objects76/ft-ja4-2.25sec-250513_1443

ft-ja4-2.25sec

This model is a fine-tuned version of pyannote/segmentation-3.0 on the objects76/synthetic-ja4-speaker-overlap-6400 dataset. It achieves the following results on the evaluation set:

• Loss: 0.3024
• Der: 0.0934
• False Alarm: 0.0422
• Missed Detection: 0.0401
• Confusion: 0.0111

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: 2048
• eval_batch_size: 2048
• 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: 200

Training results

Training Loss	Epoch	Step	Confusion	Der	False Alarm	Validation Loss	Missed Detection
No log	1.0	7	0.0611	0.2939	0.1347	0.9642	0.0982
No log	2.0	14	0.0771	0.2589	0.0527	0.8482	0.1291
No log	3.0	21	0.0839	0.2499	0.0303	0.8020	0.1358
0.8969	4.0	28	0.0664	0.2162	0.0686	0.7508	0.0813
0.8969	5.0	35	0.0654	0.2029	0.0680	0.6863	0.0695
0.8969	6.0	42	0.0615	0.1919	0.0725	0.6313	0.0580
0.8969	7.0	49	0.0535	0.1808	0.0752	0.5859	0.0522
0.6817	8.0	56	0.0468	0.1698	0.0708	0.5505	0.0522
0.6817	9.0	63	0.0415	0.1619	0.0694	0.5233	0.0510
0.6817	10.0	70	0.0366	0.1550	0.0688	0.4992	0.0496
0.5316	11.0	77	0.0316	0.1459	0.0634	0.4725	0.0509
0.5316	12.0	84	0.0271	0.1389	0.0608	0.4490	0.0510
0.5316	13.0	91	0.0250	0.1322	0.0556	0.4286	0.0517
0.5316	14.0	98	0.0220	0.1263	0.0544	0.4123	0.0499
0.4403	15.0	105	0.0203	0.1213	0.0523	0.3977	0.0487
0.4403	16.0	112	0.0191	0.1190	0.0536	0.3904	0.0462
0.4403	17.0	119	0.0188	0.1161	0.0484	0.3784	0.0489
0.3873	18.0	126	0.0171	0.1144	0.0520	0.3725	0.0453
0.3873	19.0	133	0.0183	0.1128	0.0458	0.3659	0.0487
0.3873	20.0	140	0.0177	0.1129	0.0516	0.3636	0.0437
0.3873	21.0	147	0.0171	0.1099	0.0470	0.3569	0.0458
0.3577	22.0	154	0.0178	0.1097	0.0451	0.3541	0.0468
0.3577	23.0	161	0.0164	0.1076	0.0495	0.3487	0.0417
0.3577	24.0	168	0.0165	0.1066	0.0459	0.3436	0.0443
0.3417	25.0	175	0.0166	0.1059	0.0446	0.3385	0.0447
0.3417	26.0	182	0.0155	0.1052	0.0487	0.3379	0.0411
0.3417	27.0	189	0.0155	0.1043	0.0443	0.3352	0.0445
0.3417	28.0	196	0.0157	0.1050	0.0464	0.3352	0.0429
0.3257	29.0	203	0.0151	0.1040	0.0474	0.3327	0.0415
0.3257	30.0	210	0.0150	0.1014	0.0419	0.3270	0.0445
0.3257	31.0	217	0.0133	0.1000	0.0451	0.3240	0.0417
0.3257	32.0	224	0.3204	0.0989	0.0425	0.0431	0.0133
0.3111	33.0	231	0.3191	0.0995	0.0439	0.0424	0.0133
0.3111	34.0	238	0.3159	0.0989	0.0431	0.0424	0.0134
0.3111	35.0	245	0.3153	0.0980	0.0433	0.0417	0.0130
0.3077	36.0	252	0.3133	0.0967	0.0409	0.0429	0.0128
0.3077	37.0	259	0.3145	0.0975	0.0418	0.0430	0.0127
0.3077	38.0	266	0.3164	0.0981	0.0417	0.0436	0.0128
0.3077	39.0	273	0.3145	0.0978	0.0402	0.0446	0.0130
0.3047	40.0	280	0.3138	0.0974	0.0441	0.0406	0.0128
0.3047	41.0	287	0.3182	0.0979	0.0398	0.0450	0.0131
0.3047	42.0	294	0.3117	0.0955	0.0393	0.0434	0.0128
0.3023	43.0	301	0.3078	0.0953	0.0408	0.0421	0.0124
0.3023	44.0	308	0.3103	0.0969	0.0409	0.0435	0.0126
0.3023	45.0	315	0.3067	0.0949	0.0379	0.0447	0.0124
0.3023	46.0	322	0.3076	0.0950	0.0434	0.0395	0.0121
0.295	47.0	329	0.3035	0.0937	0.0384	0.0430	0.0123
0.295	48.0	336	0.3054	0.0954	0.0411	0.0419	0.0124
0.295	49.0	343	0.3038	0.0937	0.0391	0.0428	0.0118
0.2845	50.0	350	0.3048	0.0942	0.0409	0.0415	0.0118
0.2845	51.0	357	0.3039	0.0942	0.0388	0.0435	0.0119
0.2845	52.0	364	0.3029	0.0935	0.0403	0.0416	0.0116
0.2845	53.0	371	0.2978	0.0922	0.0389	0.0418	0.0115
0.2796	54.0	378	0.2995	0.0933	0.0422	0.0399	0.0111
0.2796	55.0	385	0.2984	0.0922	0.0358	0.0448	0.0115
0.2796	56.0	392	0.3003	0.0933	0.0423	0.0390	0.0120
0.2796	57.0	399	0.3000	0.0933	0.0393	0.0417	0.0123
0.2768	58.0	406	0.3037	0.0942	0.0428	0.0395	0.0120
0.2768	59.0	413	0.2987	0.0921	0.0366	0.0438	0.0117
0.2768	60.0	420	0.3036	0.0946	0.0472	0.0362	0.0113
0.2774	61.0	427	0.3010	0.0929	0.0384	0.0431	0.0114
0.2774	62.0	434	0.3014	0.0924	0.0401	0.0409	0.0114
0.2774	63.0	441	0.3024	0.0934	0.0422	0.0401	0.0111

Framework versions

• Transformers 4.51.3
• Pytorch 2.6.0+cu124
• Datasets 3.5.0
• Tokenizers 0.21.1