whisper-multi-diar-wer

<!-- This model card has been generated automatically according to the information the Trainer had access to. You should probably proofread and complete it, then remove this comment. -->

whisper-multi-diar-wer

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: 10.0666
• Wer: 582.8864
• Cer: 168.6522
• Speech Scored: 693831
• Speech Miss: 52252
• Speech Falarm: 117030
• Speaker Miss: 52252
• Speaker Falarm: 117030
• Speaker Error: 187216
• Speaker Correct: 1437.5240
• Diarization Error: 356498
• Frames: 600
• Speaker Wide Frames: 746083
• Speech Scored Ratio: 1156.385
• Speech Miss Ratio: 87.0867
• Speech Falarm Ratio: 195.05
• Speaker Correct Ratio: 2.3959
• Speaker Miss Ratio: 0.0700
• Speaker Falarm Ratio: 0.1569
• Speaker Error Ratio: 0.2509
• Diarization Error Ratio: 0.4778

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: 24
• eval_batch_size: 32
• seed: 42
• gradient_accumulation_steps: 2
• total_train_batch_size: 48
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• lr_scheduler_warmup_ratio: 0.05
• num_epochs: 10
• mixed_precision_training: Native AMP

Training results

Framework versions

• Transformers 4.36.2
• Pytorch 2.0.0
• Datasets 2.16.1
• Tokenizers 0.15.0