Skip to content

Add performance tuning functionalities#46

Merged
albanpuech merged 59 commits intomainfrom
add_perf_fix
Apr 2, 2026
Merged

Add performance tuning functionalities#46
albanpuech merged 59 commits intomainfrom
add_perf_fix

Conversation

@romeokienzler
Copy link
Copy Markdown
Collaborator

@romeokienzler romeokienzler commented Mar 31, 2026

No description provided.

@romeokienzler romeokienzler marked this pull request as draft March 31, 2026 05:55
@romeokienzler romeokienzler changed the title Add performance tuning functionalities [WIP] Add performance tuning functionalities Mar 31, 2026
albanpuech
albanpuech reviewed Mar 31, 2026
View reviewed changes
@albanpuech
Copy link
Copy Markdown
Collaborator

albanpuech commented Mar 31, 2026

@romeokienzler Is your integration test run when running the pytests? Not yet right?

@romeokienzler
Copy link
Copy Markdown
Collaborator Author

romeokienzler commented Mar 31, 2026

@romeokienzler Is your integration test run when running the pytests? Not yet right?

No @albanpuech - in TerraTorch we used to run unit tests on PR and integration tests on release - let me know your preference, I think the IT on GT are lightweight enough to also run on PR

@albanpuech
Copy link
Copy Markdown
Collaborator

albanpuech commented Mar 31, 2026

@romeokienzler Is your integration test run when running the pytests? Not yet right?

No @albanpuech - in TerraTorch we used to run unit tests on PR and integration tests on release - let me know your preference, I think the IT on GT are lightweight enough to also run on PR

Would be good to do it also on PR :)

@romeokienzler romeokienzler marked this pull request as ready for review April 2, 2026 07:21
@romeokienzler romeokienzler changed the title [WIP] Add performance tuning functionalities Add performance tuning functionalities Apr 2, 2026
@romeokienzler
fix gitignore
ec01b3c 
Signed-off-by: Romeo Kienzler <romeo.kienzler1@ibm.com>
@romeokienzler
add torch.compile support via --compile CLI flag
2d103b2 
Signed-off-by: Romeo Kienzler <romeo.kienzler1@ibm.com>
@romeokienzler
fix torch.compile max-autotune OOM: allow ATEN GEMM fallback
6b8aed9 
Signed-off-by: Romeo Kienzler <romeo.kienzler1@ibm.com>
@romeokienzler
fix torch.compile graph breaks: replace boolean index-put with torch.…
b3fd8c4 
…where, set float32 matmul precision

