search

vasgaowei / pytorch_MELM

The pytorch implementation of the Min-Entropy Latent Model for Weakly Supervised Object Detection
104 stars 19 forks source link

So many issues on PyTorch1.0 #26

Open yarkable opened 3 years ago

yarkable commented 3 years ago

I just clone the proj and switch it to pytorch1.0 branch. While compiling nms, lots of error occur, what astonish me is that nms needs libcudart.so.9.0…… Well, we are on CUDA10.0 bro.

After I've solved all the errors, it came with [1] 122287 segmentation fault (core dumped) python ./tools/demo.py --net vgg16 --dataset pascal_voc…… quit it bro.

vasgaowei commented 3 years ago

There is an error in demo.py. Just modifying from model.nms_wrapper import in demo.py to from ops.nms import nms corrects it. And after you compile nms successfully, is there any error when you train the model using ./experiments/scripts/train_faster_rcnn.sh 0 pascal_voc vgg16 ?

yarkable commented 3 years ago

I think there are still some error, cause I've found what cause the error is roi_ring_pool.py, line 34-35

roi_ring_pool_cuda.forward(ctx.pooled_height, ctx.pooled_width, ctx.spatial_scale,
                                                   features, ctx.processed_rois, output, ctx.argmax)

when the code runs here, a segmentation fault occurs.

vasgaowei commented 3 years ago

The pytorch1.0 branch supports PyTorch version 1.0~1.4. Higher versions may encounter errors.

yarkable commented 3 years ago

yep, I'm using torch1.1