[Model] Add GraniteMoeHybrid 4.0 model #17497

s3woz · 2025-04-30T20:07:16Z

The PR adds support for upcoming Granite4.0 models. It is a companion PR to huggingface/transformers#37658 for adding the same model to HF.

Note: Running the model in vLLM depends on having HF Transformers with Granite4.0 support installed, see https://huggingface.co/ibm-granite/granite-4.0-tiny-preview

Note: It is re-opening #17461 again after fixing wrong rebase. Currently rebased on newer main, signed-off, resolved conflicts.
@DarkLight1337 - As per suggestion, Sampler removed from the model.

The HF model is not available in main yet. Model tests are currently marked with 'skip'.

@tdoublep @bohnstingl

github-actions · 2025-04-30T20:07:27Z

👋 Hi! Thank you for contributing to the vLLM project.

💬 Join our developer Slack at https://slack.vllm.ai to discuss your PR in #pr-reviews, coordinate on features in #feat- channels, or join special interest groups in #sig- channels.

Just a reminder: PRs would not trigger full CI run by default. Instead, it would only run fastcheck CI which starts running only a small and essential subset of CI tests to quickly catch errors. You can run other CI tests on top of those by going to your fastcheck build on Buildkite UI (linked in the PR checks section) and unblock them. If you do not have permission to unblock, ping simon-mo or khluu to add you in our Buildkite org.

Once the PR is approved and ready to go, your PR reviewer(s) can run CI to test the changes comprehensively before merging.

To run CI, PR reviewers can either: Add ready label to the PR or enable auto-merge.

🚀

tlrmchlsmth

First pass through looks good and implementation looks clean!

When will the model be available for testing?

tlrmchlsmth · 2025-04-30T23:09:23Z

tests/models/decoder_only/language/test_granitemoehybrid.py

Can ibm-research/granite-4.0-tiny-test be added to test_hybrid.py? It would be good to get a TP test in place

Hey @tlrmchlsmth, I'm working on the transformers side of this PR - ibm-research/granite-4.0-tiny-test won't be the name of the actual model, it's just a placeholder since the name of the model hasn't been decided quite yet 😅

I agree that it would be nice to add a hybrid model test though

(CC @DarkLight1337 since I had just sent a note to him about this PR a little while ago also)

I added the model to the test_hybrid.py, but since the name is not final and therefore the link to HF is not established yet, the tests are marked as skipped for now. Is that okay?

Thanks @bohnstingl sounds good! Were you able to run the tests on the upcoming model? I had seen some tp failures in my environment, but not just for this model, so may have something going on with my machines

alex-jw-brooks

Thanks for opening this! some thoughts

alex-jw-brooks · 2025-05-01T01:47:34Z

vllm/model_executor/models/granitemoehybrid.py

+
+        self.quant_config = vllm_config.quant_config
+
+        super().__init__()


Is there a reason the super().__init__() is called in the middle here?

No, the call to super().__init__() has been moved to the first instruction.

Cool thanks! I think you may have forgotten to push changes for this file

alex-jw-brooks · 2025-05-01T01:48:58Z

vllm/model_executor/models/granitemoehybrid.py

+                hidden_states = hidden_states * self.embedding_multiplier
+            residual = None
+        else:
+            assert intermediate_tensors is not None


Can you remove the asserts and raise errors instead?

The asserts are replaced with RuntimeErrors.

alex-jw-brooks · 2025-05-01T01:52:26Z

vllm/model_executor/models/granitemoehybrid.py

+
+        self.position_embedding_type = config.position_embedding_type
+        if self.position_embedding_type == "rope":
+            self.rotary_emb = get_rope(


Can this be changed to set self.rotary_emb to None if it's not rope? This will avoid potential attribute errors and we can do

if self.rotary_emb is not None: query, key = self.rotary_emb(positions, query, key)

in forward() instead

I think we don't need to store the position_embedding_type attribute explicitly. Now we set self.rotary_emb directly based on config.position_embedding_type and to None if there is no rope being used.

alex-jw-brooks · 2025-05-01T01:56:01Z

vllm/model_executor/models/granitemoehybrid.py

+        else:
+            assert intermediate_tensors is not None
+            hidden_states = intermediate_tensors["hidden_states"]
+            residual = intermediate_tensors["residual"]


This will be overwritten right below?

