Add pydocstyle to the ruff rules (#52136)

Signed-off-by: Harry Mellor <19981378+hmellor@users.noreply.github.com>
Signed-off-by: Harry Mellor <hej.mellor@gmail.com>
Co-authored-by: Claude Opus 5 (1M context) <noreply@anthropic.com>
This commit is contained in:
Harry Mellor
2026-09-16 16:44:29 +00:00
committed by GitHub
co-authored by Claude Opus 5
parent 804e5377aa
commit 8c1557a79c
2002 changed files with 15024 additions and 14269 deletions
+6 -13
View File
@@ -56,8 +56,7 @@ def create_expert_indices_with_redundancy(
total_physical_experts: int,
redundancy_config: list[int], # redundancy for each logical expert
) -> torch.Tensor:
"""
Create expert indices with redundancy.
"""Create expert indices with redundancy.
Args:
num_layers: number of layers
@@ -67,6 +66,7 @@ def create_expert_indices_with_redundancy(
Returns:
indices: Shape (num_layers, total_physical_experts)
"""
assert sum(redundancy_config) == total_physical_experts
assert len(redundancy_config) == num_logical_experts
@@ -95,8 +95,7 @@ def create_expert_weights(
device: torch.device,
physical_to_logical_mapping: torch.Tensor,
) -> list[list[torch.Tensor]]:
"""
Create fake expert weights tensor for testing.
"""Create fake expert weights tensor for testing.
Use `arange` to generate predictable weights values, based on logical
expert ID.
@@ -105,6 +104,7 @@ def create_expert_weights(
Args:
physical_to_logical_mapping: Shape (num_layers, num_local_experts)
mapping[layer, physical_pos] = logical_expert_id
"""
expert_weights = []
@@ -209,9 +209,7 @@ def verify_redundant_experts_have_same_weights(
world_size: int,
num_local_experts: int,
) -> bool:
"""
Verify that all replicas of the same logical expert have the same weights.
"""
"""Verify that all replicas of the same logical expert have the same weights."""
num_layers = len(expert_weights)
total_physical_experts = world_size * num_local_experts
@@ -568,7 +566,6 @@ def test_rearrange_expert_weights_with_redundancy(
eplb_communicator,
):
"""Test the functionality of rearranging expert weights with redundancy."""
if eplb_communicator == "nixl" and not has_nixl():
pytest.skip("NIXL is not available")
if torch.accelerator.device_count() < world_size:
@@ -678,7 +675,6 @@ def test_async_transfer_layer_without_mtp(
eplb_communicator: str,
):
"""Exercise async EPLB transfer path without MTP/spec decode."""
if eplb_communicator == "nixl" and not has_nixl():
pytest.skip("NIXL is not available")
if torch.accelerator.device_count() < world_size:
@@ -696,11 +692,9 @@ def test_async_transfer_layer_without_mtp(
@pytest.mark.parametrize("world_size", [2, 4])
def test_rearrange_expert_weights_no_change(world_size):
"""
Test that when the indices do not change, the weights should remain
"""Test that when the indices do not change, the weights should remain
unchanged.
"""
if torch.accelerator.device_count() < world_size:
pytest.skip(f"Need at least {world_size} GPUs to run the test")
distributed_run(
@@ -800,7 +794,6 @@ def _test_rearrange_expert_weights_profile_mode(env, world_size) -> None:
@pytest.mark.parametrize("world_size", [2, 4])
def test_rearrange_expert_weights_profile_mode(world_size):
"""Test profile mode (should not copy actual weights)"""
if torch.accelerator.device_count() < world_size:
pytest.skip(f"Need at least {world_size} GPUs to run the test")
distributed_run(