jethrowang/whisper-tiny_tat-esc_exp_nr_0.5_embed

Whisper Tiny Taiwanese (exp_nr_0.5_embed)

This model is a fine-tuned version of openai/whisper-tiny on the TAT ASR Aligned dataset. It achieves the following results on the evaluation set:

• Loss: 1.3660
• Cer: 32.6917

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: 0.0001
• train_batch_size: 64
• eval_batch_size: 32
• seed: 42
• optimizer: Use adamw_torch with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
• lr_scheduler_type: linear
• lr_scheduler_warmup_steps: 681
• training_steps: 6810
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Cer
0.4831	0.9985	681	0.9359	46.4492
0.2884	1.9971	1362	0.8960	35.3926
0.1794	2.9956	2043	0.9427	38.7004
0.1035	3.9941	2724	1.0423	35.5451
0.0585	4.9927	3405	1.1355	35.1530
0.0287	5.9912	4086	1.2120	35.6566
0.0146	6.9897	4767	1.2845	33.8159
0.0052	7.9883	5448	1.3243	33.4713
0.0018	8.9868	6129	1.3504	32.9579
0.0008	9.9853	6810	1.3660	32.6917

Framework versions

• Transformers 4.49.0
• Pytorch 2.0.0.post304
• Datasets 3.3.2
• Tokenizers 0.21.0