Skip to content

fix: fallback to dot_product for autoselected attention on CPU to prevent Tokamax NotImplementedError #4617

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.

Sign up for GitHub

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
chiajunglien wants to merge 1 commit into
base: main
Choose a base branch
from
+1 −0
Open

fix: fallback to dot_product for autoselected attention on CPU to prevent Tokamax NotImplementedError #4617

Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
1 change: 1 addition & 0 deletions src/maxtext/layers/attention_op.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -1073,6 +1073,7 @@ def apply_attention(
self.attention_kernel == "dot_product"
or (self.attention_kernel == "autoselected" and model_mode == MODEL_MODE_AUTOREGRESSIVE)
or (self.attention_kernel == "autoselected" and length < 128)
or (self.attention_kernel == "autoselected" and target_hardware == "cpu")

@NuojCheng NuojCheng Jul 27, 2026

Copy link
Copy Markdown
Collaborator

Choose a reason for hiding this comment

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

this change gonna silently twist all results using AOT on CPU platform

@NuojCheng NuojCheng Jul 27, 2026

Copy link
Copy Markdown
Collaborator

Choose a reason for hiding this comment

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

I doubt this change can pass https://github.com/AI-Hypercomputer/maxtext/blob/main/tests/integration/aot_identical_test.py

@NuojCheng NuojCheng Jul 27, 2026

Copy link
Copy Markdown
Collaborator

Choose a reason for hiding this comment

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

somehow aot identical test does not show in https://github.com/AI-Hypercomputer/maxtext/actions/runs/30232605541/job/89875651384?pr=4617

@chiajunglien chiajunglien Jul 28, 2026

Copy link
Copy Markdown
Contributor Author

Choose a reason for hiding this comment

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

Hi @NuojCheng, thanks for the review! I've double-checked the AOT compilation behavior, and this change works safely without twisting results.

1. AOT correctly bypasses the fallback:
Even when cross-compiling on a CPU host, AOT (via train_compile.py) uses get_topology_desc(platform="tpu") to generate mock devices. These mock devices strictly retain device.platform == "tpu". Thus, target_hardware remains "tpu", properly preserving the flash or autoselected kernels without triggering a fallback to dot_product.

2. aot_identical_test.py passes flawlessly:
I just ran the test manually on my TPU VM and can confirm the generated HLO graphs remain identical and it passes perfectly:

  • tests/integration/aot_identical_test.py::AotHloIdenticalTest::test_default_hlo_match PASSED

Log: https://paste.googleplex.com/4608946768838656

If it’s missing in recent CI job logs, it’s likely because the test classes are wrapped with @pytest.mark.skip_on_tpu7x. If a GitHub Action runner happens to execute on a Trillium (tpu7x) node, Pytest naturally skips it.

Let me know if this clears up the concerns!

or (self.attention_kernel == "paged")
or (self.attention_kernel in ("vllm_rpa", "vllm_batched_rpa"))
):
Expand Down
Loading