Windows Compatibility

digits/config/caffe_option.py

@@ -65,6 +65,8 @@ def validate(cls, value):
         if value == '<PATHS>':
             # Find the executable
             executable = cls.find_executable('caffe')
+            if not executable:
+                executable = cls.find_executable('caffe.exe')
             if not executable:
                 raise config_option.BadValue('caffe binary not found in PATH')
             cls.validate_version(executable)
@@ -187,6 +189,9 @@ def get_version(executable):
         elif platform.system() == 'Darwin':
             # XXX: guess and let the user figure out errors later
             return (0,11,0)
+        elif platform.system() == 'Windows':
+            # XXX: guess and let the user figure out errors later
+            return (0,11,0)
         else:
             print 'WARNING: platform "%s" not supported' % platform.system()
             return None
@@ -197,6 +202,8 @@ def _set_config_dict_value(self, value):
         else:
             if value == '<PATHS>':
                 executable = self.find_executable('caffe')
+                if not executable:
+                    executable = self.find_executable('caffe.exe')
             else:
                 executable = os.path.join(value, 'build', 'tools', 'caffe')
 
@@ -238,7 +245,7 @@ def apply(self):
                 print 'Did you forget to "make pycaffe"?'
                 raise
 
-            if platform.system() == 'Darwin':
+            if platform.system() == 'Darwin' or platform.system() == 'Windows':
                 # Strange issue with protocol buffers and pickle - see issue #32
                 sys.path.insert(0, os.path.join(
                     os.path.dirname(caffe.__file__), 'proto'))

digits/dataset/images/classification/job.py

@@ -37,7 +37,7 @@ def __setstate__(self, state):
                         import numpy as np
 
                         old_blob = caffe_pb2.BlobProto()
-                        with open(task.path(task.mean_file)) as infile:
+                        with open(task.path(task.mean_file),'rb') as infile:
                             old_blob.ParseFromString(infile.read())
                         data = np.array(old_blob.data).reshape(
                                 old_blob.channels,
@@ -48,7 +48,7 @@ def __setstate__(self, state):
                         new_blob.num = 1
                         new_blob.channels, new_blob.height, new_blob.width = data.shape
                         new_blob.data.extend(data.astype(float).flat)
-                        with open(task.path(task.mean_file), 'w') as outfile:
+                        with open(task.path(task.mean_file), 'wb') as outfile:
                             outfile.write(new_blob.SerializeToString())
                 else:
                     print '\tSetting "%s" status to ERROR because it was created with RGB channels' % self.name()

digits/dataset/images/generic/test_lmdb_creator.py

@@ -158,7 +158,7 @@ def _save_mean(mean, filename):
         blob.channels = 1
         blob.height, blob.width = mean.shape
         blob.data.extend(mean.astype(float).flat)
-        with open(filename, 'w') as outfile:
+        with open(filename, 'wb') as outfile:
             outfile.write(blob.SerializeToString())
 
     elif filename.endswith(('.jpg', '.jpeg', '.png')):

digits/device_query.py

@@ -111,12 +111,24 @@ def get_library(name):
             return ctypes.cdll.LoadLibrary('%s.so' % name)
         elif platform.system() == 'Darwin':
             return ctypes.cdll.LoadLibrary('%s.dylib' % name)
+        elif platform.system() == 'Windows':
+            return ctypes.windll.LoadLibrary('%s.dll' % name)
     except OSError:
         pass
     return None
 
 devices = None
 
+def get_cudart():
+    if not platform.system() == 'Windows':
+        return get_library('libcudart')
+
+    arch = platform.architecture()[0]
+    for ver in range(90,50,-5):
+        cudart = get_library('cudart%s_%d' % (arch[:2], ver))
+        if not cudart is None:
+            return cudart
+
 def get_devices(force_reload=False):
     """
     Returns a list of c_cudaDeviceProp's
@@ -131,7 +143,7 @@ def get_devices(force_reload=False):
         return devices
     devices = []
 
-    cudart = get_library('libcudart')
+    cudart = get_cudart()
     if cudart is None:
         return []
 
@@ -180,7 +192,9 @@ def get_nvml_info(device_id):
 
     nvml = get_library('libnvidia-ml')
     if nvml is None:
-        return None
+        nvml = get_library('nvml')
+        if nvml is None:
+            return None
 
     rc = nvml.nvmlInit()
     if rc != 0:

digits/job.py

@@ -118,7 +118,7 @@ def path(self, filename, relative=False):
             path = os.path.join(self._dir, filename)
         if relative:
             path = os.path.relpath(path, config_value('jobs_dir'))
-        return str(path)
+        return str(path).replace("\\","/")
 
     def path_is_local(self, path):
         """assert that a path is local to _dir"""

digits/model/images/classification/test_views.py

@@ -338,7 +338,7 @@ def test_classify_one(self):
         category = self.imageset_paths.keys()[0]
         image_path = self.imageset_paths[category][0]
         image_path = os.path.join(self.imageset_folder, image_path)
-        with open(image_path) as infile:
+        with open(image_path,'rb') as infile:
             # StringIO wrapping is needed to simulate POST file upload.
             image_upload = (StringIO(infile.read()), 'image.png')
 
