lr0.0001_bs16_0613_1406

This model is a fine-tuned version of nvidia/mit-b0 on the greenkwd/upwellingdetection_SST dataset. It achieves the following results on the evaluation set:

• Loss: 0.2475
• Mean Iou: 0.5250
• Mean Accuracy: 0.8644
• Overall Accuracy: 0.8974
• Accuracy Land: nan
• Accuracy Upwelling: 0.9447
• Accuracy Not Upwelling: 0.7842
• Iou Land: 0.0
• Iou Upwelling: 0.8812
• Iou Not Upwelling: 0.6937
• Dice Macro: 0.9072
• Dice Micro: 0.9350

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: 0.0001
• 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: linear
• lr_scheduler_warmup_steps: 500
• num_epochs: 100

Training results

Training Loss	Epoch	Step	Validation Loss	Mean Iou	Mean Accuracy	Overall Accuracy	Accuracy Land	Accuracy Upwelling	Accuracy Not Upwelling	Iou Land	Iou Upwelling	Iou Not Upwelling	Dice Macro	Dice Micro
0.9564	10.0	20	1.0348	0.3708	0.7051	0.7459	nan	0.8044	0.6058	0.0	0.6974	0.4150	0.6938	0.7353
0.7616	20.0	40	0.8405	0.4019	0.7384	0.7810	nan	0.8422	0.6346	0.0	0.7402	0.4655	0.7633	0.8127
0.6029	30.0	60	0.5523	0.4551	0.7901	0.8384	nan	0.9077	0.6725	0.0	0.8092	0.5561	0.8294	0.8753
0.4617	40.0	80	0.4134	0.4867	0.8413	0.8640	nan	0.8966	0.7859	0.0	0.8289	0.6311	0.8586	0.8934
0.434	50.0	100	0.3439	0.4928	0.8247	0.8760	nan	0.9496	0.6998	0.0	0.8534	0.6250	0.8750	0.9132
0.3962	60.0	120	0.3170	0.5029	0.8547	0.8784	nan	0.9125	0.7970	0.0	0.8492	0.6593	0.8833	0.9145
0.3269	70.0	140	0.2858	0.5009	0.8345	0.8798	nan	0.9447	0.7243	0.0	0.8623	0.6404	0.8874	0.9220
0.3104	80.0	160	0.2693	0.5131	0.8535	0.8893	nan	0.9409	0.7661	0.0	0.8677	0.6717	0.8960	0.9267
0.2788	90.0	180	0.2553	0.5210	0.8605	0.8971	nan	0.9497	0.7713	0.0	0.8738	0.6891	0.8998	0.9293
0.3077	100.0	200	0.2475	0.5250	0.8644	0.8974	nan	0.9447	0.7842	0.0	0.8812	0.6937	0.9072	0.9350

Framework versions

• Transformers 4.41.2
• Pytorch 2.6.0+cu124
• Datasets 3.2.0
• Tokenizers 0.19.1