-
Notifications
You must be signed in to change notification settings - Fork 33
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
Regenerate ONNX tests using newer torch-mlir. #84
Conversation
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Note that we're down to 48 import failures now, compared to just 2 upstream in torch-mlir. The difference there comes from iree_tests/onnx/import_tests.py
not handling all input and output protobufs from the ONNX test suite (e.g. sequence types, optional types). To support those (if we care), we'd need to write the conversion code from ONNX protos to .npy or .bin files. ONNX has helper functions here: https://github.com/onnx/onnx/blob/b48d763a2d22574540f51b6bc5ce1a537fb65024/onnx/backend/test/runner/__init__.py#L510-L523 but I wasn't quite sure how to get those saved to .npy files.
Ping? Want to review? (I can also just merge myself with no review :P) |
[AMD Official Use Only - General]
Sorry, could not get to it sooner. Will review by 11 am PST.
From: Scott Todd ***@***.***>
Sent: Monday, March 4, 2024 8:48 AM
To: nod-ai/SHARK-TestSuite ***@***.***>
Cc: Deepak, Kumar ***@***.***>; Review requested ***@***.***>
Subject: Re: [nod-ai/SHARK-TestSuite] Regenerate ONNX tests using newer torch-mlir. (PR #84)
Caution: This message originated from an External Source. Use proper caution when opening attachments, clicking links, or responding.
Ping? Want to review? (I can also just merge myself with no review :P)
—
Reply to this email directly, view it on GitHub<#84 (comment)>, or unsubscribe<https://github.com/notifications/unsubscribe-auth/A5OMX32TVUUPMLMGRJYCKH3YWSQVFAVCNFSM6AAAAABECIXUGWVHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMYTSNZXGAZDOOBVGM>.
You are receiving this because your review was requested.Message ID: ***@***.******@***.***>>
|
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Scanned the files changed. Looks good.
* Regenerate onnx tests using latest torch-mlir. * Update XFAIL lists in CPU and GPU Vulkan configs. * Rename config_cpu to config_cpu_llvm_sync matching IREE.
Generated with
none
for missing operands - these could be cleaned up with DCE/CSE/etc.)(Would be nice to put that version information in the repo / test case files somewhere...)
Unfortunately, none of the newly importable tests compile: