somali-may-asr-wav2vec2

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: 2.8575
• Wer: 1.0

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: 8
• 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

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
5.8543	2.75	400	1.5898	0.9924
1.1906	5.5	800	1.0469	0.8550
2.1171	8.25	1200	2.9265	1.0
3.0353	11.0	1600	2.8589	1.0
3.0736	13.75	2000	2.8575	1.0
3.0843	16.49	2400	2.8575	1.0
3.0777	19.24	2800	2.8575	1.0
3.0818	21.99	3200	2.8575	1.0
3.077	24.74	3600	2.8575	1.0
3.086	27.49	4000	2.8575	1.0

Framework versions

• Transformers 4.33.1
• Pytorch 2.1.2
• Datasets 2.19.2
• Tokenizers 0.13.3