triton-inference-server / server

The Triton Inference Server provides an optimized cloud and edge inferencing solution.
https://docs.nvidia.com/deeplearning/triton-inference-server/user-guide/docs/index.html
BSD 3-Clause "New" or "Revised" License
8.4k stars 1.49k forks source link

Revert "Fixing StringTo uint32_t used only by tracing (#6883)" #7596

Open rvroge opened 2 months ago

rvroge commented 2 months ago

We are running the server and core in a armv7 device. Build for platform linux/arm/v7 failed because the reverted commit was disabling the template for 32 unsigned int variables.

This reverts commit fbe03b34f7836198d367a5cd3519ead64b278a2d.

What does the PR do?

Checklist

Commit Type:

Check the conventional commit type box here and add the label to the github PR.

Related PRs:

https://github.com/triton-inference-server/core/pull/395

Where should the reviewer start?

Test plan:

Caveats:

Background

Related Issues: (use one of the action keywords Closes / Fixes / Resolves / Relates to)

rmccorm4 commented 2 months ago

Hi @rvroge, thanks for contributing the PR! CC'ing a couple folks who may have some more context @oandreeva-nv @nv-kmcgill53

nv-kmcgill53 commented 2 months ago

Just to confirm, you are wanting to build triton without tracing but you need access to the std::stoul template for your use case? We would need to confirm on our end that this change doesn't break our build with tracing enabled before merging.

rvroge commented 2 months ago

Just to confirm, you are wanting to build triton without tracing but you need access to the std::stoul template for your use case? We would need to confirm on our end that this change doesn't break our build with tracing enabled before merging.

That's correct, we do not want to enable tracing. However, the build in our platform attempts to use the template for uint32_t (we are building for a 32 bit architecture).