search

xiaogangw-zz / cascaded-point-completion

59 stars 6 forks source link

No OpKernel was registered to support Op 'FarthestPointSample' #10

Open Rachel-Green opened 3 years ago

Rachel-Green commented 3 years ago

Hello, I meet a problem that I can't solve. I have the same environment (tf-gpu 1.13.1, cuda 10.0) as yours and the tf-gpu is fine with gpu (I have tested with tf.test.is_gpu_available() and tf.test.is_built_with_cuda() ) but I can't use the function farthest_point_sample(), and the error info as follows:

InvalidArgumentError (see above for traceback): No OpKernel was registered to support Op 'FarthestPointSample' used by node gpu_0/generator/decoder/FarthestPointSample (defined at :148) with these attrs: [npoint=256] Registered devices: [CPU, XLA_CPU] Registered kernels: device='GPU' [[node gpu_0/generator/decoder/FarthestPointSample (defined at :148) ]]

hope somebody could help me, it's very important to me, thanks!

xiaogangw commented 3 years ago

@Rachel-Green @xiaogangw-zz Hi, can you please double-check your environment paths or your gpu usage during training, or recompile it?