whisper-tiny-khmer-aug-v6

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

• Loss: 0.2583
• Wer: 65.9478

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
0.7761	0.9994	837	0.3819	88.5358
0.3222	2.0	1675	0.2888	79.8768
0.2456	2.9994	2512	0.2535	73.9419
0.2058	4.0	3350	0.2438	73.8447
0.1777	4.9994	4187	0.2391	69.8070
0.1577	6.0	5025	0.2397	68.3152
0.1421	6.9994	5862	0.2383	68.9801
0.1283	8.0	6700	0.2418	67.2612
0.1163	8.9994	7537	0.2475	69.7097
0.1052	9.9940	8370	0.2583	65.9478

Framework versions

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