edmundchan70/Cantonese_Whisper_finetune

Whisper Small fine tune-Edmund-0818

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

• Loss: 0.1966
• Wer: 30.4762

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: 1.25e-05
• train_batch_size: 16
• eval_batch_size: 8
• seed: 42
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• lr_scheduler_warmup_ratio: 0.1
• num_epochs: 15
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
No log	1.0	156	0.1196	17.1429
No log	2.0	312	0.1553	24.6032
No log	3.0	468	0.1655	26.5079
0.0806	4.0	624	0.1820	29.5238
0.0806	5.0	780	0.1792	30.1587
0.0806	6.0	936	0.1998	31.5873
0.0131	7.0	1092	0.1954	31.2698
0.0131	8.0	1248	0.1923	30.6349
0.0131	9.0	1404	0.1905	31.2698
0.0016	10.0	1560	0.1954	31.2698
0.0016	11.0	1716	0.1931	31.1111
0.0016	12.0	1872	0.1953	30.4762
0.0005	13.0	2028	0.1960	30.6349
0.0005	14.0	2184	0.1964	30.6349
0.0005	15.0	2340	0.1966	30.4762

Framework versions

• Transformers 4.44.0
• Pytorch 2.4.0+cu124
• Datasets 2.21.0
• Tokenizers 0.19.1