whisper-large-v3-cantonese-hk-v3 API & Inference Endpoint

More information needed

More information needed

More information needed

The following hyperparameters were used during training:

learning_rate: 1e-05
train_batch_size: 4
eval_batch_size: 4
seed: 42
gradient_accumulation_steps: 4
total_train_batch_size: 16
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: linear
lr_scheduler_warmup_steps: 100
training_steps: 3000
mixed_precision_training: Native AMP

Table with columns: Training Loss, Epoch, Step, Validation Loss, Cer
Training Loss	Epoch	Step	Validation Loss	Cer
1.2215	0.5003	500	0.2381	0.1352
1.3911	1.0	1000	0.2229	0.1063
0.7399	1.5003	1500	0.2261	0.1101
0.7151	2.0	2000	0.2189	0.0904
0.4398	2.5003	2500	0.2260	0.0865
0.4779	3.0	3000	0.2230	0.0799

More information needed

More information needed

More information needed

The following hyperparameters were used during training:

learning_rate: 1e-05
train_batch_size: 4
eval_batch_size: 4
seed: 42
gradient_accumulation_steps: 4
total_train_batch_size: 16
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: linear
lr_scheduler_warmup_steps: 100
training_steps: 3000
mixed_precision_training: Native AMP

Table with columns: Training Loss, Epoch, Step, Validation Loss, Cer
Training Loss	Epoch	Step	Validation Loss	Cer
1.2215	0.5003	500	0.2381	0.1352
1.3911	1.0	1000	0.2229	0.1063
0.7399	1.5003	1500	0.2261	0.1101
0.7151	2.0	2000	0.2189	0.0904
0.4398	2.5003	2500	0.2260	0.0865
0.4779	3.0	3000	0.2230	0.0799

whisper-large-v3-cantonese-hk-v3