This repository has been archived by the owner on Nov 17, 2023. It is now read-only.
[FEATURE] Add feature of attach_grad to nonleaf variables in HybridizedBlock. #21089
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
Description
This is the same as this PR. The PR adds the support for fetching the gradients of intermediate variables in a gluon HybridizedBlock. This applies uniformly to both when
block.hybridize()
is on and off. This generates theattach_grad
implemented in implemented in PR#20500.The motivation of this feature comes from this issue#11865.
Checklist
Essentials
Changes
block.py
wheremark_vars
andget_mark_vars
are added along withMXNDArrayMarkDCVariables
.cached_op.invoke
in cpp backend andCachedOp.__call__
have been editted to include the pass of marked nonleaf ndarrays.set_nleafs
method is added intoCachedOp
class to store the marked nonleaf ndarrays.void RunGraph
, marked nonleaf ndarrays are linked to the marked computational node for autograd computation.Comments