Whisper Small Burmese v4

This model is a fine-tuned version of myatsu/whisper-small-burmese-v3 on the Myanmar Speech Dataset For ASR dataset. It achieves the following results on the evaluation set:

Loss: 0.1227
Wer: 79.3736
Cer: 42.8927

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: 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: 50
training_steps: 1000
mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer	Cer
0.2979	0.1797	100	0.1976	95.7134	53.4017
0.2608	0.3594	200	0.1669	85.1946	45.1336
0.1953	0.5391	300	0.1459	84.7042	45.3455
0.1938	0.7188	400	0.1380	82.5846	43.8820
0.1834	0.8985	500	0.1312	81.3350	43.9509
0.1265	1.0773	600	0.1317	80.5283	43.0450
0.1182	1.2570	700	0.1282	80.1171	43.0384
0.1178	1.4367	800	0.1262	80.1329	43.3006
0.1091	1.6164	900	0.1242	80.1645	43.3443
0.1013	1.7960	1000	0.1227	79.3736	42.8927