Siyong/MC_RN

Millad_Customer_RN

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: 4.5635
• Wer: 0.8113
• Cer: 0.4817

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: 0.0001
• train_batch_size: 8
• eval_batch_size: 8
• seed: 42
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• lr_scheduler_warmup_steps: 4000
• num_epochs: 600
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer	Cer
1.9257	13.33	2000	2.0606	0.9767	0.5500
1.4828	26.67	4000	2.1161	0.9019	0.4932
1.2582	40.0	6000	2.0589	0.8504	0.4942
0.9804	53.33	8000	2.4633	0.8745	0.4763
0.7862	66.67	10000	2.4794	0.8861	0.4944
0.6492	80.0	12000	2.8693	0.8554	0.4928
0.5375	93.33	14000	2.6125	0.8296	0.4802
0.4462	106.67	16000	2.7591	0.8770	0.4974
0.3873	120.0	18000	3.0325	0.8379	0.4800
0.3445	133.33	20000	2.9965	0.8761	0.4986
0.3087	146.67	22000	3.3437	0.8221	0.4923
0.2755	160.0	24000	3.3022	0.8803	0.5211
0.2467	173.33	26000	3.2348	0.8479	0.4933
0.2281	186.67	28000	3.8010	0.8695	0.5081
0.2119	200.0	30000	3.0446	0.8545	0.4902
0.194	213.33	32000	3.0873	0.8454	0.4840
0.1677	226.67	34000	3.6184	0.8645	0.5019
0.1642	240.0	36000	3.2480	0.8412	0.4903
0.1656	253.33	38000	3.4379	0.8362	0.4816
0.1371	266.67	40000	3.5117	0.8479	0.5040
0.1301	280.0	42000	3.4360	0.8404	0.4870
0.128	293.33	44000	3.6589	0.8537	0.4977
0.1152	306.67	46000	4.2359	0.8545	0.5051
0.1119	320.0	48000	3.5818	0.7980	0.4882
0.1026	333.33	50000	3.7618	0.8013	0.4865
0.0945	346.67	52000	4.2197	0.8404	0.5028
0.0962	360.0	54000	3.9231	0.8653	0.5030
0.088	373.33	56000	3.8400	0.8354	0.4914
0.0743	386.67	58000	3.4924	0.8088	0.4824
0.0811	400.0	60000	3.8370	0.8396	0.4861
0.0696	413.33	62000	4.2808	0.8412	0.5065
0.0692	426.67	64000	4.0161	0.8088	0.4744
0.0622	440.0	66000	3.9080	0.8163	0.4910
0.0591	453.33	68000	3.9838	0.8113	0.4823
0.0527	466.67	70000	3.8067	0.8329	0.4914
0.056	480.0	72000	4.1415	0.8096	0.4782
0.0535	493.33	74000	4.3350	0.8229	0.4828
0.0531	506.67	76000	3.9808	0.8071	0.4807
0.0451	520.0	78000	4.0301	0.7988	0.4816
0.044	533.33	80000	4.4680	0.8371	0.4921
0.0389	546.67	82000	4.1380	0.8121	0.4819
0.0392	560.0	84000	4.3910	0.7930	0.4763
0.0389	573.33	86000	4.5086	0.8055	0.4802
0.0355	586.67	88000	4.6259	0.8113	0.4821
0.0307	600.0	90000	4.5635	0.8113	0.4817

Framework versions

• Transformers 4.17.0
• Pytorch 1.12.0+cu113
• Datasets 1.18.3
• Tokenizers 0.12.1