Python 3 support for model prediction (+ misc) #98
Conversation
coremlpython/CoreMLPythonUtils.mm
Outdated
| #define PyBytes_Check(name) PyString_Check(name) | ||
| #define PyAnyInteger_Check(name) (PyLong_Check(name) || PyInt_Check(name)) | ||
| #else | ||
| #define PyAnyInteger_Check(name) (PyLong_Check(name) || (_import_array(), PyArray_IsScalar(name, Integer))) |
There was a problem hiding this comment.
There is probably a better location where I can place _import_array(), but I wasn't sure of an appropriate entry point. This should be very fast after the first import though.
coremlpython/CoreMLPythonUtils.mm
Outdated
| @@ -4,6 +4,15 @@ | |||
| #include <pybind11/eval.h> | |||
| #include <pybind11/numpy.h> | |||
|
|
|||
| #include <numpy/arrayobject.h> | |||
There was a problem hiding this comment.
This should be moved into the PY_MAJOR_VERSION=3, since numpy is not needed in Python 2. I will push a commit that moves this in.
gustavla
force-pushed
the
dev/py3-support
branch
from
83fce03 to
4cb1c3c
Compare
coremltools/models/utils.py
Outdated
| @@ -21,6 +21,13 @@ | |||
| import scipy.sparse as _sp | |||
|
|
|||
|
|
|||
| def to_unicode(x): | |||
There was a problem hiding this comment.
I think this should be _to_unicode
There was a problem hiding this comment.
Good point. I'll update!
|
I have now tested this with libsvm, xgboost, and keras, and as far as I can tell only tests that were already failing are failing. I'm merging this now. |
|
I don't know why github wanted two commits to resolve the merge conflict, so sorry for cluttering the history a bit! |
* added kwargs for easier intialisation of random model * initial commit for conversion script * current debug script * update * Update * done * add updated debug conversion script * style * clean conversion script
This gets me passed all non-slow unit tests (with lots skipped - libsvm/xgboost). I will update with comments when I am able to run more tests and confirm they also work.
This PR does not include build configuration for Python 3.