@@ -360,7 +360,7 @@ def test_classify_one_json(self):
         category = self.imageset_paths.keys()[0]
         image_path = self.imageset_paths[category][0]
         image_path = os.path.join(self.imageset_folder, image_path)
-        with open(image_path) as infile:
+        with open(image_path,'rb') as infile:
             # StringIO wrapping is needed to simulate POST file upload.
             image_upload = (StringIO(infile.read()), 'image.png')
 

digits/model/images/classification/views.py

@@ -25,6 +25,7 @@
 from forms import ImageClassificationModelForm
 from job import ImageClassificationModelJob
 from digits.status import Status
+import platform
 
 NAMESPACE   = '/models/images/classification'
 
@@ -246,9 +247,11 @@ def image_classification_model_classify_one():
     if 'image_url' in flask.request.form and flask.request.form['image_url']:
         image = utils.image.load_image(flask.request.form['image_url'])
     elif 'image_file' in flask.request.files and flask.request.files['image_file']:
-        with tempfile.NamedTemporaryFile() as outfile:
-            flask.request.files['image_file'].save(outfile.name)
-            image = utils.image.load_image(outfile.name)
+        outfile = tempfile.mkstemp(suffix='.bin')
+        flask.request.files['image_file'].save(outfile[1])
+        image = utils.image.load_image(outfile[1])
+        os.close(outfile[0])
+        os.remove(outfile[1])
     else:
         raise werkzeug.exceptions.BadRequest('must provide image_url or image_file')
 

digits/model/images/generic/test_views.py

@@ -338,7 +338,7 @@ def test_model_json(self):
 
     def test_infer_one(self):
         image_path = os.path.join(self.imageset_folder, self.test_image)
-        with open(image_path) as infile:
+        with open(image_path,'rb') as infile:
             # StringIO wrapping is needed to simulate POST file upload.
             image_upload = (StringIO(infile.read()), 'image.png')
 
@@ -355,7 +355,7 @@ def test_infer_one(self):
 
     def test_infer_one_json(self):
         image_path = os.path.join(self.imageset_folder, self.test_image)
-        with open(image_path) as infile:
+        with open(image_path,'rb') as infile:
             # StringIO wrapping is needed to simulate POST file upload.
             image_upload = (StringIO(infile.read()), 'image.png')
 

digits/model/images/generic/views.py

@@ -22,6 +22,7 @@
 from forms import GenericImageModelForm
 from job import GenericImageModelJob
 from digits.status import Status
+import platform
 
 NAMESPACE   = '/models/images/generic'
 
@@ -222,9 +223,11 @@ def generic_image_model_infer_one():
     if 'image_url' in flask.request.form and flask.request.form['image_url']:
         image = utils.image.load_image(flask.request.form['image_url'])
     elif 'image_file' in flask.request.files and flask.request.files['image_file']:
-        with tempfile.NamedTemporaryFile() as outfile:
-            flask.request.files['image_file'].save(outfile.name)
-            image = utils.image.load_image(outfile.name)
+        outfile = tempfile.mkstemp(suffix='.bin')
+        flask.request.files['image_file'].save(outfile[1])
+        image = utils.image.load_image(outfile[1])
+        os.close(outfile[0])
+        os.remove(outfile[1])
     else:
         raise werkzeug.exceptions.BadRequest('must provide image_url or image_file')
 

digits/model/tasks/caffe_train.py

@@ -233,7 +233,7 @@ def save_files_classification(self):
             val_data_layer.data_param.backend = caffe_pb2.DataParameter.LMDB
         if self.use_mean:
             mean_pixel = None
-            with open(self.dataset.path(self.dataset.train_db_task().mean_file)) as f:
+            with open(self.dataset.path(self.dataset.train_db_task().mean_file),'rb') as f:
                 blob = caffe_pb2.BlobProto()
                 blob.MergeFromString(f.read())
                 mean = np.reshape(blob.data,
@@ -1312,7 +1312,7 @@ def get_transformer(self):
                 channel_swap = (2,1,0)
 
             if self.use_mean:
-                with open(self.dataset.path(self.dataset.train_db_task().mean_file)) as infile:
+                with open(self.dataset.path(self.dataset.train_db_task().mean_file),'rb') as infile:
                     blob = caffe_pb2.BlobProto()
                     blob.MergeFromString(infile.read())
                     mean_pixel = np.reshape(blob.data,
@@ -1331,7 +1331,7 @@ def get_transformer(self):
                 channel_swap = (2,1,0)
 
             if self.dataset.mean_file:
-                with open(self.dataset.path(self.dataset.mean_file)) as infile:
+                with open(self.dataset.path(self.dataset.mean_file),'rb') as infile:
                     blob = caffe_pb2.BlobProto()
                     blob.MergeFromString(infile.read())
                     mean_pixel = np.reshape(blob.data,

digits/task.py

@@ -15,6 +15,8 @@
 from config import config_value
 from status import Status, StatusCls
 
