eric-haibin-lin commented on a change in pull request #18228: URL: https://github.com/apache/incubator-mxnet/pull/18228#discussion_r424742207
########## File path: python/mxnet/rnn/rnn_cell.py ########## @@ -459,6 +459,11 @@ def __call__(self, inputs, states): name='%so'%name) next_c = symbol._internal._plus(forget_gate * states[1], in_gate * in_transform, name='%sstate'%name) + next_c._set_attr(force_mirroring='0') Review comment: Does it still requires hints from the user's model code - can the subgraph pattern be detected automatically in the backend? ########## File path: docs/static_site/src/pages/api/faq/env_var.md ########## @@ -189,7 +189,7 @@ $env:MXNET_STORAGE_FALLBACK_LOG_VERBOSE=0 - The maximum size of an NDArray slice in terms of number of parameters. - This parameter is used to slice an NDArray before synchronizing through P3Store (dist_p3). -## Memonger +## Memory Optimizations Review comment: Since the version is bumped to mxnet 2.0 already, we have the opportunity to change some of the user facing API/options. What about consolidating `MXNET_BACKWARD_DO_MIRROR` and `MXNET_MEMORY_OPT` into one variable `MXNET_MEMORY_OPT_LEVEL`, where: - 0 means disabled - 1 means basic optimizations - 2 includes the latest optimizations from this PR ? ---------------------------------------------------------------- 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