target_hold

This model is a fine-tuned version of facebook/detr-resnet-50 on the c14kevincardenas/beta_caller_284_target_hold dataset. It achieves the following results on the evaluation set:

• Loss: 0.8720
• Iou: 0.0008

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: 5e-05
• train_batch_size: 64
• eval_batch_size: 64
• seed: 2014
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• lr_scheduler_warmup_steps: 250
• num_epochs: 20.0
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Iou
1.2348	1.0	100	1.1666	0.0001
1.0043	2.0	200	0.9816	0.0023
0.9101	3.0	300	0.9058	0.0020
0.8846	4.0	400	0.8883	0.0013
0.8755	5.0	500	0.8819	0.0011
0.8714	6.0	600	0.8789	0.0010
0.8684	7.0	700	0.8773	0.0009
0.8664	8.0	800	0.8764	0.0008
0.8677	9.0	900	0.8752	0.0009
0.863	10.0	1000	0.8747	0.0009
0.8619	11.0	1100	0.8737	0.0009
0.8637	12.0	1200	0.8732	0.0009
0.8632	13.0	1300	0.8730	0.0009
0.8581	14.0	1400	0.8727	0.0009
0.8615	15.0	1500	0.8724	0.0009
0.8604	16.0	1600	0.8724	0.0008
0.8606	17.0	1700	0.8720	0.0009
0.8592	18.0	1800	0.8720	0.0009
0.8621	19.0	1900	0.8720	0.0008
0.8629	20.0	2000	0.8720	0.0008

Framework versions

• Transformers 4.45.2
• Pytorch 2.5.0+cu124
• Datasets 3.0.1
• Tokenizers 0.20.1