finetuning

[junfu1115]

finetuning

[yiqigo0215]

Hi, did you train the model with single gpu successively

[sbharadwajj]

This is what I'm running from my terminal
CUDA_VISIBLE_DEVICES=0,1 python train.py --dataset balloons --model danet --backbone resnet50 --checkname danet101 --base-size 512 --crop-size 384 --epochs 240 --batch-size 8 --lr 0.003 --ft --ft-resume /danet_model/danet101/model_best.pth.tar --pre-class 1 --workers 2 --multi-grid --multi-dilation 4 8 16

My dataset has only one class, i.e balloons. To load the dataset, I edited the cityscapes.py and modified it to accept a single class.

But I am not sure as to which weights i have to use to fine-tune the model.