You are viewing a plain text version of this content. The canonical link for it is here.
Posted to commits@tvm.apache.org by "vacu9708 (via GitHub)" <gi...@apache.org> on 2025/05/18 18:35:31 UTC

[PR] (Re-upload due to mistakes) [Doc][Relax][MetaSchedule] Fix errors in the e2e_opt_model.py tutorial [tvm]

vacu9708 opened a new pull request, #17986:
URL: https://github.com/apache/tvm/pull/17986

   Sorry for re-uploading this, I made some mistakes in my branch name.
   
   # Summary
   This PR resolves issue [#17899](https://github.com/apache/tvm/issues/17899) where the end-to-end tutorial `e2e_opt_model.py` fails with errors and cannot be executed successfully.
   
   # Bug fixes
   - Memory binding error("Did you forget to bind?")
   This error occurs when the compiler attempts to use operations that are not properly bound to GPU memory, due to occasionally generated invalid GPU schedules that result in memory misalignment. (probably due to vectorization)
   	- **Fix**: Added GPU code verification to SpaceGenerator() that filters out schedules involving memory misalignment.
   - <class 'tvm.ir.container.Array'> has no attribute numpy
   This error occurs because vm\["main"\]() returns an Array of NDArray
   	- **Fix**: Extracted the NDArray out of the returned Array before calling .numpy().


-- 
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.

To unsubscribe, e-mail: commits-unsubscribe@tvm.apache.org

For queries about this service, please contact Infrastructure at:
users@infra.apache.org


Re: [PR] [Doc][Relax][MetaSchedule] Resolve issue #17899 where the end-to-end tutorial e2e_opt_model.py fails with errors and cannot be executed successfully. [tvm]

Posted by "vacu9708 (via GitHub)" <gi...@apache.org>.
vacu9708 closed pull request #17986: [Doc][Relax][MetaSchedule] Resolve issue #17899 where the end-to-end tutorial e2e_opt_model.py fails with errors and cannot be executed successfully.
URL: https://github.com/apache/tvm/pull/17986


-- 
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.

To unsubscribe, e-mail: commits-unsubscribe@tvm.apache.org

For queries about this service, please contact Infrastructure at:
users@infra.apache.org


Re: [PR] [Doc][Relax][MetaSchedule] Resolve issue #17899 where the end-to-end tutorial e2e_opt_model.py fails with errors and cannot be executed successfully. [tvm]

Posted by "vacu9708 (via GitHub)" <gi...@apache.org>.
vacu9708 commented on PR #17986:
URL: https://github.com/apache/tvm/pull/17986#issuecomment-2898633595

   I need to look more into this issue


-- 
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.

To unsubscribe, e-mail: commits-unsubscribe@tvm.apache.org

For queries about this service, please contact Infrastructure at:
users@infra.apache.org


Re: [PR] [Doc][Relax][MetaSchedule] Resolve issue #17899 where the end-to-end tutorial e2e_opt_model.py fails with errors and cannot be executed successfully. [tvm]

Posted by "vacu9708 (via GitHub)" <gi...@apache.org>.
vacu9708 commented on PR #17986:
URL: https://github.com/apache/tvm/pull/17986#issuecomment-2897951582

   I confirmed that the change I made works but
   **this seems more correct**
   ```python
   SpaceGenerator.create(postprocs=target.kind.name)
   ```
   **instead of**
   ```python
   if "gpu" in target.keys:
                       space_generator = SpaceGenerator.create(
                           space_generator, postprocs=[postproc.VerifyGPUCode()]
                       )
                   else:
                       space_generator = SpaceGenerator.create(space_generator)
   ```
   
   so that PostprocDefaultCUDA can be used from postproc.py
   However, SpaceGenerator.create(postprocs=target.kind.name) does not work. This issue needs to be looked into more.


-- 
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.

To unsubscribe, e-mail: commits-unsubscribe@tvm.apache.org

For queries about this service, please contact Infrastructure at:
users@infra.apache.org