Skip to content

Update llama_serving.md doc to reflect the latest changes #1386

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

Open
wants to merge 1 commit into
base: main
Choose a base branch
from

Conversation

pravg-amd
Copy link
Contributor

  1. Update the shortfin response format, which contains the entire json instead of just the response text.
  2. Update the export and compilation commands for sharded models.

1. Update the shortfin response format, which contains the entire json instead of
   just the response text.
2. Update the export and compilation commands for sharded models.

Signed-off-by: Praveen G <[email protected]>
@@ -404,7 +409,11 @@ iree-compile /path/to/output/llama3.1-405b.mlir \
--iree-hal-target-device=hip[5] \
--iree-hal-target-device=hip[6] \
--iree-hal-target-device=hip[7] \
--iree-hip-target=gfx942
--iree-hip-target=gfx942 \
--iree-opt-level=O3 \
Copy link
Contributor

Choose a reason for hiding this comment

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

Do we know why these extra flags are required now for the compile?

Copy link
Contributor Author

Choose a reason for hiding this comment

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

I am not aware of exactly why we need these optimizations. Without these, was running into the following error

ValueError: shortfin_iree-src/runtime/src/iree/vm/bytecode/verifier.c:344: RESOURCE_EXHAUSTED; register count overflow

These flags are used in the llama benchmark tests in sharktank. Maybe @AmosLewis can provide more insight to this ?

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