wav2vec2-demo-M04-2

This model is a fine-tuned version of facebook/wav2vec2-large-xlsr-53 on the None dataset. It achieves the following results on the evaluation set:

• Loss: 2.0168
• Wer: 1.2882

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: 1000
• num_epochs: 30

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
21.8298	0.88	500	3.2643	1.0
3.2319	1.75	1000	2.8027	1.0
2.769	2.63	1500	2.4684	1.0
2.0823	3.5	2000	1.9137	1.6482
1.3094	4.38	2500	1.7267	1.6094
0.9654	5.25	3000	1.7523	1.4882
0.7505	6.13	3500	1.5588	1.5353
0.6364	7.01	4000	1.5428	1.4706
0.5307	7.88	4500	1.6277	1.4765
0.4664	8.76	5000	1.6817	1.3718
0.4243	9.63	5500	1.7682	1.4541
0.3911	10.51	6000	1.8567	1.4094
0.3555	11.38	6500	1.7248	1.3694
0.3252	12.26	7000	1.8712	1.4012
0.3072	13.13	7500	2.0088	1.4424
0.2956	14.01	8000	1.8649	1.3576
0.283	14.89	8500	1.8951	1.4035
0.2682	15.76	9000	1.8762	1.3976
0.2465	16.64	9500	1.8406	1.34
0.2344	17.51	10000	1.9975	1.3294
0.2269	18.39	10500	1.9207	1.3176
0.2053	19.26	11000	2.0406	1.3412
0.1934	20.14	11500	1.9039	1.2859
0.2018	21.02	12000	1.8337	1.3212
0.169	21.89	12500	1.9120	1.3071
0.1742	22.77	13000	2.0650	1.3153
0.1571	23.64	13500	2.0369	1.3165
0.1403	24.52	14000	2.0420	1.2894
0.1474	25.39	14500	1.9529	1.2847
0.1373	26.27	15000	2.0818	1.3129
0.1222	27.15	15500	1.9551	1.2753
0.1182	28.02	16000	2.0138	1.2659
0.1357	28.9	16500	1.9976	1.2859
0.1158	29.77	17000	2.0168	1.2882

Framework versions

• Transformers 4.23.1
• Pytorch 1.12.1+cu113
• Datasets 1.18.3
• Tokenizers 0.13.2