CoolFace
Modelpublic

amirhosein-prdv/speaker-segmentation-fine-tuned-synthetic_v3

sourceHugging Facemitupdated 15d agoView on Hugging Face
0likes372downloads
Model Card

<!-- 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. -->

speaker-segmentation-fine-tuned-synthetic_v3

This model is a fine-tuned version of pyannote/segmentation-3.0 on the uncleMehrzad/synthetic-speaker-diarization-dataset-fa-large-3000 dataset. It achieves the following results on the evaluation set:

  • —Loss: 0.5584
  • —Model Preparation Time: 0.0027
  • —Der: 0.1661
  • —False Alarm: 0.0517
  • —Missed Detection: 0.0375
  • —Confusion: 0.0769

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
  • —trainbatchsize: 32
  • —evalbatchsize: 32
  • —seed: 42
  • —optimizer: Use adamwtorchfused with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
  • —lrschedulertype: cosine
  • —num_epochs: 50.0

Training results

Training LossEpochStepValidation LossModel Preparation TimeDerFalse AlarmMissed DetectionConfusion
No log000.91410.00270.27510.06960.06580.1397
0.58211.06010.58910.00270.20840.06430.03580.1083
0.54362.012020.55750.00270.19300.05580.04070.0965
0.52533.018030.55230.00270.19160.06020.03300.0984
0.50344.024040.53010.00270.18740.05900.03450.0939
0.48975.030050.51730.00270.18050.05740.03270.0904
0.47946.036060.51570.00270.17860.05710.03380.0877
0.47027.042070.51470.00270.17890.05330.03640.0892
0.46878.048080.51210.00270.17520.05580.03210.0873
0.44499.054090.49880.00270.17340.05770.03110.0846
0.447410.060100.50400.00270.17430.05800.03060.0858
0.439811.066110.49250.00270.17080.05530.03280.0827
0.432912.072120.49030.00270.16800.05570.03140.0809
0.427913.078130.49390.00270.16850.05150.03650.0805
0.421514.084140.49240.00270.16830.05410.03290.0813
0.411215.090150.50180.00270.17020.05480.03280.0826
0.404716.096160.49020.00270.16560.05270.03330.0797
0.381517.0102170.50500.00270.16770.05480.03260.0803
0.388218.0108180.49300.00270.16660.05340.03360.0796
0.382319.0114190.50580.00270.16780.05580.03040.0815
0.378220.0120200.49600.00270.16850.05530.03390.0794
0.375321.0126210.49600.00270.16310.05370.03250.0769
0.360822.0132220.49530.00270.16550.05400.03410.0774
0.359623.0138230.50330.00270.16650.05160.03720.0777
0.352224.0144240.49580.00270.16590.05400.03490.0770
0.344525.0150250.51210.00270.16520.05060.03630.0783
0.335226.0156260.51760.00270.16770.05150.03670.0795
0.337227.0162270.52840.00270.16540.05410.03220.0791
0.331828.0168280.51840.00270.16590.05320.03450.0783
0.327729.0174290.52740.00270.16680.05160.03590.0792
0.323530.0180300.52530.00270.16650.05330.03540.0779
0.310931.0186310.52990.00270.16550.05060.03790.0770
0.313432.0192320.53730.00270.16580.05180.03700.0771
0.3133.0198330.54110.00270.16640.05320.03480.0784
0.305534.0204340.53500.00270.16540.05320.03690.0753
0.301435.0210350.53860.00270.16600.05230.03720.0765
0.298136.0216360.55040.00270.16570.05140.03760.0767
0.303837.0222370.54470.00270.16490.05270.03550.0767
0.296938.0228380.54490.00270.16470.05040.03820.0761
0.289539.0234390.55040.00270.16600.05160.03760.0768
0.295440.0240400.54860.00270.16620.05160.03750.0772
0.2941.0246410.55030.00270.16620.05180.03790.0765
0.29142.0252420.54990.00270.16670.05160.03790.0771
0.292743.0258430.55030.00270.16610.05270.03700.0764
0.292544.0264440.55310.00270.16610.05110.03820.0768
0.285845.0270450.55460.00270.16630.05200.03710.0771
0.284846.0276460.55660.00270.16680.05150.03780.0775
0.289447.0282470.55780.00270.16630.05190.03730.0771
0.284248.0288480.55800.00270.16630.05180.03750.0770
0.282149.0294490.55840.00270.16620.05170.03750.0770
0.281650.0300500.55840.00270.16610.05170.03750.0769

Framework versions

  • —Transformers 4.57.6
  • —Pytorch 2.14.0+cu126
  • —Datasets 5.0.1
  • —Tokenizers 0.22.2