+import platform
+
 # NOTE: Increment this everytime the pickled version changes
 PICKLE_VERSION = 1
 
@@ -128,7 +130,7 @@ def path(self, filename, relative=False):
             path = os.path.join(self.job_dir, filename)
         if relative:
             path = os.path.relpath(path, config_value('jobs_dir'))
-        return str(path)
+        return str(path).replace("\\","/")
 
     def ready_to_queue(self):
         """
@@ -193,7 +195,7 @@ def run(self, resources):
                 stdout=subprocess.PIPE,
                 stderr=subprocess.STDOUT,
                 cwd=self.job_dir,
-                close_fds=True,
+                close_fds=False if platform.system() == 'Windows' else True,
                 )
 
         try:

digits/utils/__init__.py

@@ -2,17 +2,23 @@
 
 import os
 import math
-import fcntl
 import locale
 from random import uniform
 from urlparse import urlparse
 from io import BlockingIOError
 import inspect
+import platform
+
+
+if not platform.system() == 'Windows':
+    import fcntl
+else:
+    import gevent.os
 
 HTTP_TIMEOUT = 6.05
 
 def is_url(url):
-    return url is not None and urlparse(url).scheme != ""
+    return url is not None and urlparse(url).scheme != "" and not os.path.exists(url)
 
 def wait_time():
     """Wait a random number of seconds"""
@@ -27,22 +33,25 @@ def nonblocking_readlines(f):
        Newlines are normalized to the Unix standard.
     """
     fd = f.fileno()
-    fl = fcntl.fcntl(fd, fcntl.F_GETFL)
-    fcntl.fcntl(fd, fcntl.F_SETFL, fl | os.O_NONBLOCK)
+    if not platform.system() == 'Windows':
+        fl = fcntl.fcntl(fd, fcntl.F_GETFL)
+        fcntl.fcntl(fd, fcntl.F_SETFL, fl | os.O_NONBLOCK)
     enc = locale.getpreferredencoding(False)
 
     buf = bytearray()
     while True:
         try:
-            block = os.read(fd, 8192)
+            if not platform.system() == 'Windows':
+                block = os.read(fd, 8192)
+            else:
+                block = gevent.os.tp_read(fd, 8192)
         except (BlockingIOError, OSError):
             yield ""
             continue
 
         if not block:
             if buf:
                 yield buf.decode(enc)
-                buf.clear()
             break
 
         buf.extend(block)

digits/utils/test_image.py

@@ -7,6 +7,8 @@
 import mock
 import PIL.Image
 import numpy as np
+import os
+import platform
 
 from . import image as _, errors
 
@@ -52,9 +54,19 @@ def check_good_file(self, args):
         orig_mode, suffix, pixel, new_mode = args
 
         orig = PIL.Image.new(orig_mode, (10,10), pixel)
-        with tempfile.NamedTemporaryFile(suffix='.' + suffix) as tmp:
-            orig.save(tmp.name)
-            new = _.load_image(tmp.name)
+
+        # temp files cause permission errors so just generate the name
+        tmp = tempfile.mkstemp(suffix='.' + suffix)
+        orig.save(tmp[1])
+        new = _.load_image(tmp[1])
+        try:
+            # sometimes on windows the file is not closed yet
+            # which can cause an exception
+            os.close(tmp[0])
+            os.remove(tmp[1])
+        except:
+            pass
+
         assert new is not None, 'load_image should never return None'
         assert new.mode == new_mode, 'Image mode should be "%s", not "%s\nargs - %s' % (new_mode, new.mode, args)
 
@@ -94,16 +106,21 @@ def test_corrupted_file(self):
         corrupted = encoded[:size/2] + encoded[size/2:][::-1]
 
         # Save the corrupted image to a temporary file.
-        f = tempfile.NamedTemporaryFile(delete=False)
+        fname = tempfile.mkstemp(suffix='.bin')
+        f = os.fdopen(fname[0],'wb')
+        fname = fname[1]
+
         f.write(corrupted)
         f.close()
 
         assert_raises(
                 errors.LoadImageError,
                 _.load_image,
-                f.name,
+                fname,
                 )
 
+        os.remove(fname)
+
 class TestResizeImage():
 
     @classmethod