Accumulate loop dyn size #657
Merged
Conversation
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
albertz
force-pushed
the
albert-accum-loop-dyn-size
branch
3 times, most recently
from
bd72b69 to
3fe1cac
Compare
|
Finally. Again more effort than expected. This still needs to be cleaned up. Maybe some aspects will be separated. Like the |
albertz
force-pushed
the
albert-accum-loop-dyn-size
branch
2 times, most recently
from
70672a3 to
fb7f54d
Compare
This comment has been minimized.
This comment has been minimized.
This comment has been minimized.
This comment has been minimized.
This was referenced Sep 21, 2021
Merged
Merged
Merged
albertz
force-pushed
the
albert-accum-loop-dyn-size
branch
2 times, most recently
from
8f2c0c9 to
37f4eca
Compare
This was referenced Sep 21, 2021
albertz
force-pushed
the
albert-accum-loop-dyn-size
branch
from
37f4eca to
38a6e62
Compare
albertz
added a commit
that referenced
this pull request
Sep 21, 2021
albertz
force-pushed
the
albert-accum-loop-dyn-size
branch
from
38a6e62 to
3706730
Compare
albertz
force-pushed
the
albert-accum-loop-dyn-size
branch
from
3706730 to
3d17449
Compare
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
2 participants
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Outdated suggestions cannot be applied.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.
Support of accumulation of dynamic sizes (
DimensionTag.dyn_size_ext) within a recurrent context (control_flow_context). E.g. when the dyn size (seq lengths) is of shape [B] inside the loop, and this dyn size changes in every iteration (i.e.control_flow_contextis set to the rec loop), then the accumulated dyn size outside the loop would have shape [T,B].Further, when some layer inside the loop has such a dynamic shape [B,T_per_frame], where T_per_frame is the dyn size as mentioned before, this means that we cannot use the current
TensorArraylogic, which expects that elements always have the same shape in every frame. This PR also extends that logic to support that.This was a request by #635 and also makes sense for #589, #579 (and other related commits for #391).