model.no1_expe.dia.1.A_data_ESLO_06.05.25

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

• Loss: 0.7445
• Model Preparation Time: 0.004
• Der: 0.4569
• False Alarm: 0.1475
• Missed Detection: 0.2178
• Confusion: 0.0916

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: 20

Training results

Training Loss	Epoch	Step	Validation Loss	Model Preparation Time	Der	False Alarm	Missed Detection	Confusion
0.8584	1.0	270	0.7939	0.004	0.4959	0.1382	0.2534	0.1042
0.8018	2.0	540	0.7538	0.004	0.4728	0.1353	0.2455	0.0919
0.7842	3.0	810	0.7547	0.004	0.4815	0.1657	0.2003	0.1155
0.7471	4.0	1080	0.7575	0.004	0.4750	0.1686	0.2022	0.1042
0.764	5.0	1350	0.7550	0.004	0.4673	0.1492	0.2237	0.0944
0.7684	6.0	1620	0.7544	0.004	0.4614	0.1459	0.2246	0.0910
0.7191	7.0	1890	0.7383	0.004	0.4587	0.1561	0.2133	0.0892
0.7204	8.0	2160	0.7347	0.004	0.4568	0.1504	0.2148	0.0916
0.7063	9.0	2430	0.7343	0.004	0.4586	0.1494	0.2188	0.0904
0.716	10.0	2700	0.7365	0.004	0.4622	0.1499	0.2203	0.0920
0.7117	11.0	2970	0.7410	0.004	0.4597	0.1437	0.2253	0.0907
0.7169	12.0	3240	0.7319	0.004	0.4526	0.1553	0.2039	0.0933
0.6566	13.0	3510	0.7381	0.004	0.4550	0.1518	0.2126	0.0907
0.6799	14.0	3780	0.7486	0.004	0.4564	0.1438	0.2230	0.0896
0.6755	15.0	4050	0.7425	0.004	0.4542	0.1500	0.2138	0.0904
0.6789	16.0	4320	0.7456	0.004	0.4572	0.1501	0.2140	0.0931
0.6793	17.0	4590	0.7425	0.004	0.4551	0.1467	0.2177	0.0907
0.672	18.0	4860	0.7445	0.004	0.4551	0.1479	0.2159	0.0914
0.69	19.0	5130	0.7442	0.004	0.4567	0.1473	0.2180	0.0914
0.7042	20.0	5400	0.7445	0.004	0.4569	0.1475	0.2178	0.0916

Framework versions

• Transformers 4.45.2
• Pytorch 2.4.1+cu121
• Datasets 3.0.1
• Tokenizers 0.20.0