9685	2.1092	0.2805	0.5532	0.9672	1.1267
1.1125	14.02	10430	1.9269	0.2294	0.5589	0.8275	1.0737
0.9919	15.02	11175	1.8918	0.2097	0.5555	0.7500	1.1043
0.8889	16.02	11920	1.8554	0.1982	0.5646	0.6987	1.1131
0.8228	17.02	12665	1.7697	0.1940	0.5610	0.6591	1.0697
0.7558	18.02	13410	1.6513	0.1864	0.5718	0.6363	0.9853
0.7122	19.03	14155	1.7786	0.1836	0.5614	0.6145	1.1156
0.6696	20.03	14900	1.6690	0.1796	0.5638	0.5980	1.0350
0.6352	21.03	15645	1.6720	0.1770	0.5670	0.5816	1.0522
0.6045	22.03	16390	1.6025	0.1739	0.5746	0.5745	0.9995
0.5786	23.03	17135	1.5682	0.1718	0.5675	0.5662	0.9784
0.5503	24.03	17880	1.5467	0.1707	0.5626	0.5568	0.9688
0.53	25.03	18625	1.5534	0.1684	0.5729	0.5535	0.9784
0.5161	26.03	19370	1.5191	0.1672	0.5679	0.5489	0.9540
0.4972	27.04	20115	1.5131	0.1668	0.5698	0.5468	0.9514
0.492	28.04	20860	1.5420	0.1664	0.5683	0.5422	0.9800
0.4795	29.04	21605	1.5279	0.1659	0.5703	0.5415	0.9690
0.4731	30.04	22350	1.5624	0.1656	0.5701	0.5415	0.9981

Framework versions

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