sit-justin/whisper-small-test-c2c

Whisper Small Chinese MOE Response

This model is a fine-tuned version of sit-justin/whisper-small-test on the MOE Response Chinese dataset. It achieves the following results on the evaluation set:

• Loss: 0.0158
• Cer: 2.5487

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: 8
• eval_batch_size: 8
• seed: 42
• gradient_accumulation_steps: 2
• total_train_batch_size: 16
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• lr_scheduler_warmup_steps: 250
• training_steps: 2000
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Cer
0.0804	0.5405	200	0.0771	5.5753
0.0695	1.0811	400	0.0632	5.4650
0.0374	1.6216	600	0.0467	4.6808
0.0228	2.1622	800	0.0439	4.9994
0.0142	2.7027	1000	0.0339	3.4310
0.0068	3.2432	1200	0.0257	5.1587
0.005	3.7838	1400	0.0216	2.7815
0.0019	4.3243	1600	0.0176	2.3772
0.0023	4.8649	1800	0.0158	2.5242
0.0012	5.4054	2000	0.0158	2.5487

Framework versions

• Transformers 4.42.3
• Pytorch 2.3.1+cu121
• Datasets 2.20.0
• Tokenizers 0.19.1