Shreyask09/speaker-segmentation-fine-tuned-callhome-jpn

speaker-segmentation-fine-tuned-hindi

This model is a fine-tuned version of pyannote/speaker-diarization-3.1 on the Samyak29/synthetic-speaker-diarization-dataset-hindi-large dataset. It achieves the following results on the evaluation set:

• Loss: 0.4284
• Model Preparation Time: 0.0095
• Der: 0.1417
• False Alarm: 0.0235
• Missed Detection: 0.0281
• Confusion: 0.0901

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: 0.001
• train_batch_size: 32
• eval_batch_size: 32
• seed: 42
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: cosine
• num_epochs: 5

Training results

Training Loss	Epoch	Step	Validation Loss	Model Preparation Time	Der	False Alarm	Missed Detection	Confusion
0.4708	1.0	194	0.4808	0.0095	0.1613	0.0255	0.0323	0.1035
0.388	2.0	388	0.4553	0.0095	0.1499	0.0225	0.0314	0.0960
0.3654	3.0	582	0.4368	0.0095	0.1433	0.0242	0.0278	0.0913
0.363	4.0	776	0.4296	0.0095	0.1410	0.0239	0.0279	0.0893
0.3388	5.0	970	0.4284	0.0095	0.1417	0.0235	0.0281	0.0901

Framework versions

• Transformers 4.44.2
• Pytorch 2.5.0+cu121
• Datasets 3.1.0
• Tokenizers 0.19.1