Help needed with running DiMP on CPU

[orilifs]

Taken from issue #128:

Right now, there is no way to run the PrROIPool layer on CPU that I know of. However, you can run the tracker without the IoUNet part (although it wont be nearly as good).

Try for example using this ATOM parameter file or set params.use_iou_net = False in any DiMP parameter file.

Hello @martin-danelljan I've tried what you suggested: I added this line to my super_dimp_no_iou_net.py params file: params.use_iou_net = False I also modified this line: params.use_gpu = False

But I am still getting this error:

File "../ltr/external/PreciseRoIPooling/pytorch/prroi_pool/prroi_pool.py", line 28, in forward return prroi_pool2d(features, rois, self.pooled_height, self.pooled_width, self.spatial_scale) File "../ltr/external/PreciseRoIPooling/pytorch/prroi_pool/functional.py", line 63, in forward raise NotImplementedError('Precise RoI Pooling only supports GPU (cuda) implememtations.') NotImplementedError: Precise RoI Pooling only supports GPU (cuda) implememtations.

I even went so far as trying to use the ltr/admin/loading/load_network and making and tried to make all the relevant changes along the dimp config to take a FeatureParams instead of a "net" object but it was a rabbit hole that I did not come out of :)

Any thoughts/suggestions please?