torgo_xlsr_finetune_F03_keep_all

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: 0.1972
• Wer: 0.0443

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
3.573	0.56	1000	3.3927	1.0
1.9872	1.11	2000	1.0729	0.7588
1.0338	1.67	3000	0.6272	0.4530
0.7363	2.22	4000	0.5054	0.3579
0.6181	2.78	5000	0.3583	0.2448
0.5327	3.33	6000	0.2902	0.1771
0.5046	3.89	7000	0.3202	0.1628
0.4227	4.44	8000	0.2460	0.1400
0.4405	5.0	9000	0.2750	0.1287
0.3927	5.55	10000	0.2235	0.0928
0.3496	6.11	11000	0.2621	0.1083
0.3294	6.66	12000	0.2056	0.0688
0.3201	7.22	13000	0.2155	0.0838
0.2973	7.77	14000	0.2909	0.0844
0.2758	8.33	15000	0.3096	0.0832
0.2688	8.88	16000	0.2420	0.0664
0.2537	9.44	17000	0.2438	0.0772
0.2642	9.99	18000	0.1836	0.0598
0.2611	10.55	19000	0.2562	0.0670
0.2019	11.1	20000	0.2084	0.0658
0.205	11.66	21000	0.1959	0.0628
0.2019	12.22	22000	0.2316	0.0700
0.1923	12.77	23000	0.2051	0.0664
0.1614	13.33	24000	0.2322	0.0724
0.1955	13.88	25000	0.1920	0.0551
0.1705	14.44	26000	0.2071	0.0521
0.1906	14.99	27000	0.2112	0.0533
0.1503	15.55	28000	0.1914	0.0491
0.1692	16.1	29000	0.1870	0.0557
0.149	16.66	30000	0.2172	0.0467
0.1413	17.21	31000	0.2313	0.0479
0.1466	17.77	32000	0.1989	0.0437
0.1533	18.32	33000	0.2078	0.0437
0.1176	18.88	34000	0.2034	0.0431
0.1189	19.43	35000	0.1929	0.0413
0.1056	19.99	36000	0.1972	0.0443

Framework versions

• Transformers 4.26.1
• Pytorch 2.2.0
• Datasets 2.16.1
• Tokenizers 0.13.3