torgo_xlsr_finetune_F03

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: 0.2366
• Wer: 0.0535

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
3.4672	0.85	1000	3.4147	1.0
2.149	1.69	2000	0.9590	0.7064
0.9304	2.54	3000	0.3999	0.3896
0.6396	3.39	4000	0.2892	0.2332
0.5114	4.24	5000	0.2362	0.1701
0.454	5.08	6000	0.2241	0.1468
0.3899	5.93	7000	0.2356	0.0919
0.3466	6.78	8000	0.2607	0.0796
0.316	7.63	9000	0.2358	0.0919
0.2889	8.47	10000	0.2159	0.0617
0.2476	9.32	11000	0.2129	0.0713
0.2608	10.17	12000	0.1812	0.0878
0.2338	11.02	13000	0.2010	0.0631
0.2302	11.86	14000	0.1623	0.0576
0.2037	12.71	15000	0.2469	0.0604
0.1995	13.56	16000	0.1836	0.0562
0.1668	14.41	17000	0.2070	0.0741
0.1597	15.25	18000	0.2269	0.0658
0.1574	16.1	19000	0.2087	0.0713
0.1397	16.95	20000	0.1998	0.0576
0.1325	17.8	21000	0.1844	0.0549
0.1273	18.64	22000	0.2306	0.0535
0.1204	19.49	23000	0.2366	0.0535

Framework versions

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