[CI] Ignore ruff D209, rejoin the docstrings it split, and silence incompatible-rule warnings (#57212)
Signed-off-by: Harry Mellor <19981378+hmellor@users.noreply.github.com> Co-authored-by: Claude Opus 5 (1M context) <noreply@anthropic.com>
This commit is contained in:
co-authored by
Claude Opus 5
parent
1085b64425
commit
651a88c09a
@@ -42,8 +42,7 @@ def _run_warmup(monkeypatch, layers) -> list[dict]:
|
||||
@pytest.mark.parametrize("scale_name", ["weight_scale", "weight_scale_inv"])
|
||||
def test_registered_deep_gemm_layer_is_warmed_up(monkeypatch, scale_name) -> None:
|
||||
"""Any layer stamped by the DeepGEMM kernel is warmed, regardless of the
|
||||
quantization method that owns it or the name of its scale parameter.
|
||||
"""
|
||||
quantization method that owns it or the name of its scale parameter."""
|
||||
layer = _block_fp8_layer(256, 128, scale_name)
|
||||
|
||||
calls = _run_warmup(monkeypatch, [layer])
|
||||
@@ -55,8 +54,7 @@ def test_registered_deep_gemm_layer_is_warmed_up(monkeypatch, scale_name) -> Non
|
||||
|
||||
def test_n_multiple_of_64_matches_kernel_selection(monkeypatch) -> None:
|
||||
"""DeepGEMM accepts N % 64 == 0 (e.g. DeepSeek kv_a_proj_with_mqa, N=576),
|
||||
so warmup must not require N % 128 == 0.
|
||||
"""
|
||||
so warmup must not require N % 128 == 0."""
|
||||
assert len(_run_warmup(monkeypatch, [_block_fp8_layer(576, 256)])) == 1
|
||||
|
||||
|
||||
|
||||
Reference in New Issue
Block a user