The initial value of residual is now not fixed to None anymore.

DarkLight1337 · 2025-05-01T06:05:40Z

Can you move the test based on #17459?

mergify · 2025-05-01T17:00:02Z

This pull request has merge conflicts that must be resolved before it can be
merged. Please rebase the PR, @s3woz.

https://docs.github.com/en/pull-requests/collaborating-with-pull-requests/working-with-forks/syncing-a-fork

bohnstingl · 2025-05-01T17:01:22Z

@DarkLight1337, I moved the test_granitemoehybrid.py to tests/models/language/generation/test_granitemoehybrid.py. Is that okay?

DarkLight1337 · 2025-05-01T17:31:19Z

Yes sounds good.

alex-jw-brooks · 2025-05-01T23:20:59Z

tests/models/language/generation/test_hybrid.py

+    max_tokens: int,
+    num_logprobs: int,
+) -> None:
+    if model == "ibm-research/granite-4.0-tiny-test":


I think for now, we can just comment it out in the hybrid model list instead of skipping like this for these tests

I reverted the changes of test_hybrid.py and just added the models as a comment to the HYBRID_MODELS

alex-jw-brooks · 2025-05-01T23:21:55Z

tests/models/language/generation/test_hybrid.py

+    )
+
+
+@pytest.mark.parametrize("model", SSM_MODELS[0:1] + HYBRID_MODELS[0:2])


Is this supposed to be SSM_MODELS[0:1] + HYBRID_MODELS[0:2]) rather than [SSM_MODELS[0], HYBRID_MODELS[0]]?

This was just an approach to include the new model in the test, because with [SSM_MODELS[0], HYBRID_MODELS[0]], it would not be tested, right? I reverted those changes though for the moment.

Yup! I think that is on purpose (there is a comment at the top of the file), although now sure about the reason. I assume probably because they take awhile to run

Yeah it is to reduce CI cost

alex-jw-brooks · 2025-05-01T23:25:06Z

tests/models/language/generation/test_hybrid.py

@@ -0,0 +1,346 @@
+# SPDX-License-Identifier: Apache-2.0


This is the same as tests/models/decoder_only/language/test_hybrid.py with the new model added right? I think this PR has both files

I apologize, I missed to commit the deletion of the old file. This should now be fixed.

Signed-off-by: Thomas Ortner <[email protected]>

bohnstingl · 2025-05-02T05:53:45Z

@alex-jw-brooks. I apologize, I have missed to push the changes to vllm/model_executor/models/granitemoehybrid.py. I also did a rebase to the latest main, in order to have the new test structure correct. What do you think?

Signed-off-by: Stanislaw Wozniak <[email protected]>

Signed-off-by: Thomas Ortner <[email protected]>

Signed-off-by: Stanislaw Wozniak <[email protected]>

Co-authored-by: Thomas Ortner <[email protected]> Signed-off-by: Stanislaw Wozniak <[email protected]>

Signed-off-by: Stanislaw Wozniak <[email protected]>

docs/source/models/supported_models.md

Co-authored-by: Cyrus Leung <[email protected]> Signed-off-by: Stanislaw Wozniak <[email protected]>

s3woz · 2025-05-05T08:00:02Z

@DarkLight1337 , thank you for fixing the spacing.
I've also fixed model URLs, as we have it now uploaded in preview version.
Still, the HF code is not in the main yet, so skipping tests.
BTW, Does vLLM CI always pull the HF main? I.e. is it fine to enable test immediately after HF PR goes into main, or do we need to wait for HF release?

DarkLight1337 · 2025-05-05T08:08:08Z

The CI only uses the HF transformers version defined in the requirements file

… in CI Signed-off-by: Stanislaw Wozniak <[email protected]>

tests/models/registry.py

…rs in HF Transformers Signed-off-by: Stanislaw Wozniak <[email protected]>

DarkLight1337 · 2025-05-05T11:30:43Z

Maybe need to also add min_transformers_version

…in HF Transformers Signed-off-by: Stanislaw Wozniak <[email protected]>

s3woz · 2025-05-05T12:22:46Z

Maybe need to also add min_transformers_version

I don't know which exact version would include it. To be on the safe side, I've commented out for now.

