[Kernel] Add b12x FP4 MoE backend (#52018)

Signed-off-by: Luke Alonso <lalonso@gmail.com>
Signed-off-by: mgoin <mgoin64@gmail.com>
Signed-off-by: Misha Goin <mgoin64@gmail.com>
Co-authored-by: mgoin <mgoin64@gmail.com>
This commit is contained in:
Luke Alonso
2026-08-21 08:04:36 -07:00
committed by GitHub
co-authored by mgoin
parent 5ee84d3c52
commit 2740c817ff
23 changed files with 2417 additions and 60 deletions
+9 -3
View File
@@ -181,8 +181,14 @@ def test_b12x_warmup_deduplicates_registered_signatures(monkeypatch) -> None:
worker = SimpleNamespace(
get_model=lambda: SimpleNamespace(modules=modules),
scheduler_config=SimpleNamespace(max_num_batched_tokens=8),
scheduler_config=SimpleNamespace(
max_num_batched_tokens=8,
max_num_scheduled_tokens=32,
),
model_config=SimpleNamespace(dtype=torch.float32),
vllm_config=SimpleNamespace(
compilation_config=SimpleNamespace(compile_sizes=[4, 16])
),
)
platform = SimpleNamespace(
is_cuda=lambda: True,
@@ -200,7 +206,7 @@ def test_b12x_warmup_deduplicates_registered_signatures(monkeypatch) -> None:
assert scans == 1
assert calls == [
("first", (1, 2, 8), torch.bfloat16),
("second", (1, 2, 8), torch.bfloat16),
("first", (1, 2, 4, 8, 16, 32), torch.bfloat16),
("second", (1, 2, 4, 8, 16, 32), torch.bfloat16),
]
assert synchronized == [True]