k2e-20s_asr-scr_w2v2-base_001

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: 1.5230
• Per: 0.1454
• Pcc: 0.5490
• Ctc Loss: 0.5155
• Mse Loss: 0.9906

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: 1111
• 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
42.4921	1.0	745	19.1354	0.9890	0.1755	6.0201	13.1446
9.6761	2.0	1490	4.8873	0.9890	0.3766	3.8628	1.0511
4.7811	3.0	2235	4.6462	0.9890	0.5747	3.8063	0.9061
4.5636	4.01	2980	4.4298	0.9890	0.5829	3.7773	0.7581
4.3976	5.01	3725	4.3800	0.9890	0.6088	3.7597	0.7604
4.2244	6.01	4470	4.4381	0.9890	0.5888	3.6791	0.9234
4.0281	7.01	5215	4.4452	0.9890	0.5979	3.6172	1.0127
3.8406	8.01	5960	4.3227	0.9884	0.5790	3.5061	1.0160
3.4504	9.01	6705	3.7651	0.9557	0.5562	2.8520	1.0726
2.6451	10.01	7450	3.2489	0.6173	0.5703	1.9227	1.3898
1.89	11.01	8195	2.1831	0.3574	0.5481	1.2651	0.9472
1.4355	12.02	8940	2.1442	0.2583	0.5619	0.9769	1.1527
1.2033	13.02	9685	1.8016	0.2317	0.5534	0.8432	0.9477
1.0366	14.02	10430	1.9141	0.2145	0.5525	0.7478	1.1287
0.9253	15.02	11175	1.9080	0.2019	0.5479	0.6880	1.1717
0.8488	16.02	11920	1.6636	0.1923	0.5558	0.6417	0.9913
0.7648	17.02	12665	1.5709	0.1837	0.5517	0.6131	0.9345
0.7179	18.02	13410	1.6913	0.1798	0.5501	0.5893	1.0623
0.6645	19.03	14155	1.6498	0.1760	0.5565	0.5766	1.0380
0.6345	20.03	14900	1.7144	0.1741	0.5650	0.5604	1.1090
0.5919	21.03	15645	1.6624	0.1719	0.5581	0.5480	1.0756
0.5616	22.03	16390	1.5461	0.1695	0.5629	0.5467	0.9780
0.5371	23.03	17135	1.5791	0.1674	0.5533	0.5360	1.0165
0.5074	24.03	17880	1.5947	0.1662	0.5474	0.5267	1.0386
0.4922	25.03	18625	1.4868	0.1652	0.5489	0.5250	0.9494
0.473	26.03	19370	1.5373	0.1646	0.5576	0.5226	0.9952
0.4671	27.04	20115	1.5479	0.1638	0.5540	0.5201	1.0069
0.452	28.04	20860	1.5199	0.1635	0.5555	0.5163	0.9869
0.4435	29.04	21605	1.5116	0.1634	0.5544	0.5154	0.9810
0.439	30.04	22350	1.5230	0.1633	0.5567	0.5155	0.9906

Framework versions

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