reductionista commented on a change in pull request #462: DL: Add asymmetric
cluster support for fit and evaluate
URL: https://github.com/apache/madlib/pull/462#discussion_r352941669
##########
File path: src/ports/postgres/modules/deep_learning/madlib_keras_wrapper.py_in
##########
@@ -69,15 +70,34 @@ def get_device_name_and_set_cuda_env(gpus_per_host, seg):
set_cuda_env('-1')
return device_name
-def set_keras_session(device_name, gpus_per_host, segments_per_host):
+def get_device_name_and_set_cuda_env(use_gpus, gpus_per_seg, seg):
+
+ if use_gpus:
+ gpu_count = gpus_per_seg
Review comment:
maybe `gpu_counts` could be that 3rd alternative to `gpus_per_seg` or
`gpus_per_host`? Since we end up calling each element a `gpu_count` anyway.
Only problem with that I see is, it's less descriptive than
`gpus_per_host` and might leave someone wondering whether the numbers represent
# gpus per host or per segment.
But maybe less descriptive would be less prone to someone making the
wrong assumption and getting confused; it would force whoever is modifying the
code to actually check where it's set?
----------------------------------------------------------------
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:
[email protected]
With regards,
Apache Git Services
