torgo_xlsr_finetune_F04

This model is a fine-tuned version of facebook/wav2vec2-large-xlsr-53 on an unknown dataset. It achieves the following results on the evaluation set:

• Loss: 1.4132
• Wer: 0.2275

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: 4
• eval_batch_size: 4
• seed: 42
• gradient_accumulation_steps: 2
• total_train_batch_size: 8
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• lr_scheduler_warmup_steps: 1000
• num_epochs: 20

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
3.4699	0.85	1000	3.2861	1.0
2.1971	1.69	2000	2.0008	0.8514
0.9545	2.54	3000	1.4512	0.6358
0.6665	3.39	4000	1.4047	0.5008
0.5094	4.24	5000	1.3973	0.4457
0.4719	5.08	6000	1.4290	0.4066
0.4183	5.93	7000	1.4807	0.3761
0.3525	6.78	8000	1.5710	0.3667
0.3112	7.63	9000	1.4555	0.3268
0.2876	8.47	10000	1.4537	0.2988
0.2321	9.32	11000	1.6268	0.3200
0.2456	10.17	12000	1.3804	0.2852
0.2376	11.02	13000	1.6112	0.3141
0.2169	11.86	14000	1.4480	0.2988
0.2106	12.71	15000	1.6790	0.2929
0.2055	13.56	16000	1.5383	0.2963
0.1601	14.41	17000	1.4142	0.2555
0.1631	15.25	18000	1.5318	0.2470
0.1481	16.1	19000	1.6078	0.2453
0.1374	16.95	20000	1.3588	0.2360
0.1349	17.8	21000	1.3788	0.2309
0.1284	18.64	22000	1.4818	0.2326
0.1328	19.49	23000	1.4132	0.2275

Framework versions

• Transformers 4.26.1
• Pytorch 2.2.1
• Datasets 2.18.0
• Tokenizers 0.13.3