tokens

This model is a fine-tuned version of facebook/wav2vec2-xls-r-300m on the common_voice dataset. It achieves the following results on the evaluation set:

• Loss: 0.9811
• Wer: 0.4608

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: 0.0003
• train_batch_size: 16
• eval_batch_size: 16
• seed: 42
• gradient_accumulation_steps: 2
• total_train_batch_size: 32
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• lr_scheduler_warmup_steps: 500
• num_epochs: 30
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
6.5212	0.59	400	3.3776	1.0
2.4798	1.18	800	1.0697	0.7740
1.0057	1.77	1200	0.7077	0.6487
0.7731	2.36	1600	0.6113	0.5883
0.6917	2.94	2000	0.5618	0.5573
0.5844	3.53	2400	0.5610	0.5532
0.5606	4.12	2800	0.5584	0.5484
0.4973	4.71	3200	0.5466	0.5333
0.4721	5.3	3600	0.5495	0.5178
0.4439	5.89	4000	0.5667	0.5237
0.3965	6.48	4400	0.5865	0.5322
0.3876	7.07	4800	0.6099	0.5135
0.3407	7.66	5200	0.5891	0.5228
0.33	8.25	5600	0.6135	0.5072
0.3032	8.84	6000	0.6004	0.5028
0.2706	9.43	6400	0.6321	0.4991
0.2709	10.01	6800	0.6541	0.5051
0.2373	10.6	7200	0.6613	0.5119
0.2284	11.19	7600	0.6798	0.5086
0.212	11.78	8000	0.6509	0.4910
0.1983	12.37	8400	0.7018	0.5043
0.1947	12.96	8800	0.6826	0.4965
0.1717	13.55	9200	0.7056	0.4828
0.1741	14.14	9600	0.7544	0.5060
0.1626	14.73	10000	0.7331	0.4915
0.1529	15.32	10400	0.7518	0.4772
0.1504	15.91	10800	0.7362	0.4732
0.1401	16.49	11200	0.7179	0.4769
0.1335	17.08	11600	0.7716	0.4826
0.1185	17.67	12000	0.7465	0.4798
0.1182	18.26	12400	0.8105	0.4733
0.1135	18.85	12800	0.7693	0.4743
0.1098	19.44	13200	0.8362	0.4888
0.1023	20.03	13600	0.8427	0.4768
0.1003	20.62	14000	0.8079	0.4741
0.0936	21.21	14400	0.8551	0.4651
0.0875	21.8	14800	0.8462	0.4712
0.0843	22.39	15200	0.9177	0.4782
0.0846	22.97	15600	0.8618	0.4735
0.08	23.56	16000	0.9017	0.4687
0.0789	24.15	16400	0.9034	0.4659
0.0717	24.74	16800	0.9690	0.4734
0.0714	25.33	17200	0.9395	0.4677
0.0699	25.92	17600	0.9222	0.4608
0.0658	26.51	18000	0.9222	0.4621
0.0612	27.1	18400	0.9691	0.4586
0.0583	27.69	18800	0.9647	0.4581
0.0596	28.28	19200	0.9820	0.4614
0.056	28.87	19600	0.9795	0.4596
0.055	29.45	20000	0.9811	0.4608

Framework versions

• Transformers 4.11.3
• Pytorch 1.10.0+cu113
• Datasets 1.18.3
• Tokenizers 0.10.3