DickJC123 commented on a change in pull request #14200: Bulked op segments to allow Variable nodes URL: https://github.com/apache/incubator-mxnet/pull/14200#discussion_r260107119
########## File path: src/imperative/cached_op.h ########## @@ -53,10 +53,12 @@ struct CachedOpConfig : public dmlc::Parameter<CachedOpConfig> { .set_default(2) .describe("Maximum number of operators that can be inlined."); DMLC_DECLARE_FIELD(forward_bulk_size) - .set_default(dmlc::GetEnv("MXNET_EXEC_BULK_EXEC_MAX_NODE_TRAIN", 15)) Review comment: Please review the latest commits, which address this valid concern about consistency. I also consolidated all op-bulking env var references to a central place and added timing-based tests for the perf impact of bulking. I'm happy with the PR now (assuming it passes CI). Anyone else you want to pull into the review @eric-haibin-lin? ---------------------------------------------------------------- This is an automated message from the Apache Git Service. To respond to the message, please log on 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