whisper-small-khmer-aug-v6-2

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

• Loss: 0.3497
• Wer: 68.6233

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: 16
• eval_batch_size: 16
• seed: 42
• gradient_accumulation_steps: 2
• total_train_batch_size: 32
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: constant
• lr_scheduler_warmup_steps: 1000
• num_epochs: 10

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
0.8376	0.9994	837	0.4499	100.4702
0.3482	2.0	1675	0.3490	79.3903
0.2732	2.9994	2512	0.3141	74.6230
0.231	4.0	3350	0.3190	75.0608
0.2002	4.9994	4187	0.3118	72.5799
0.1743	6.0	5025	0.3104	72.2556
0.1553	6.9994	5862	0.3216	71.2826
0.1375	8.0	6700	0.3307	73.7311
0.1217	8.9994	7537	0.3497	69.3854
0.1089	9.9940	8370	0.3497	68.6233

Framework versions

• Transformers 4.44.0
• Pytorch 2.3.1
• Datasets 2.21.0
• Tokenizers 0.19.1