synthetic-speaker-jpn

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

• Loss: 0.3546
• Model Preparation Time: 0.0018
• Der: 0.1098
• False Alarm: 0.0178
• Missed Detection: 0.0198
• Confusion: 0.0722

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: 64
• eval_batch_size: 64
• seed: 42
• 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
• num_epochs: 20

Training results

Training Loss	Epoch	Step	Validation Loss	Model Preparation Time	Der	False Alarm	Missed Detection	Confusion
0.4222	1.0	198	0.4012	0.0018	0.1316	0.0195	0.0240	0.0880
0.3767	2.0	396	0.3893	0.0018	0.1267	0.0176	0.0237	0.0854
0.3784	3.0	594	0.3935	0.0018	0.1233	0.0172	0.0232	0.0829
0.3596	4.0	792	0.3747	0.0018	0.1216	0.0192	0.0204	0.0820
0.352	5.0	990	0.3807	0.0018	0.1231	0.0184	0.0207	0.0840
0.3111	6.0	1188	0.3585	0.0018	0.1134	0.0183	0.0203	0.0748
0.3139	7.0	1386	0.3460	0.0018	0.1123	0.0181	0.0202	0.0740
0.3176	8.0	1584	0.3610	0.0018	0.1134	0.0184	0.0198	0.0752
0.3142	9.0	1782	0.3542	0.0018	0.1127	0.0172	0.0211	0.0745
0.2834	10.0	1980	0.3485	0.0018	0.1116	0.0178	0.0201	0.0737
0.2875	11.0	2178	0.3537	0.0018	0.1095	0.0174	0.0204	0.0717
0.2704	12.0	2376	0.3582	0.0018	0.1111	0.0177	0.0201	0.0733
0.2802	13.0	2574	0.3589	0.0018	0.1106	0.0177	0.0200	0.0728
0.2577	14.0	2772	0.3547	0.0018	0.1102	0.0180	0.0198	0.0725
0.261	15.0	2970	0.3511	0.0018	0.1086	0.0181	0.0196	0.0709
0.2647	16.0	3168	0.3544	0.0018	0.1096	0.0182	0.0194	0.0719
0.2554	17.0	3366	0.3537	0.0018	0.1093	0.0174	0.0202	0.0717
0.2624	18.0	3564	0.3547	0.0018	0.1095	0.0178	0.0199	0.0718
0.2667	19.0	3762	0.3542	0.0018	0.1098	0.0178	0.0198	0.0722
0.2613	20.0	3960	0.3546	0.0018	0.1098	0.0178	0.0198	0.0722

Framework versions

• Transformers 4.50.3
• Pytorch 2.6.0+cu124
• Datasets 3.5.0
• Tokenizers 0.21.1