Signed-off-by: Romeo Kienzler <romeo.kienzler1@ibm.com>
@romeokienzler
add -tf32 and --bfloat16 parameters
84e613f 
Signed-off-by: Romeo Kienzler <romeo.kienzler1@ibm.com>
@romeokienzler
pass bfloat to lightning trainer as well if set
2bb6f13 
Signed-off-by: Romeo Kienzler <romeo.kienzler1@ibm.com>
@romeokienzler
enable autocast for HeteroData objects
6dae6d5 
Signed-off-by: Romeo Kienzler <romeo.kienzler1@ibm.com>
@romeokienzler
add smoke test
c9fc63d 
Signed-off-by: Romeo Kienzler <romeo.kienzler1@ibm.com>
@romeokienzler
fix value range
86fb524 
Signed-off-by: Romeo Kienzler <romeo.kienzler1@ibm.com>
@romeokienzler
fix doc, improve test
e2835e3 
Signed-off-by: Romeo Kienzler <romeo.kienzler1@ibm.com>
@romeokienzler
add missing doc
6a15194 
Signed-off-by: Romeo Kienzler <romeo.kienzler1@ibm.com>
@romeokienzler
improve test
6735e77 
Signed-off-by: Romeo Kienzler <romeo.kienzler1@ibm.com>
@romeokienzler
fix docs
5abe149 
Signed-off-by: Romeo Kienzler <romeo.kienzler1@ibm.com>
@romeokienzler
fix ordering
345167b 
Signed-off-by: Romeo Kienzler <romeo.kienzler1@ibm.com>
@romeokienzler
add fixure for cleanup
c02271c 
Signed-off-by: Romeo Kienzler <romeo.kienzler1@ibm.com>
@romeokienzler
fix test name
bed1e9e 
Signed-off-by: Romeo Kienzler <romeo.kienzler1@ibm.com>
@romeokienzler
remove bad doc
512dfc3 
Signed-off-by: Romeo Kienzler <romeo.kienzler1@ibm.com>
@romeokienzler
fix broken links
2f557ae 
Signed-off-by: Romeo Kienzler <romeo.kienzler1@ibm.com>
@romeokienzler
fix documentation
9bf689d 
Signed-off-by: Romeo Kienzler <romeo.kienzler1@ibm.com>
@romeokienzler
strip down
afb569b 
Signed-off-by: Romeo Kienzler <romeo.kienzler1@ibm.com>
@romeokienzler
simplify doc
69d35ea 
Signed-off-by: Romeo Kienzler <romeo.kienzler1@ibm.com>
@romeokienzler
add support for dataset wrapper
1e032be 
Signed-off-by: Romeo Kienzler <romeo.kienzler1@ibm.com>
@romeokienzler
precommit fix
a93f2ba 
Signed-off-by: Romeo Kienzler <romeo.kienzler1@ibm.com>
@romeokienzler
bump trivy
abacdc4 
Signed-off-by: Romeo Kienzler <romeo.kienzler1@ibm.com>
@romeokienzler
fix precommit
d49c8d1 
Signed-off-by: Romeo Kienzler <romeo.kienzler1@ibm.com>
@romeokienzler
fix trivy
2691248 
Signed-off-by: Romeo Kienzler <romeo.kienzler1@ibm.com>
@romeokienzler
add support for dataloading performance tests
d702616 
Signed-off-by: Romeo Kienzler <romeo.kienzler1@ibm.com>
@romeokienzler
fix validation order
8954822 
Signed-off-by: Romeo Kienzler <romeo.kienzler1@ibm.com>
@romeokienzler
precommit fix
698c8a4 
Signed-off-by: Romeo Kienzler <romeo.kienzler1@ibm.com>
@romeokienzler
security fix
0a0c5db 
Signed-off-by: Romeo Kienzler <romeo.kienzler1@ibm.com>
@romeokienzler
fix missing package and circular import
5de3045 
Signed-off-by: Romeo Kienzler <romeo.kienzler1@ibm.com>
@romeokienzler
ignore security CVE-2026-4539 as not relevant
7651ad6 
Signed-off-by: Romeo Kienzler <romeo.kienzler1@ibm.com>
@romeokienzler
fix tests
1684ffa 
Signed-off-by: Romeo Kienzler <romeo.kienzler1@ibm.com>
@romeokienzler
fix precommit
d62f5bf 
Signed-off-by: Romeo Kienzler <romeo.kienzler1@ibm.com>
@romeokienzler
add profiler cli argunent
7ff6c4c 
Signed-off-by: Romeo Kienzler <romeo.kienzler1@ibm.com>
@romeokienzler
add profiler cli argunent
7c1c245 
Signed-off-by: Romeo Kienzler <romeo.kienzler1@ibm.com>
@romeokienzler
add batch elapsed time logging
cb80c4c 
Signed-off-by: Romeo Kienzler <romeo.kienzler1@ibm.com>
@romeokienzler
test for semantic equivalenze emprirically
6e3e6a2 
Signed-off-by: Romeo Kienzler <romeo.kienzler1@ibm.com>
@romeokienzler
implement in aten safe way
6d95793 
Signed-off-by: Romeo Kienzler <romeo.kienzler1@ibm.com>
@romeokienzler
log (to mlflow) only every 1000 steps
7153039 
Signed-off-by: Romeo Kienzler <romeo.kienzler1@ibm.com>
romeokienzler and others added 8 commits April 2, 2026 13:06
@romeokienzler
fix syntax error
76e6bd5 
Signed-off-by: Romeo Kienzler <romeo.kienzler1@ibm.com>
@romeokienzler
Merge branch 'main' into add_perf_fix
b793b67 
Signed-off-by: Romeo Kienzler <5694071+romeokienzler@users.noreply.github.com>
@romeokienzler
fix merge error
1908ef5 
Signed-off-by: Romeo Kienzler <romeo.kienzler1@ibm.com>
@romeokienzler
fix
0dfa6e3 
Signed-off-by: Romeo Kienzler <romeo.kienzler1@ibm.com>
@romeokienzler
remove crashing code
e092dbd 
Signed-off-by: Romeo Kienzler <romeo.kienzler1@ibm.com>
@romeokienzler
fix precommit hook
775be84 
Signed-off-by: Romeo Kienzler <romeo.kienzler1@ibm.com>
@romeokienzler
fix received 0 items of ancdata
458eecc 
Signed-off-by: Romeo Kienzler <romeo.kienzler1@ibm.com>
@romeokienzler
fix precommit
51ca57f 
Signed-off-by: Romeo Kienzler <romeo.kienzler1@ibm.com>
romeokienzler
romeokienzler commented Apr 2, 2026
View reviewed changes
albanpuech
albanpuech approved these changes Apr 2, 2026
View reviewed changes
Copy link
Copy Markdown
Collaborator

@albanpuech albanpuech left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Thank you so much @romeokienzler

@albanpuech albanpuech merged commit ca05c73 into main Apr 2, 2026
11 checks passed
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Reviewers

@albanpuech albanpuech albanpuech approved these changes

Assignees

No one assigned

Labels

None yet

Projects

None yet

Milestone

No milestone

Development

Successfully merging this pull request may close these issues.

2 participants

@romeokienzler @albanpuech