Merge pull request #4771 from nouiz/cudnn_conv_precision_fix

theano/configdefaults.py

@@ -375,7 +375,7 @@ AddConfigVar('dnn.enabled',
              " to not using it if not present."
              " If True and cuDNN can not be used, raise an error."
              " If False, disable cudnn",
-             StrParam("auto", "True", "False"),
+             EnumStr("auto", "True", "False"),
              in_c_key=False)
 
 # This flag determines whether or not to raise error/warning message if

theano/gpuarray/dnn.py

@@ -873,7 +873,7 @@ def dnn_conv(img, kerns, border_mode='valid', subsample=(1, 1),
         Convolution implementation to use. Some of its values may
         require certain versions of cuDNN to be installed. Default is
         the value of :attr:`config.dnn.conv.algo_fwd`.
-    precision : {'as_input', 'float16', 'float32', 'float64'}
+    precision : {'as_input_f32', 'as_input', 'float16', 'float32', 'float64'}
         Description of the dtype in which the computation of the convolution
         should be done. Possible values are 'as_input', 'float16', 'float32'
         and 'float64'. Default is the value of

theano/sandbox/cuda/dnn.py

@@ -1108,7 +1108,7 @@ def dnn_conv(img, kerns, border_mode='valid', subsample=(1, 1),
         Convolution implementation to use. Some of its  values may require certain
         versions of cuDNN to be installed. Default is the value of
         :attr:`config.dnn.conv.algo_fwd`.
-    precision : {'as_input', 'float16', 'float32', 'float64'}
+    precision : {'as_input_f32', 'as_input', 'float16', 'float32', 'float64'}
         Description of the dtype in which the computation of the convolution
         should be done. Possible values are 'as_input', 'float16', 'float32'
         and 'float64'. Default is the value of
@@ -1122,8 +1122,12 @@ def dnn_conv(img, kerns, border_mode='valid', subsample=(1, 1),
     # Establish dtype in which to perform the computation of the convolution
     if precision is None:
         precision = theano.config.dnn.conv.precision
-    if precision == 'as_input':
-        precision = theano.scalar.upcast(img.dtype, kerns.dtype)
+    if precision == 'as_input' or precision == 'as_input_f32':
+        nprec = theano.scalar.upcast(img.dtype, kerns.dtype)
+        if nprec == 'float16' and precision == 'as_input_f32':
+            precision = 'float32'
+        else:
+            precision = nprec
 
     # Check if deprecated param 'workmem' is used
     if workmem is not None:
@@ -1218,8 +1222,8 @@ def dnn_conv3d(img, kerns, border_mode='valid', subsample=(1, 1, 1),
         for the conv3d. Default is the value of
         :attr:`config.dnn.conv.algo_fwd`.
     :param precision : dtype in which the computation of the convolution
-        should be done. Possible values are 'as_input', 'float16', 'float32'
-        and 'float64'. Default is the value of
+        should be done. Possible values are 'as_input_f32', 'as_input',
+        'float16', 'float32' and 'float64'. Default is the value of
         :attr:`config.dnn.conv.precision`.
 
     :warning: The cuDNN library only works with GPU that have a compute
@@ -1234,8 +1238,12 @@ def dnn_conv3d(img, kerns, border_mode='valid', subsample=(1, 1, 1),
     # Establish dtype in which to perform the computation of the convolution
     if precision is None:
         precision = theano.config.dnn.conv.precision
-    if precision == 'as_input':
-        precision = theano.scalar.upcast(img.dtype, kerns.dtype)
+    if precision == 'as_input' or precision == 'as_input_f32':
+        nprec = theano.scalar.upcast(img.dtype, kerns.dtype)
+        if nprec == 'float16' and precision == 'as_input_f32':
+            precision = 'float32'
+        else:
+            precision = nprec
 
     # Check if deprecated param 'workmem' is used
     if workmem is not None: