Skip to content

Conversation

@Aaraviitkgp
Copy link

@Aaraviitkgp Aaraviitkgp commented Nov 27, 2025

issue: #42451

This PR fixes a bug in KernelConfig.create_compatible_mapping where defining a kernel mapping for multiple devices (e.g., both cuda and rocm) caused the last defined device to overwrite the active device configuration.

@Rocketknight1 @MekkCyber

@Aaraviitkgp Aaraviitkgp force-pushed the kernel_mapping_error_resolve branch from 9623bfe to 04e27cb Compare November 27, 2025 23:20
@Aaraviitkgp Aaraviitkgp force-pushed the kernel_mapping_error_resolve branch from 5a4a08a to e328d0c Compare November 27, 2025 23:25
@Aaraviitkgp
Copy link
Author

Aaraviitkgp commented Nov 27, 2025

@Rocketknight1 I noticed that the run test issues that is occuring ColQwen2ForRetrieval: Tensor vlm.lm_head.weight seems to have nothing to do with the changes made in pr.

@Rocketknight1
Copy link
Member

cc @MekkCyber for kernels I think! also @Aaraviitkgp that ColQwen2 issue should be fixed now if you rebase, sorry about that!

@Aaraviitkgp
Copy link
Author

@Rocketknight1 the test_processor and test_tokenization have similar issue they fail but no changes in pr related to them can you look into that aswell 😁.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

2 participants