speaker-segmentation-fine-tuned-callhome-jpn

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

• Loss: 0.6518
• Der: 0.2003
• False Alarm: 0.0204
• Missed Detection: 0.0126
• Confusion: 0.1673

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: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: cosine
• num_epochs: 25

Training results

Training Loss	Epoch	Step	Validation Loss	Der	False Alarm	Missed Detection	Confusion
0.6084	1.0	157	0.6361	0.2140	0.0209	0.0101	0.1829
0.5157	2.0	314	0.6039	0.2079	0.0213	0.0101	0.1765
0.4718	3.0	471	0.6117	0.2094	0.0218	0.0101	0.1775
0.5069	4.0	628	0.6086	0.2129	0.0215	0.0101	0.1813
0.47	5.0	785	0.5974	0.2040	0.0215	0.0101	0.1724
0.4539	6.0	942	0.6047	0.2065	0.0219	0.0102	0.1745
0.4325	7.0	1099	0.5944	0.2009	0.0214	0.0104	0.1691
0.434	8.0	1256	0.6110	0.2059	0.0214	0.0105	0.1740
0.4199	9.0	1413	0.6045	0.2050	0.0212	0.0106	0.1733
0.4479	10.0	1570	0.6101	0.1990	0.0212	0.0105	0.1673
0.392	11.0	1727	0.6106	0.2003	0.0208	0.0107	0.1687
0.3858	12.0	1884	0.6279	0.2009	0.0211	0.0108	0.1689
0.3686	13.0	2041	0.6279	0.1976	0.0209	0.0114	0.1653
0.3963	14.0	2198	0.6263	0.1991	0.0211	0.0112	0.1668
0.3521	15.0	2355	0.6313	0.1970	0.0206	0.0116	0.1649
0.348	16.0	2512	0.6307	0.2001	0.0204	0.0123	0.1673
0.3668	17.0	2669	0.6425	0.2012	0.0206	0.0124	0.1682
0.3592	18.0	2826	0.6328	0.2001	0.0205	0.0124	0.1672
0.3485	19.0	2983	0.6489	0.2006	0.0202	0.0128	0.1675
0.3529	20.0	3140	0.6501	0.2007	0.0206	0.0123	0.1678
0.35	21.0	3297	0.6473	0.2003	0.0205	0.0124	0.1674
0.3549	22.0	3454	0.6518	0.2003	0.0205	0.0126	0.1672
0.3439	23.0	3611	0.6523	0.2002	0.0204	0.0127	0.1671
0.3495	24.0	3768	0.6527	0.2002	0.0204	0.0126	0.1672
0.34	25.0	3925	0.6518	0.2003	0.0204	0.0126	0.1673

Framework versions

• Transformers 4.42.4
• Pytorch 2.4.0+cu121
• Datasets 2.21.0
• Tokenizers 0.19.1