whisper-malagasy-medium-v2

This model is a fine-tuned version of openai/whisper-medium on an unknown dataset. It achieves the following results on the evaluation set:

• Loss: 0.3881
• Wer: 0.2681

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: 1e-05
• 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: cosine
• lr_scheduler_warmup_steps: 500
• training_steps: 15000
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
0.36	0.6229	1000	0.4077	0.3387
0.2375	1.2457	2000	0.3526	0.2866
0.2368	1.8686	3000	0.3276	0.2774
0.1612	2.4914	4000	0.3237	0.2648
0.1107	3.1143	5000	0.3293	0.2733
0.1087	3.7372	6000	0.3294	0.2581
0.0649	4.3600	7000	0.3534	0.2645
0.0693	4.9829	8000	0.3526	0.2618
0.0383	5.6057	9000	0.3881	0.2681

Framework versions

• Transformers 4.40.0
• Pytorch 2.4.1+cu124
• Datasets 2.19.0
• Tokenizers 0.19.1