xlsr-a-nomimose

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: 0.4623
• Wer: 0.3473

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: 0.0004
• train_batch_size: 8
• eval_batch_size: 8
• seed: 42
• gradient_accumulation_steps: 2
• total_train_batch_size: 16
• optimizer: Use adamw_torch with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
• lr_scheduler_type: linear
• lr_scheduler_warmup_steps: 132
• num_epochs: 100
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
4.7536	1.8387	200	2.6165	1.0
1.8115	3.6728	400	0.6163	0.6689
0.469	5.5069	600	0.5160	0.5442
0.2745	7.3410	800	0.2917	0.4108
0.1694	9.1751	1000	0.3568	0.4624
0.1782	11.0092	1200	0.3223	0.3901
0.1406	12.8479	1400	0.2959	0.3923
0.1046	14.6820	1600	0.3131	0.3658
0.109	16.5161	1800	0.3343	0.3739
0.1135	18.3502	2000	0.3368	0.3577
0.0875	20.1843	2200	0.4722	0.3695
0.0956	22.0184	2400	0.3427	0.3614
0.0749	23.8571	2600	0.3377	0.3695
0.0727	25.6912	2800	0.3489	0.3599
0.0655	27.5253	3000	0.3348	0.3665
0.0603	29.3594	3200	0.3636	0.3606
0.0537	31.1935	3400	0.3923	0.3555
0.0567	33.0276	3600	0.3476	0.3555
0.0461	34.8664	3800	0.3589	0.3628
0.0459	36.7005	4000	0.4104	0.3584
0.0467	38.5346	4200	0.3686	0.3555
0.0419	40.3687	4400	0.3889	0.3555
0.0381	42.2028	4600	0.4013	0.3540
0.0403	44.0369	4800	0.4077	0.3555
0.0371	45.8756	5000	0.4502	0.3577
0.0354	47.7097	5200	0.4884	0.3739
0.0372	49.5438	5400	0.4227	0.3614
0.0344	51.3779	5600	0.3949	0.3532
0.0288	53.2120	5800	0.4088	0.3532
0.0314	55.0461	6000	0.4194	0.3496
0.0276	56.8848	6200	0.4184	0.3532
0.0232	58.7189	6400	0.4184	0.3496
0.0267	60.5530	6600	0.4028	0.3496
0.0271	62.3871	6800	0.3804	0.3488
0.0205	64.2212	7000	0.4735	0.3466
0.0243	66.0553	7200	0.4108	0.3473
0.0198	67.8940	7400	0.4157	0.3488
0.0193	69.7281	7600	0.4226	0.3547
0.0222	71.5622	7800	0.4147	0.3451
0.0176	73.3963	8000	0.4553	0.3488
0.0209	75.2304	8200	0.4135	0.3525
0.0222	77.0645	8400	0.4300	0.3481
0.0169	78.9032	8600	0.4139	0.3481
0.0174	80.7373	8800	0.4510	0.3473
0.019	82.5714	9000	0.4664	0.3488
0.0145	84.4055	9200	0.4768	0.3459
0.0146	86.2396	9400	0.4678	0.3466
0.0184	88.0737	9600	0.4906	0.3488
0.0141	89.9124	9800	0.4676	0.3481
0.0128	91.7465	10000	0.4612	0.3473
0.0134	93.5806	10200	0.4649	0.3459
0.0134	95.4147	10400	0.4606	0.3481
0.0126	97.2488	10600	0.4646	0.3488
0.014	99.0829	10800	0.4623	0.3473

Framework versions

• Transformers 4.47.0.dev0
• Pytorch 2.4.0
• Datasets 3.0.1
• Tokenizers 0.20.0