xlsr-aiish-nomo

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: 0.0000
• Wer: 0.3105

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: Adam with betas=(0.9,0.999) and epsilon=1e-08
• 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
4.5816	1.7167	200	2.5505	1.0
1.5977	3.4335	400	0.1411	0.5024
0.2835	5.1502	600	0.0321	0.3716
0.1323	6.8670	800	0.0247	0.3166
0.0826	8.5837	1000	0.0093	0.3289
0.0816	10.3004	1200	0.0032	0.3105
0.0543	12.0172	1400	0.0022	0.3130
0.0516	13.7339	1600	0.0045	0.3105
0.0399	15.4506	1800	0.0028	0.3130
0.0368	17.1674	2000	0.0099	0.3252
0.032	18.8841	2200	0.0007	0.3105
0.0349	20.6009	2400	0.0152	0.3240
0.025	22.3176	2600	0.0004	0.3105
0.0267	24.0343	2800	0.0003	0.3105
0.0279	25.7511	3000	0.0003	0.3105
0.0231	27.4678	3200	0.0014	0.3142
0.0311	29.1845	3400	0.0005	0.3105
0.0209	30.9013	3600	0.0015	0.3105
0.0175	32.6180	3800	0.0023	0.3105
0.0158	34.3348	4000	0.0002	0.3105
0.0189	36.0515	4200	0.0004	0.3105
0.0137	37.7682	4400	0.0001	0.3105
0.0138	39.4850	4600	0.0002	0.3105
0.0141	41.2017	4800	0.0002	0.3105
0.0147	42.9185	5000	0.0011	0.3105
0.013	44.6352	5200	0.0020	0.3130
0.0204	46.3519	5400	0.0011	0.3154
0.0131	48.0687	5600	0.0018	0.3117
0.0109	49.7854	5800	0.0053	0.3105
0.0135	51.5021	6000	0.0038	0.3068
0.0112	53.2189	6200	0.0003	0.3081
0.0075	54.9356	6400	0.0001	0.3081
0.0116	56.6524	6600	0.0001	0.3068
0.0048	58.3691	6800	0.0013	0.3093
0.0077	60.0858	7000	0.0000	0.3081
0.009	61.8026	7200	0.0002	0.3081
0.0043	63.5193	7400	0.0001	0.3081
0.0043	65.2361	7600	0.0017	0.3093
0.0063	66.9528	7800	0.0000	0.3081
0.0051	68.6695	8000	0.0000	0.3081
0.0062	70.3863	8200	0.0001	0.3081
0.0027	72.1030	8400	0.0000	0.3081
0.0043	73.8197	8600	0.0000	0.3068
0.0031	75.5365	8800	0.0000	0.3081
0.004	77.2532	9000	0.0001	0.3081
0.0045	78.9700	9200	0.0000	0.3093
0.0049	80.6867	9400	0.0000	0.3081
0.0026	82.4034	9600	0.0001	0.3093
0.0025	84.1202	9800	0.0000	0.3081
0.0029	85.8369	10000	0.0000	0.3093
0.0025	87.5536	10200	0.0000	0.3105
0.0029	89.2704	10400	0.0000	0.3105
0.0018	90.9871	10600	0.0000	0.3105
0.0027	92.7039	10800	0.0000	0.3105
0.0025	94.4206	11000	0.0000	0.3105
0.0014	96.1373	11200	0.0000	0.3105
0.0021	97.8541	11400	0.0000	0.3105
0.0016	99.5708	11600	0.0000	0.3105

Framework versions

• Transformers 4.45.0.dev0
• Pytorch 2.4.0
• Datasets 2.21.0
• Tokenizers 0.19.1