speaker-segmentation-fine-tuned-hindi-v3

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

• Loss: 0.3447
• Model Preparation Time: 0.007
• Der: 0.0985
• False Alarm: 0.0375
• Missed Detection: 0.0235
• Confusion: 0.0375

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: 8
• eval_batch_size: 8
• 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
No log	1.0	24	0.4600	0.007	0.1443	0.0349	0.0256	0.0837
0.5196	2.0	48	0.3562	0.007	0.1304	0.0325	0.0242	0.0737
0.306	3.0	72	0.3732	0.007	0.1251	0.0402	0.0253	0.0596
0.2116	4.0	96	0.3712	0.007	0.1265	0.0408	0.0242	0.0615
0.1944	5.0	120	0.3846	0.007	0.1223	0.0337	0.0260	0.0627
0.1538	6.0	144	0.3544	0.007	0.1191	0.0375	0.0228	0.0587
0.1417	7.0	168	0.4045	0.007	0.1213	0.0358	0.0241	0.0614
0.1122	8.0	192	0.4213	0.007	0.1267	0.0438	0.0228	0.0601
0.1053	9.0	216	0.4171	0.007	0.1178	0.0368	0.0255	0.0555
0.0897	10.0	240	0.3561	0.007	0.1142	0.0409	0.0228	0.0505
0.1043	11.0	264	0.3738	0.007	0.1122	0.0380	0.0248	0.0495
0.0825	12.0	288	0.3383	0.007	0.1025	0.0377	0.0237	0.0411
0.0894	13.0	312	0.3328	0.007	0.0995	0.0388	0.0237	0.0370
0.0699	14.0	336	0.3272	0.007	0.0988	0.0376	0.0237	0.0375
0.0785	15.0	360	0.3374	0.007	0.0991	0.0378	0.0235	0.0378
0.0759	16.0	384	0.3414	0.007	0.0978	0.0383	0.0233	0.0362
0.0653	17.0	408	0.3417	0.007	0.0973	0.0375	0.0234	0.0364
0.0726	18.0	432	0.3439	0.007	0.0981	0.0374	0.0236	0.0370
0.0684	19.0	456	0.3445	0.007	0.0984	0.0374	0.0235	0.0375
0.0731	20.0	480	0.3447	0.007	0.0985	0.0375	0.0235	0.0375

Framework versions

• Transformers 4.52.4
• Pytorch 2.6.0+cu124
• Datasets 3.6.0
• Tokenizers 0.21.1