Update: Apparently I cannot comment out, as tests fail then. As suggested, I've used min_transformers_version and set it to the next minor version 4.52.0, similarly to some other models in registry there.

Signed-off-by: Stanislaw Wozniak <[email protected]>

tests/models/language/generation/test_granitemoehybrid.py

tests/models/language/generation/test_hybrid.py

Co-authored-by: Tyler Michael Smith <[email protected]> Signed-off-by: Stanislaw Wozniak <[email protected]>

Signed-off-by: Stanislaw Wozniak <[email protected]>

s3woz · 2025-05-05T20:51:32Z

Two CI tests failed, both unrelated to this PR:

buildkite/ci/pr/v1-test — Failed (exit status 1)
 v1/engine/test_engine_core_client.py::test_startup_failure
 ...
 Start compiling function <code object forward at 0xda82150, file "/usr/local/lib/python3.12/dist-packages/vllm/model_executor/models/llama.py", line 345>
[2025-05-05T18:30:39Z] +++++++++++++++++++++++++++++++++++ Timeout ++++++++++++++++++++++++++++++++++++

and

buildkite/ci/pr/2-node-tests-4-gpus-in-total — Failed (exit status 1)
 docker run -d --gpus '"device=2,3"' ...
 nvidia-container-cli: device error: 3: unknown device: unknown.

I've force-pushed to retrigger CI checks and see if this persists. (Generally, I'm not sure if this is the right way to retrigger or is there a way to retrigger selected tests only.)

