torgo_xlsr_finetune_M01_word_only_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: 1.5197
• Wer: 0.4523

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: 10

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
No log	0.28	250	5.5883	1.0
32.6551	0.57	500	3.4592	1.0
32.6551	0.85	750	3.4807	1.0
3.536	1.13	1000	3.4312	1.0
3.536	1.42	1250	3.4500	1.0
3.4905	1.7	1500	3.4075	1.0
3.4905	1.98	1750	3.1965	1.0
3.1337	2.27	2000	2.4374	1.0
3.1337	2.55	2250	1.9984	0.9512
1.796	2.83	2500	1.7787	0.8337
1.796	3.11	2750	1.7683	0.8248
1.0903	3.4	3000	1.4810	0.6452
1.0903	3.68	3250	1.5670	0.6408
0.8207	3.96	3500	1.4513	0.6585
0.8207	4.25	3750	1.2161	0.5632
0.6562	4.53	4000	1.5037	0.6208
0.6562	4.81	4250	1.2742	0.5698
0.5897	5.1	4500	1.4467	0.5543
0.5897	5.38	4750	1.5197	0.5477
0.5008	5.66	5000	1.6018	0.5654
0.5008	5.95	5250	1.4544	0.5299
0.4653	6.23	5500	1.4768	0.5122
0.4653	6.51	5750	1.4434	0.5277
0.4165	6.8	6000	1.4197	0.4834
0.4165	7.08	6250	1.5469	0.5299
0.3873	7.36	6500	1.5274	0.5233
0.3873	7.64	6750	1.4026	0.4812
0.355	7.93	7000	1.4866	0.4723
0.355	8.21	7250	1.5028	0.4989
0.3382	8.49	7500	1.5790	0.4834
0.3382	8.78	7750	1.4626	0.4590
0.2974	9.06	8000	1.5658	0.4856
0.2974	9.34	8250	1.5510	0.4701
0.2975	9.63	8500	1.5392	0.4501
0.2975	9.91	8750	1.5197	0.4523

Framework versions

• Transformers 4.26.1
• Pytorch 2.1.0+cu121
• Datasets 2.16.1
• Tokenizers 0.13.3