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

Loss: 0.4068
Model Preparation Time: 0.0039
Der: 0.1364
False Alarm: 0.0204
Missed Detection: 0.0118
Confusion: 0.1043

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: 0.0003
train_batch_size: 64
eval_batch_size: 64
seed: 100
optimizer: Use OptimizerNames.ADAMW_TORCH with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
lr_scheduler_type: cosine
lr_scheduler_warmup_ratio: 0.15
num_epochs: 10

Training results

Training Loss	Epoch	Step	Validation Loss	Model Preparation Time	Der	False Alarm	Missed Detection	Confusion
0.5533	1.0	400	0.5588	0.0039	0.1910	0.0248	0.0182	0.1480
0.4941	2.0	800	0.4830	0.0039	0.1649	0.0206	0.0150	0.1293
0.4496	3.0	1200	0.4526	0.0039	0.1540	0.0199	0.0137	0.1204
0.4371	4.0	1600	0.4332	0.0039	0.1474	0.0202	0.0129	0.1143
0.4173	5.0	2000	0.4234	0.0039	0.1444	0.0204	0.0121	0.1119
0.4072	6.0	2400	0.4115	0.0039	0.1400	0.0201	0.0123	0.1076
0.4007	7.0	2800	0.4112	0.0039	0.1383	0.0205	0.0117	0.1061
0.3945	8.0	3200	0.4072	0.0039	0.1378	0.0204	0.0117	0.1057
0.4086	9.0	3600	0.4061	0.0039	0.1364	0.0204	0.0118	0.1042
0.3951	10.0	4000	0.4068	0.0039	0.1364	0.0204	0.0118	0.1043

Framework versions

Transformers 4.48.3
Pytorch 2.5.1+cu124
Datasets 3.5.0
Tokenizers 0.21.0

whitneyten/pydiarize-Dataset-2_4-updated_04

speaker-segmentation-fine-tuned-datasetID-hugging_2_4_updated_04

Model description

Intended uses & limitations

Training and evaluation data

Training procedure

Training hyperparameters

Training results

Framework versions

whitneyten/
pydiarize-Dataset-2_4-updated_04