* [Model] Add GraniteMoeHybrid 4.0 model (vllm-project#17497) Signed-off-by: Thomas Ortner <[email protected]> Signed-off-by: Stanislaw Wozniak <[email protected]> Co-authored-by: Thomas Ortner <[email protected]> Co-authored-by: Cyrus Leung <[email protected]> Co-authored-by: Tyler Michael Smith <[email protected]> * [easy] Fix logspam on PiecewiseBackend errors (vllm-project#17138) Signed-off-by: rzou <[email protected]> * [Bugfix] Fixed prompt length for random dataset (vllm-project#17408) Signed-off-by: Mikhail Podvitskii <[email protected]> * [Doc] Update notes for H2O-VL and Gemma3 (vllm-project#17219) Signed-off-by: DarkLight1337 <[email protected]> * [Misc] Fix ScalarType float4 naming (vllm-project#17690) Signed-off-by: Lucas Wilkinson <[email protected]> * Fix `dockerfilegraph` pre-commit hook (vllm-project#17698) Signed-off-by: Harry Mellor <[email protected]> * [Bugfix] Fix triton import with local TritonPlaceholder (vllm-project#17446) Signed-off-by: Mengqing Cao <[email protected]> * [V1] Enable TPU V1 backend by default (vllm-project#17673) Signed-off-by: mgoin <[email protected]> * [V1][PP] Support PP for MultiprocExecutor (vllm-project#14219) Signed-off-by: jiang1.li <[email protected]> Signed-off-by: jiang.li <[email protected]> * [v1] AttentionMetadata for each layer (vllm-project#17394) Signed-off-by: Chen Zhang <[email protected]> * [Feat] Add deprecated=True to CLI args (vllm-project#17426) Signed-off-by: Aaron Pham <[email protected]> * [Docs] Use gh-file to add links to tool_calling.md (vllm-project#17709) Signed-off-by: windsonsea <[email protected]> * [v1] Introduce KVCacheBlocks as interface between Scheduler and KVCacheManager (vllm-project#17479) Signed-off-by: Chen Zhang <[email protected]> * [doc] Add RAG Integration example (vllm-project#17692) Signed-off-by: reidliu41 <[email protected]> Co-authored-by: reidliu41 <[email protected]> * [Bugfix] Fix modality limits in vision language example (vllm-project#17721) Signed-off-by: DarkLight1337 <[email protected]> * Make right sidebar more readable in "Supported Models" (vllm-project#17723) Signed-off-by: Harry Mellor <[email protected]> * [TPU] Increase block size and reset block shapes (vllm-project#16458) * [Misc] Add Next Edit Prediction (NEP) datasets support in `benchmark_serving.py` (vllm-project#16839) Signed-off-by: dtransposed <damian@damian-ml-machine.europe-west3-b.c.jetbrains-grazie.internal> Signed-off-by: dtransposed <> Co-authored-by: dtransposed <damian@damian-ml-machine.europe-west3-b.c.jetbrains-grazie.internal> * [Bugfix] Fix for the condition to accept empty encoder inputs for mllama (vllm-project#17732) Signed-off-by: Gregory Shtrasberg <[email protected]> * [Kernel] Unified Triton kernel that doesn't distinguish between prefill + decode (vllm-project#16828) Signed-off-by: Thomas Parnell <[email protected]> Signed-off-by: Lucas Wilkinson <[email protected]> Co-authored-by: Lucas Wilkinson <[email protected]> --------- Signed-off-by: Thomas Ortner <[email protected]> Signed-off-by: Stanislaw Wozniak <[email protected]> Signed-off-by: rzou <[email protected]> Signed-off-by: Mikhail Podvitskii <[email protected]> Signed-off-by: DarkLight1337 <[email protected]> Signed-off-by: Lucas Wilkinson <[email protected]> Signed-off-by: Harry Mellor <[email protected]> Signed-off-by: Mengqing Cao <[email protected]> Signed-off-by: mgoin <[email protected]> Signed-off-by: jiang1.li <[email protected]> Signed-off-by: jiang.li <[email protected]> Signed-off-by: Chen Zhang <[email protected]> Signed-off-by: Aaron Pham <[email protected]> Signed-off-by: windsonsea <[email protected]> Signed-off-by: reidliu41 <[email protected]> Signed-off-by: dtransposed <damian@damian-ml-machine.europe-west3-b.c.jetbrains-grazie.internal> Signed-off-by: dtransposed <> Signed-off-by: Gregory Shtrasberg <[email protected]> Signed-off-by: Thomas Parnell <[email protected]> Signed-off-by: [email protected] <[email protected]> Co-authored-by: Stan Wozniak <[email protected]> Co-authored-by: Thomas Ortner <[email protected]> Co-authored-by: Cyrus Leung <[email protected]> Co-authored-by: Tyler Michael Smith <[email protected]> Co-authored-by: Richard Zou <[email protected]> Co-authored-by: Mikhail Podvitskii <[email protected]> Co-authored-by: Cyrus Leung <[email protected]> Co-authored-by: Lucas Wilkinson <[email protected]> Co-authored-by: Harry Mellor <[email protected]> Co-authored-by: Mengqing Cao <[email protected]> Co-authored-by: Michael Goin <[email protected]> Co-authored-by: Li, Jiang <[email protected]> Co-authored-by: Chen Zhang <[email protected]> Co-authored-by: Aaron Pham <[email protected]> Co-authored-by: Michael Yao <[email protected]> Co-authored-by: Reid <[email protected]> Co-authored-by: reidliu41 <[email protected]> Co-authored-by: Jevin Jiang <[email protected]> Co-authored-by: d.transposed <[email protected]> Co-authored-by: dtransposed <damian@damian-ml-machine.europe-west3-b.c.jetbrains-grazie.internal> Co-authored-by: Gregory Shtrasberg <[email protected]> Co-authored-by: Thomas Parnell <[email protected]> Co-authored-by: Lucas Wilkinson <[email protected]>

Signed-off-by: Thomas Ortner <[email protected]> Signed-off-by: Stanislaw Wozniak <[email protected]> Co-authored-by: Thomas Ortner <[email protected]> Co-authored-by: Cyrus Leung <[email protected]> Co-authored-by: Tyler Michael Smith <[email protected]> Signed-off-by: Mu Huai <[email protected]>

