anirudh2290 commented on a change in pull request #15245: FP16 Support for C
Predict API
URL: https://github.com/apache/incubator-mxnet/pull/15245#discussion_r303110259
##########
File path: amalgamation/python/mxnet_predict.py
##########
@@ -160,10 +249,18 @@ def forward(self, **kwargs):
>>> predictor.forward(data=mydata)
>>> out = predictor.get_output(0)
"""
+ if self.type_dict and len(self.type_dict) != len(kwargs.items()):
+ raise ValueError("number of kwargs should be same as len of
type_dict" \
+ "Please check your forward pass inputs" \
+ "or type_dict passed to Predictor instantiation")
+
for k, v in kwargs.items():
if not isinstance(v, np.ndarray):
raise ValueError("Expect numpy ndarray as input")
- v = np.asarray(v, dtype=np.float32, order='C')
+ if self.type_dict and k in self.type_dict:
+ v = np.asarray(v, dtype=self.type_dict[k], order='C')
+ else:
+ v = np.asarray(v, dtype=np.float32, order='C')
Review comment:
we use the lookup table to map the dtype int, which comes from mxnet backend
to numpy dtype. For cases, we are not determining it from backend dtype I think
it is better to use np.float32 rather than _DTYPE_MX_TO_NP[0].
----------------------------------------------------------------
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
For queries about this service, please contact Infrastructure at:
us...@infra.apache.org
With regards,
Apache Git Services
