pranetk/whisper-large-v3-hi

Whisper Large V3 Hi - Paraspeak

This model is a fine-tuned version of openai/whisper-large-v3 on the Paraspeak Dataset 1.0 dataset. It achieves the following results on the evaluation set:

• Loss: 0.7832
• Wer: 69.2308

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: 2
• eval_batch_size: 8
• seed: 42
• gradient_accumulation_steps: 8
• 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: 10
• training_steps: 200
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
0.5306	1.7094	25	1.0175	100.0
0.2893	3.4188	50	0.8659	84.6154
0.2662	5.1282	75	0.8183	92.3077
0.0551	6.8376	100	1.1448	100.0
0.018	8.5470	125	1.1831	76.9231
0.0028	10.2564	150	0.7150	69.2308
0.0481	11.9658	175	0.9248	69.2308
0.0002	13.6752	200	0.7832	69.2308

Framework versions

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