revert: remove adapters / intrinsics / alora / lora from openai code - #543
Merged
Conversation
Contributor
|
The PR description has been updated. Please fill out the template for your PR to be reviewed. |
Merge ProtectionsYour pull request matches the following merge protections and will not be merged until they are valid. 🟢 Enforce conventional commitWonderful, this rule succeeded.Make sure that we follow https://www.conventionalcommits.org/en/v1.0.0/
|
jakelorocco
force-pushed
the
jal/remove-vllm-adapters
branch
8 times, most recently
from
February 12, 2026 20:01
777d43d to
54262dc
Compare
Contributor
Author
|
Note: while reverting these changes, I found an issue with using format in generate_from_raw requests. Newer vllm server version expect different parameter structures. I added some code to try to find the vllm version and then use that version in the request. Added tests to test the underlying function. We don't have a great way to test different vllm server versions, but I tested older and newer versions manually. Both worked. |
Contributor
Author
|
@guicho271828, could you please review this when you get a chance. Thanks! |
jakelorocco
force-pushed
the
jal/remove-vllm-adapters
branch
3 times, most recently
from
February 12, 2026 20:55
52eb26b to
6848fd9
Compare
Member
|
@guicho271828 context: the vLLM aLoRA PR will not accepted, so the alora/intrinsics code for openai is now all dead code. |
openai backend no longer supports intrinsics for vllm; vllm branch to support aloras was closed with prejudice.
guicho271828
force-pushed
the
jal/remove-vllm-adapters
branch
from
February 17, 2026 18:25
dce4862 to
d6a9d62
Compare
guicho271828
force-pushed
the
jal/remove-vllm-adapters
branch
from
February 17, 2026 21:11
0206743 to
065f7f3
Compare
Contributor
|
rewrote the tests; test pass. Since there is no need for using the forked vllm, it now simply launches a child vllm process. |
jakelorocco
enabled auto-merge
February 17, 2026 21:31
jakelorocco
disabled auto-merge
February 17, 2026 21:31
planetf1
added a commit
to planetf1/mellea
that referenced
this pull request
May 8, 2026
- Prominent linked header to Epic generative-computing#929 + reference table of all related issues/PRs at the top of the doc so context is immediate on share. - §1: add evidence table of seven recent fix-up commits in the adapter area to demonstrate the friction is real. - §2: rewrite customer-adapter goal to distinguish current partial state (m alora train exists; consuming needs hacks; generative-computing#424 open) from the refactor's first-class intent. - §4.1: new backend × reality matrix showing Reality A/B/C support today and planned (with generative-computing#1018 for HF + embedded, generative-computing#27 for vLLM). - New §6 "Impact and blast radius": API surface, user-archetype impact table, code reach, release planning, blocking/unblocking, performance notes, risk register. - §6.3: fix vLLM/aLoRA factual error — upstream vLLM never had aLoRA; we ran a custom vLLM build; PR generative-computing#543 was triggered by upstream declining the aLoRA PR, not by vLLM "dropping" support. Cite generative-computing#27 as the live tracking item. - §5 reviewer questions: drop telemetry coupling as a reviewer question (move to implementation note); reframe Reality C question around generative-computing#27 upstream status rather than C1-vs-C2 choice. - §11: acknowledge TestBasedEval and BenchDrift are part of the broader LLM-unit-testing conversation, not invented for this refactor. Assisted-by: Claude Code Signed-off-by: Nigel Jones <jonesn@uk.ibm.com>
akihikokuroda
pushed a commit
to akihikokuroda/mellea
that referenced
this pull request
May 27, 2026
…enerative-computing#543) * revert: remove adapters / intrinsics / alora / lora from openai code openai backend no longer supports intrinsics for vllm; vllm branch to support aloras was closed with prejudice. * fix: fix generate_from_raw with format for newer vllm server versions * fix: remove unnecessary vllm flags * doc: removed more mentions to VLLM_USE_V1=0 * refactor: make openai vllm test 'normal' * fix: vllm server complains numpy version --------- Co-authored-by: Masataro Asai <guicho2.71828@gmail.com>
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
3 participants
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Misc PR
Type of PR
Description
Currently removed all adapter based code;
will confirm if that's the intent or if we should just remove support for activate loras while still allowing for lora adapters.Consensus was to just remove all support and keep huggingface as the backend that supports adapters/aloras/intrinsics. The logic for this is that the vllm implementation that will eventually support aloras / intrinsics / arbitrary adapters will likely be a much different implementation. As a result, we are removing this code.Testing