can i trained it with cpu?

ijkguo / mx-rcnn

Parallel Faster R-CNN implementation with MXNet.

Other

669 stars 292 forks source link

can i trained it with cpu? #124

Open shadow12138 opened 4 years ago

shadow12138 commented 4 years ago

how should i config to train it with cpu? the default option for "gpus" is 0, and the gpu environment is set in variable list "ctx" when i set ctx to [], it raises index out of bound error

@ijkguo @li-haoran

shadow12138 commented 4 years ago

i changed the Module's context parameter to mx.cpu() it started to run now, but it seemed like it's stucked. with a warning in bbox.py:

RuntimeWarning: invalid value encountered in log targets_dw = np.log(gt_widths / ex_widths) / box_stds[2]