microsoft-wavlm-fleurs-ur

This model is a fine-tuned version of microsoft/wavlm-large on the fleurs dataset. It achieves the following results on the evaluation set:

• Loss: 0.7294
• Wer: 0.4026

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: 0.0003
• train_batch_size: 4
• eval_batch_size: 4
• seed: 42
• distributed_type: multi-GPU
• num_devices: 2
• total_train_batch_size: 8
• total_eval_batch_size: 8
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• lr_scheduler_warmup_steps: 500
• num_epochs: 15.0
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
3.911	0.35	100	3.7784	1.0
3.0833	0.71	200	3.0964	1.0
3.028	1.06	300	3.0377	1.0
2.5114	1.41	400	2.4941	0.9922
1.0583	1.77	500	1.0753	0.7579
0.715	2.12	600	0.8524	0.6410
0.6779	2.47	700	0.7711	0.6063
0.6123	2.83	800	0.7170	0.5706
0.8183	3.18	900	0.6897	0.5368
0.5195	3.53	1000	0.6586	0.5303
0.4774	3.89	1100	0.6306	0.5014
0.4242	4.24	1200	0.6138	0.4817
0.4549	4.59	1300	0.6027	0.4678
0.2576	4.95	1400	0.5878	0.4600
0.1578	5.3	1500	0.6144	0.4585
0.3556	5.65	1600	0.5884	0.4582
0.2427	6.01	1700	0.6071	0.4572
0.267	6.36	1800	0.6303	0.4514
0.2468	6.71	1900	0.6358	0.4495
0.159	7.07	2000	0.6242	0.4312
0.1527	7.42	2100	0.6372	0.4400
0.1401	7.77	2200	0.6252	0.4292
0.1211	8.13	2300	0.6358	0.4251
0.1022	8.48	2400	0.6529	0.4356
0.0818	8.83	2500	0.6773	0.4200
0.0918	9.19	2600	0.6879	0.4267
0.119	9.54	2700	0.6948	0.4254
0.1615	9.89	2800	0.6920	0.4259
0.0953	10.25	2900	0.7019	0.4218
0.1008	10.6	3000	0.6933	0.4133
0.0729	10.95	3100	0.6950	0.4164
0.0636	11.31	3200	0.7151	0.4121
0.0395	11.66	3300	0.7053	0.4098
0.0391	12.01	3400	0.7081	0.3984
0.0507	12.37	3500	0.7012	0.4111
0.0598	12.72	3600	0.7169	0.4035
0.0515	13.07	3700	0.7358	0.4102
0.0429	13.43	3800	0.7236	0.4013
0.0398	13.78	3900	0.7404	0.4026
0.0946	14.13	4000	0.7285	0.4029
0.0428	14.49	4100	0.7271	0.3991
0.0329	14.84	4200	0.7294	0.4026

Framework versions

• Transformers 4.27.0.dev0
• Pytorch 1.13.1
• Datasets 2.8.0
• Tokenizers 0.13.2