Skip to content

fix(perf): recognize RTX PRO 6000 for MFU #3598

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

Merged
yuhezhang-ai merged 3 commits into from
Aug 22, 2026
+13 −0
Merged

fix(perf): recognize RTX PRO 6000 for MFU #3598

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
3 commits
Select commit Hold shift + click to select a range
e58b139
fix(perf): recognize RTX PRO 6000 for MFU
ralovets Aug 20, 2026
f4eaa06
fix(perf): distinguish RTX PRO 6000 variants
ralovets Aug 20, 2026
e473443
Merge branch 'main' into ralovets/fix/rtx-pro-6000-mfu
akoumpa Aug 21, 2026
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
2 changes: 2 additions & 0 deletions nemo_automodel/_transformers/mfu.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -57,6 +57,8 @@
"910B": 354e12,
"Ascend910": 354e12,
"RTX 3070 Ti": 21.75e12,
"RTX PRO 6000 Blackwell Max-Q Workstation Edition": 438.9e12,
"RTX PRO 6000 Blackwell Workstation Edition": 503.8e12,
}

_UNIT_TO_SCALE = {
Expand Down
11 changes: 11 additions & 0 deletions tests/unit_tests/utils/test_flops_utils.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -318,6 +318,17 @@ def test_automfu_h100_reference_uses_dense_bf16_peak():
assert h100_tflops == 989.0


@pytest.mark.parametrize(
"device_name, expected_tflops",
[
("NVIDIA RTX PRO 6000 Blackwell Max-Q Workstation Edition", 438.9),
("NVIDIA RTX PRO 6000 Blackwell Workstation Edition", 503.8),
],
)
def test_get_device_flops_distinguishes_rtx_pro_6000_variants(device_name, expected_tflops):
assert get_device_flops(unit="T", device_name=device_name) == expected_tflops


def test_get_device_flops_detects_current_cuda_device(monkeypatch):
monkeypatch.setattr(mfu_module.torch.cuda, "is_available", lambda: True)
monkeypatch.setattr(mfu_module.torch.cuda, "current_device", lambda: 3)
Expand Down
Loading