Signed-off-by: Thomas Ortner <[email protected]> Signed-off-by: Stanislaw Wozniak <[email protected]> Co-authored-by: Thomas Ortner <[email protected]> Co-authored-by: Cyrus Leung <[email protected]> Co-authored-by: Tyler Michael Smith <[email protected]>

s3woz force-pushed the granitemoehybrid_clean branch from 55bc3e7 to 3569fe4 Compare April 30, 2025 21:09

s3woz marked this pull request as ready for review April 30, 2025 22:56

s3woz requested review from DarkLight1337 and ywang96 as code owners April 30, 2025 22:56

tlrmchlsmth reviewed Apr 30, 2025

View reviewed changes

alex-jw-brooks reviewed May 1, 2025

View reviewed changes

mergify bot added the needs-rebase label May 1, 2025

alex-jw-brooks reviewed May 1, 2025

View reviewed changes

Added test to functionally verify match between HF and vLLM

24bf59b

Signed-off-by: Thomas Ortner <[email protected]>

bohnstingl force-pushed the granitemoehybrid_clean branch from 590412a to 0dbb1b9 Compare May 2, 2025 05:48

mergify bot removed the needs-rebase label May 2, 2025

s3woz force-pushed the granitemoehybrid_clean branch 2 times, most recently from bbff670 to 0d4f3cd Compare May 2, 2025 07:57

s3woz and others added 11 commits May 2, 2025 08:42

GraniteMoeHybrid model

b7e89a0

Signed-off-by: Stanislaw Wozniak <[email protected]>

Removed MLA and added RoPE

e0136b3

Signed-off-by: Thomas Ortner <[email protected]>

Updated basic examples

4cfef42

Signed-off-by: Thomas Ortner <[email protected]>

TensorParallel and cleanup

859e473

Signed-off-by: Stanislaw Wozniak <[email protected]>

Fixing previous commit

bcb5e77

Signed-off-by: Stanislaw Wozniak <[email protected]>

Cleanup

18bb63d

Co-authored-by: Thomas Ortner <[email protected]> Signed-off-by: Stanislaw Wozniak <[email protected]>

Cleanup

7cb6a81

Co-authored-by: Thomas Ortner <[email protected]> Signed-off-by: Stanislaw Wozniak <[email protected]>

Removing sampler. Fixing URLs

b69ca16

Signed-off-by: Stanislaw Wozniak <[email protected]>

Fixing pre-hook errors.

6f8b1f4

Signed-off-by: Stanislaw Wozniak <[email protected]>

ruff reformatting pre-commit fix

c3b6460

Signed-off-by: Stanislaw Wozniak <[email protected]>

Skip tests until HF models become available

171532c

Signed-off-by: Stanislaw Wozniak <[email protected]>

DarkLight1337 reviewed May 5, 2025

View reviewed changes

docs/source/models/supported_models.md Outdated Show resolved Hide resolved

DarkLight1337 added the ready ONLY add when PR is ready to merge/full CI is needed label May 5, 2025

Update docs/source/models/supported_models.md

495fe33

Co-authored-by: Cyrus Leung <[email protected]> Signed-off-by: Stanislaw Wozniak <[email protected]>

DarkLight1337 enabled auto-merge (squash) May 5, 2025 07:55

auto-merge was automatically disabled May 5, 2025 07:55
Head branch was pushed to by a user without write access

s3woz force-pushed the granitemoehybrid_clean branch from f7de40b to 495fe33 Compare May 5, 2025 07:55

DarkLight1337 enabled auto-merge (squash) May 5, 2025 08:08

Commenting out failing HF registry test as code is not yet in HF used…

e114f1b

… in CI Signed-off-by: Stanislaw Wozniak <[email protected]>

auto-merge was automatically disabled May 5, 2025 09:32
Head branch was pushed to by a user without write access

DarkLight1337 reviewed May 5, 2025

View reviewed changes

tests/models/registry.py Outdated Show resolved Hide resolved

Temporarily marking model as is_available_online=False until it appea…

ca1d999

…rs in HF Transformers Signed-off-by: Stanislaw Wozniak <[email protected]>

Temporarily commenting out registration test until the model appears …

dc04497

…in HF Transformers Signed-off-by: Stanislaw Wozniak <[email protected]>

s3woz added 2 commits May 5, 2025 12:31

Fixing pre-commit error

a636074

Signed-off-by: Stanislaw Wozniak <[email protected]>

Marking model with next minor min_transformers_version

8c68720

Signed-off-by: Stanislaw Wozniak <[email protected]>

tlrmchlsmth reviewed May 5, 2025

View reviewed changes

tests/models/language/generation/test_granitemoehybrid.py Outdated Show resolved Hide resolved

tests/models/language/generation/test_hybrid.py Outdated Show resolved Hide resolved

