arampacha/wav2vec2-xls-r-300m-hy

This model is a fine-tuned version of facebook/wav2vec2-xls-r-300m on the /WORKSPACE/DATA/HY/NOIZY_STUDENT_3/ - NA dataset. It achieves the following results on the evaluation set:

• Loss: 0.2293
• Wer: 0.3333
• Cer: 0.0602

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: 7e-05
• train_batch_size: 64
• eval_batch_size: 64
• seed: 842
• gradient_accumulation_steps: 2
• total_train_batch_size: 128
• optimizer: Adam with betas=(0.9,0.98) and epsilon=1e-08
• lr_scheduler_type: cosine
• lr_scheduler_warmup_ratio: 0.1
• training_steps: 4000
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer	Cer
3.1471	7.02	400	3.1599	1.0	1.0
1.8691	14.04	800	0.7674	0.7361	0.1686
1.3227	21.05	1200	0.3849	0.5336	0.1007
1.163	28.07	1600	0.3015	0.4559	0.0823
1.0768	35.09	2000	0.2721	0.4032	0.0728
1.0224	42.11	2400	0.2586	0.3825	0.0691
0.9817	49.12	2800	0.2458	0.3653	0.0653
0.941	56.14	3200	0.2306	0.3388	0.0605
0.9235	63.16	3600	0.2315	0.3380	0.0615
0.9141	70.18	4000	0.2293	0.3333	0.0602

Framework versions

• Transformers 4.17.0.dev0
• Pytorch 1.10.2
• Datasets 1.18.4.dev0
• Tokenizers 0.11.0