Skip to content
This repository was archived by the owner on Jun 5, 2025. It is now read-only.

Enable the integration tests for vllm #806

Merged
merged 7 commits into from
Jan 30, 2025
Merged

Enable the integration tests for vllm #806

merged 7 commits into from
Jan 30, 2025

Conversation

rdimitrov
Copy link
Member

@rdimitrov rdimitrov commented Jan 29, 2025 •
edited
Loading

The following PR enables the integration tests for vllm.

Details:

  • Starts the CodeGate container in the host network (so everything can talk to each other without the need to specify explict ports), we don't have port conflicts anyway
  • Starts the vllm container, runs the Qwen/Qwen2.5-Coder-0.5B-Instruct model and verifies if it is serving as expected
  • Updates the vllm tests to use Qwen/Qwen2.5-Coder-0.5B-Instruct
  • Updates the chat test to expect a simpler welcome message
  • Updates the vllm provider making the Authorization header optional

Fixes: #816

Nit: For some reason I wasn't able to get vllm running on my Mac (issue).

What's left:

  • Figure out why the FIM test is failing. Note that I tried with 1.5B and 3B models too. The 1.5B failed, whereas the 3B model took more than 1hr of CI to compute and I canceled it at the end, most certainly because we are running on CPU mode only in a low-performant environment).

Logs:

@rdimitrov rdimitrov force-pushed the vllm-integration-tests branch 4 times, most recently from 4b58358 to 0d56e43 Compare January 29, 2025 10:39
@rdimitrov rdimitrov force-pushed the vllm-integration-tests branch 3 times, most recently from 3d703a0 to 4cd30f4 Compare January 29, 2025 15:03
@rdimitrov rdimitrov force-pushed the vllm-integration-tests branch from 4cd30f4 to 7a70414 Compare January 29, 2025 15:20
jhrozek
jhrozek approved these changes Jan 29, 2025
View reviewed changes
Copy link
Contributor

@jhrozek jhrozek left a comment

Choose a reason for hiding this comment

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

really nice work!

lukehinds
lukehinds approved these changes Jan 30, 2025
View reviewed changes
Copy link

@lukehinds lukehinds left a comment

Choose a reason for hiding this comment

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

+1 awesome stuff

@lukehinds lukehinds merged commit ddbffb3 into main Jan 30, 2025
4 checks passed
@lukehinds lukehinds deleted the vllm-integration-tests branch January 30, 2025 07:22
lukehinds pushed a commit that referenced this pull request Jan 31, 2025
@rdimitrov
Enable the integration tests for vllm (#806)
17e949d 
* Add integration tests for vllm

Signed-off-by: Radoslav Dimitrov <[email protected]>

* Try using Qwen/Qwen2.5-Coder-3B-Instruct

Signed-off-by: Radoslav Dimitrov <[email protected]>

* Go back to using Qwen/Qwen2.5-Coder-0.5B-Instruct

Signed-off-by: Radoslav Dimitrov <[email protected]>

* Reformat the vllm_fim test

Signed-off-by: Radoslav Dimitrov <[email protected]>

* Use Qwen/Qwen2.5-Coder-0.5B

Signed-off-by: Radoslav Dimitrov <[email protected]>

* Revert "Use Qwen/Qwen2.5-Coder-0.5B"

This reverts commit 32b2d8c.

* Update the expected result for vllm_fim

Signed-off-by: Radoslav Dimitrov <[email protected]>

---------

Signed-off-by: Radoslav Dimitrov <[email protected]>
Sign up for free to subscribe to this conversation on GitHub. Already have an account? Sign in.
Reviewers

@jhrozek jhrozek jhrozek approved these changes

@lukehinds lukehinds lukehinds approved these changes

Assignees
No one assigned
Labels
None yet
Projects
None yet
Milestone
No milestone
Development

Successfully merging this pull request may close these issues.

Enable the integration tests for vllm
3 participants
@rdimitrov @jhrozek @lukehinds