Update tests/models/language/generation/test_hybrid.py

6c8e664

Co-authored-by: Tyler Michael Smith <[email protected]> Signed-off-by: Stanislaw Wozniak <[email protected]>

s3woz force-pushed the granitemoehybrid_clean branch from 711dcd4 to ca47978 Compare May 5, 2025 17:00

Clarifying comments

95ede00

Signed-off-by: Stanislaw Wozniak <[email protected]>

s3woz force-pushed the granitemoehybrid_clean branch from ca47978 to 95ede00 Compare May 5, 2025 20:46

tlrmchlsmth approved these changes May 5, 2025

View reviewed changes

DarkLight1337 merged commit 999328b into vllm-project:main May 6, 2025
52 checks passed

cyang49 mentioned this pull request May 6, 2025

[Model] Mamba2 causal conv1d Refactor to Split Prefill and Decode Requests for Corresponding Kernels #17146

Merged

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

[Model] Add GraniteMoeHybrid 4.0 model #17497

[Model] Add GraniteMoeHybrid 4.0 model #17497

s3woz commented Apr 30, 2025 •

edited by github-actions bot

Loading

github-actions bot commented Apr 30, 2025

tlrmchlsmth left a comment

tlrmchlsmth Apr 30, 2025

alex-jw-brooks Apr 30, 2025 •

edited

Loading

bohnstingl May 1, 2025

alex-jw-brooks May 1, 2025

alex-jw-brooks left a comment

alex-jw-brooks May 1, 2025

bohnstingl May 1, 2025

alex-jw-brooks May 1, 2025

alex-jw-brooks May 1, 2025

bohnstingl May 1, 2025

alex-jw-brooks May 1, 2025 •

edited

Loading

bohnstingl May 1, 2025

alex-jw-brooks May 1, 2025

bohnstingl May 1, 2025

DarkLight1337 commented May 1, 2025

mergify bot commented May 1, 2025

bohnstingl commented May 1, 2025

DarkLight1337 commented May 1, 2025

alex-jw-brooks May 1, 2025

bohnstingl May 2, 2025

alex-jw-brooks May 1, 2025

bohnstingl May 2, 2025

alex-jw-brooks May 2, 2025

DarkLight1337 May 2, 2025

alex-jw-brooks May 1, 2025

bohnstingl May 2, 2025

bohnstingl commented May 2, 2025

s3woz commented May 5, 2025

DarkLight1337 commented May 5, 2025

DarkLight1337 commented May 5, 2025

s3woz commented May 5, 2025 •

edited

Loading

s3woz commented May 5, 2025


		self.quant_config = vllm_config.quant_config

		super().__init__()

		)


		@pytest.mark.parametrize("model", SSM_MODELS[0:1] + HYBRID_MODELS[0:2])

[Model] Add GraniteMoeHybrid 4.0 model #17497

[Model] Add GraniteMoeHybrid 4.0 model #17497

Conversation

s3woz commented Apr 30, 2025 • edited by github-actions bot Loading

github-actions bot commented Apr 30, 2025

tlrmchlsmth left a comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

alex-jw-brooks Apr 30, 2025 • edited Loading

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

alex-jw-brooks left a comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

alex-jw-brooks May 1, 2025 • edited Loading

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

DarkLight1337 commented May 1, 2025

mergify bot commented May 1, 2025

bohnstingl commented May 1, 2025

DarkLight1337 commented May 1, 2025

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

bohnstingl commented May 2, 2025

s3woz commented May 5, 2025

DarkLight1337 commented May 5, 2025

DarkLight1337 commented May 5, 2025

s3woz commented May 5, 2025 • edited Loading

s3woz commented May 5, 2025

s3woz commented Apr 30, 2025 •

edited by github-actions bot

Loading

alex-jw-brooks Apr 30, 2025 •

edited

Loading

alex-jw-brooks May 1, 2025 •

edited

Loading

s3woz commented May 5, 2025 •

edited

Loading