add map_matmul and fc_act_fuse passes to quant2_int8_mkldnn_pass #38023
Conversation
This file contains 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
|
Thanks for your contribution! |
sfraczek
changed the title
|
@baoachun @Aganlengzi Could you please merge this PR? |
Caozhou1995
pushed a commit
to Caozhou1995/Paddle
that referenced
this pull request
Dec 29, 2021
…dlePaddle#38023) * add map_matmul passes to quant2_int8_mkldnn_pass * fix fc+act fuse (activation scale) * ci fix, c++17 structured bindings not available * fix ci static check
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
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Outdated suggestions cannot be applied.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.
PR types
Performance optimization
PR changes
Others
Describe
Fixed mkldnn fc with activation.
Also by adding map_matmul passes, all matmul_v2 ops are converted to either matmul or mul. Afterwards mul+elementiwse_add are fused to fc by fc_fuse_pass.
Adding those 3 map passes fix a problem of crashing introduced by putting matmul_v2_transpose_reshape_fuse_pass in this file. They improve performance. Those passes together with matmul_v2_transpose_reshape fuse pass change accuracy from acc: 0.5478 to 0.5383, but accuracy before this change was higher than reference obtained from tnewst_quant_model.
I also added fc_act_mkldnn_fuse pass which fuses fc+gelu. It improves performance.
Accuracy written in here: #36962 (comment)