easonliao/cudatree

Building random forests on GPU

cudatree's Issues

As I pip install the library I encountered with the following error?

As I pip install the library I encountered with the following error?

libcurand.so.5.5: cannot open shared object file: No such file or directory

How can I solve it, any idea?

I installed pycuda by the way, it works...

TypeError: 'numpy.uint32' does not have the buffer interface

Install from pip / source code all give this error:

driver.memcpy_htod(const_stride, np.uint32(self.stride))

when calling

forest.fit(x_train, y_train, bfs_threshold = 4196) # from the example in readme

Running ubuntu 14.04,64bits, nvidia-331

Save Classifier

Hi so is there any way to train a classifier and then save it? I might just be missing something obvious but currently I try to use cPickle and dump the trained classifier to a file. When I reload the classifier, however, it does not have the kernels initialized. I saw in the latest commit that kernels were not picklable but is there some other way to store the classifier? Having to retrain all the time is quite time consuming.

Thanks!