kale
End of training
fb6ffa0 verified
metadata
library_name: transformers
language:
  - hi
license: mit
base_model: pyannote/speaker-diarization-3.1
tags:
  - speaker-diarization
  - speaker-segmentation
  - generated_from_trainer
datasets:
  - Shreyask09/synthetic-speaker-diarization-dataset-hindi
model-index:
  - name: speaker-segmentation-fine-tuned-hindi
    results: []

speaker-segmentation-fine-tuned-hindi

This model is a fine-tuned version of pyannote/speaker-diarization-3.1 on the Shreyask09/synthetic-speaker-diarization-dataset-hindi dataset. It achieves the following results on the evaluation set:

  • Loss: 0.3013
  • Model Preparation Time: 0.004
  • Der: 0.1018
  • False Alarm: 0.0131
  • Missed Detection: 0.0241
  • Confusion: 0.0646

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: Use adamw_torch with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
  • lr_scheduler_type: cosine
  • num_epochs: 5

Training results

Training Loss Epoch Step Validation Loss Model Preparation Time Der False Alarm Missed Detection Confusion
0.3506 1.0 219 0.3386 0.004 0.1165 0.0156 0.0288 0.0722
0.2898 2.0 438 0.3218 0.004 0.1080 0.0131 0.0267 0.0682
0.2479 3.0 657 0.3004 0.004 0.1034 0.0134 0.0245 0.0655
0.2413 4.0 876 0.3027 0.004 0.1020 0.0129 0.0244 0.0647
0.2506 5.0 1095 0.3013 0.004 0.1018 0.0131 0.0241 0.0646

Framework versions

  • Transformers 4.47.1
  • Pytorch 2.5.1+cu121
  • Datasets 3.2.0
  • Tokenizers 0.21.0