njayaram2 commented on a change in pull request #375: DL: Add GPU support for
evaluate
URL: https://github.com/apache/madlib/pull/375#discussion_r278349068
##########
File path: src/ports/postgres/modules/deep_learning/madlib_keras.py_in
##########
@@ -391,14 +373,17 @@ def fit_transition(state, ind_var, dep_var,
current_seg_id, num_classes,
with K.tf.device(device_name):
updated_weights = segment_model.get_weights()
-
- if SD['buffer_count'] == total_buffers:
- if total_buffers == 0:
- plpy.error('total buffers is 0')
-
- agg_loss /= total_buffers
- agg_accuracy /= total_buffers
- if not is_pg:
+ if is_platform_pg():
+ total_buffers_per_seg = total_buffers_per_seg[0]
+ else:
+ total_buffers_per_seg =
total_buffers_per_seg[all_seg_ids.index(current_seg_id)]
+ if total_buffers_per_seg == 0:
+ plpy.error('total buffers is 0')
Review comment:
Ideally, we should not be hitting this error. So, can we have a more verbose
error message, and print out the `current_seg_id` too?
----------------------------------------------------------------
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
