wav2vec2-timit-xls-r-300m-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: 0.3293
• Wer: 0.2879
• Cer: 0.0927

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	3.1501	1.0	0.9865
4.3829	1.38	800	2.9694	1.0	0.9865
2.6389	2.08	1200	0.6558	0.5876	0.1878
0.9293	2.77	1600	0.4232	0.4722	0.1462
0.5686	3.46	2000	0.3513	0.4118	0.1279
0.5686	4.15	2400	0.3246	0.3994	0.1227
0.4388	4.84	2800	0.3037	0.3716	0.1158
0.3431	5.54	3200	0.3055	0.3674	0.1158
0.3164	6.23	3600	0.2973	0.3589	0.1128
0.2678	6.92	4000	0.3053	0.3421	0.1080
0.2678	7.61	4400	0.3058	0.3435	0.1083
0.2376	8.3	4800	0.3144	0.3408	0.1094
0.2199	9.0	5200	0.3177	0.3371	0.1052
0.1988	9.69	5600	0.3123	0.3299	0.1057
0.1816	10.38	6000	0.2918	0.3282	0.1049
0.1816	11.07	6400	0.3195	0.3270	0.1049
0.1652	11.76	6800	0.3080	0.3280	0.1056
0.1576	12.46	7200	0.2859	0.3218	0.1031
0.1558	13.15	7600	0.3143	0.3179	0.1018
0.1411	13.84	8000	0.3354	0.3171	0.1045
0.1411	14.53	8400	0.3285	0.3149	0.1018
0.1381	15.22	8800	0.3048	0.3138	0.1010
0.1178	15.92	9200	0.3421	0.3140	0.1012
0.1182	16.61	9600	0.3258	0.3109	0.1001
0.1131	17.3	10000	0.3220	0.3120	0.1002
0.1131	17.99	10400	0.3156	0.3098	0.0991
0.1031	18.69	10800	0.3198	0.3062	0.0980
0.1023	19.38	11200	0.3227	0.3021	0.0972
0.0959	20.07	11600	0.3187	0.3025	0.0973
0.0881	20.76	12000	0.3177	0.3004	0.0965
0.0881	21.45	12400	0.3435	0.2976	0.0960
0.0919	22.15	12800	0.3142	0.2958	0.0954
0.0787	22.84	13200	0.3010	0.3000	0.0970
0.0794	23.53	13600	0.3528	0.3008	0.0973
0.0751	24.22	14000	0.3352	0.2954	0.0961
0.0751	24.91	14400	0.3314	0.2977	0.0963
0.0778	25.61	14800	0.3214	0.2955	0.0953
0.0711	26.3	15200	0.3277	0.2936	0.0944
0.0681	26.99	15600	0.3237	0.2915	0.0940
0.0682	27.68	16000	0.3284	0.2918	0.0943
0.0682	28.37	16400	0.3304	0.2904	0.0933
0.0731	29.07	16800	0.3307	0.2881	0.0927
0.0619	29.76	17200	0.3293	0.2879	0.0927

Framework versions

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