wav2vec2-timit-xls-r-53-wandb-colab

This model is a fine-tuned version of facebook/wav2vec2-xls-r-300m on the None dataset. It achieves the following results on the evaluation set:

• Loss: 0.3325
• Wer: 0.2897
• Cer: 0.0940

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: 30

Training results

Training Loss	Epoch	Step	Validation Loss	Wer	Cer
No log	0.69	400	3.1507	1.0	0.9806
4.3857	1.38	800	3.0109	1.0	0.9806
2.6835	2.08	1200	0.6181	0.5756	0.1795
0.9327	2.77	1600	0.4239	0.4718	0.1456
0.5602	3.46	2000	0.3691	0.4141	0.1301
0.5602	4.15	2400	0.3386	0.3894	0.1231
0.4407	4.84	2800	0.3122	0.3676	0.1177
0.3437	5.54	3200	0.3149	0.3601	0.1152
0.3154	6.23	3600	0.3146	0.3495	0.1119
0.267	6.92	4000	0.3039	0.3427	0.1089
0.267	7.61	4400	0.3313	0.3409	0.1092
0.2354	8.3	4800	0.2986	0.3365	0.1064
0.2191	9.0	5200	0.3235	0.3353	0.1074
0.1937	9.69	5600	0.3117	0.3320	0.1071
0.1803	10.38	6000	0.3102	0.3233	0.1040
0.1803	11.07	6400	0.3176	0.3196	0.1030
0.1635	11.76	6800	0.3166	0.3220	0.1036
0.1551	12.46	7200	0.2836	0.3160	0.1021
0.1566	13.15	7600	0.3146	0.3186	0.1032
0.1424	13.84	8000	0.3392	0.3167	0.1036
0.1424	14.53	8400	0.3254	0.3109	0.1001
0.1379	15.22	8800	0.3249	0.3127	0.1009
0.1192	15.92	9200	0.3408	0.3119	0.1010
0.1178	16.61	9600	0.3551	0.3061	0.0997
0.1112	17.3	10000	0.3250	0.3059	0.0991
0.1112	17.99	10400	0.3127	0.3037	0.0983
0.1022	18.69	10800	0.3370	0.3067	0.0994
0.1031	19.38	11200	0.3351	0.3048	0.0991
0.0926	20.07	11600	0.3433	0.2994	0.0974
0.0861	20.76	12000	0.3145	0.3003	0.0971
0.0861	21.45	12400	0.3367	0.2980	0.0973
0.0935	22.15	12800	0.3139	0.3016	0.0986
0.0784	22.84	13200	0.3181	0.2990	0.0972
0.078	23.53	13600	0.3347	0.2938	0.0961
0.0761	24.22	14000	0.3371	0.2921	0.0949
0.0761	24.91	14400	0.3274	0.2916	0.0952
0.0784	25.61	14800	0.3152	0.2927	0.0942
0.0714	26.3	15200	0.3237	0.2924	0.0943
0.0671	26.99	15600	0.3183	0.2914	0.0945
0.0684	27.68	16000	0.3307	0.2931	0.0950
0.0684	28.37	16400	0.3383	0.2913	0.0940
0.07	29.07	16800	0.3318	0.2901	0.0940
0.0624	29.76	17200	0.3325	0.2897	0.0940

Framework versions

• Transformers 4.32.0.dev0
• Pytorch 2.0.1+cu118
• Datasets 1.18.3
• Tokenizers 0.13.3