segformer-b1-finetuned-segments-pv_v1_normalized_t4_16batch

This model is a fine-tuned version of nvidia/segformer-b3-finetuned-ade-512-512 on the mouadenna/satellite_PV_dataset_train_test_v1 dataset. It achieves the following results on the evaluation set:

• Loss: nan
• Mean Iou: 0.0
• Precision: 1.0

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: 0.0004
• train_batch_size: 4
• eval_batch_size: 4
• seed: 42
• gradient_accumulation_steps: 4
• total_train_batch_size: 16
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• num_epochs: 40
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Mean Iou	Precision
0.0056	0.9989	229	0.0099	0.7568	0.7827
0.0092	1.9978	458	0.0113	0.7247	0.7360
0.0069	2.9967	687	0.0051	0.8440	0.9106
0.0159	4.0	917	0.0057	0.8317	0.8635
0.0124	4.9989	1146	0.0099	0.8223	0.9292
0.0033	5.9978	1375	0.2415	0.0	1.0
0.0	6.9967	1604	nan	0.0	1.0
0.0	8.0	1834	nan	0.0	1.0
0.0	8.9989	2063	nan	0.0	1.0
0.0	9.9978	2292	nan	0.0	1.0
0.0	10.9967	2521	nan	0.0	1.0
0.0	12.0	2751	nan	0.0	1.0
0.0	12.9989	2980	nan	0.0	1.0
0.0	13.9978	3209	nan	0.0	1.0
0.0	14.9967	3438	nan	0.0	1.0
0.0	16.0	3668	nan	0.0	1.0
0.0	16.9989	3897	nan	0.0	1.0
0.0	17.9978	4126	nan	0.0	1.0
0.0	18.9967	4355	nan	0.0	1.0
0.0	20.0	4585	nan	0.0	1.0
0.0	20.9989	4814	nan	0.0	1.0
0.0	21.9978	5043	nan	0.0	1.0
0.0	22.9967	5272	nan	0.0	1.0
0.0	24.0	5502	nan	0.0	1.0
0.0	24.9989	5731	nan	0.0	1.0
0.0	25.9978	5960	nan	0.0	1.0
0.0	26.9967	6189	nan	0.0	1.0
0.0	28.0	6419	nan	0.0	1.0
0.0	28.9989	6648	nan	0.0	1.0
0.0	29.9978	6877	nan	0.0	1.0
0.0	30.9967	7106	nan	0.0	1.0
0.0	32.0	7336	nan	0.0	1.0
0.0	32.9989	7565	nan	0.0	1.0
0.0	33.9978	7794	nan	0.0	1.0
0.0	34.9967	8023	nan	0.0	1.0
0.0	36.0	8253	nan	0.0	1.0
0.0	36.9989	8482	nan	0.0	1.0
0.0	37.9978	8711	nan	0.0	1.0
0.0	38.9967	8940	nan	0.0	1.0
0.0	39.9564	9160	nan	0.0	1.0

Framework versions

• Transformers 4.42.3
• Pytorch 2.1.2
• Datasets 2.20.0
• Tokenizers 0.19.1