[AMP] Register some new ops #9849
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
AndrewZhaoLuo
requested review from
MarisaKirisame,
ZihengJiang,
anijain2305,
icemelon,
jroesch,
junrushao,
jwfromm,
mbrookhart,
slyubomirsky,
vinx13,
wweic,
yzhliu and
zhiics
as code owners
Contributor
Author
|
cc @anwang2009 |
Co-authored-by: Cody Yu <[email protected]>
ylc
pushed a commit
to ylc/tvm
that referenced
this pull request
Feb 16, 2022
* add new mixed precision ops * remove space * Update python/tvm/relay/transform/mixed_precision.py Co-authored-by: Cody Yu <[email protected]> Co-authored-by: Cody Yu <[email protected]>
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
3 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Outdated suggestions cannot be applied.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.
Add some unregistered ops to the default AMP lists.
By default the pass places unregistered ops to the follow list so this does not change behavior. It does suppress some errors though.
nn.depth_to_space-- this is just a fancy reshape op so should be in follow listargmax,argmin-- this one is pretty much the same computation as inminandmaxwhich are in the follow listsfast_*-- these fast operations use approximations involving polynomials in a clipped domain of the normal functions. They seem safe, thoughfast_tanhhas coefficients outside of representable range of fp16 (which I think is fine)clip-- usually used to implement relu6, might have problems if clip range are outside of fp16 range but this should be rareimage.resize2d-- The amount of accumulation is pretty minimalExcepting the
fast_*functions which have no analog I know of, this matches closely with TFs mixed precision list (github.com/tensorflow/tensorflow/blob/v2.5.0/tensorflow/core/grappler/optimizers/auto_mixed_precision_lists.h) so feel confident about this change.