whisper-tiny-myanmar-phase2

This model is a fine-tuned version of thantzinphyo/whisper-tiny-myanmar-phase1 on the None dataset. It achieves the following results on the evaluation set:

• Loss: 1.7102
• Cer: 0.1999
• Wer: 0.6554

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: 5e-05
• train_batch_size: 32
• eval_batch_size: 32
• seed: 42
• gradient_accumulation_steps: 4
• total_train_batch_size: 128
• optimizer: Use OptimizerNames.ADAMW_TORCH_FUSED with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
• lr_scheduler_type: cosine
• lr_scheduler_warmup_steps: 50
• training_steps: 800
• mixed_precision_training: Native AMP
• label_smoothing_factor: 0.1

Training results

Training Loss	Epoch	Step	Validation Loss	Cer	Wer
11.3014	2.7778	50	1.6821	0.2024	0.6708
11.3240	5.5556	100	1.6546	0.2074	0.6594
11.3136	8.3333	150	1.6573	0.2069	0.6576
11.3026	11.1111	200	1.6722	0.2032	0.6506
11.2992	13.8889	250	1.6777	0.2022	0.6567
11.3006	16.6667	300	1.6764	0.1999	0.6475
11.2987	19.4444	350	1.6761	0.2059	0.6607
11.2959	22.2222	400	1.6833	0.2034	0.6607
11.2916	25.0	450	1.6897	0.1933	0.6489
11.2902	27.7778	500	1.6973	0.1986	0.655
11.2897	30.5556	550	1.7026	0.1977	0.6559
11.2893	33.3333	600	1.7064	0.1984	0.6572
11.2892	36.1111	650	1.7086	0.2002	0.6567
11.2891	38.8889	700	1.7097	0.2006	0.6567
11.2890	41.6667	750	1.7101	0.2	0.6554
11.2890	44.4444	800	1.7102	0.1999	0.6554

Framework versions

• Transformers 5.0.0
• Pytorch 2.10.0+cu128
• Datasets 5.0.0
• Tokenizers 0.22.2