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
+11 -16
View File
@@ -88,9 +88,7 @@ def _get_vit_attn_backend(
*,
attn_backend_override: AttentionBackendEnum | None = None,
) -> AttentionBackendEnum:
"""
Get the available attention backend for Vision Transformer.
"""
"""Get the available attention backend for Vision Transformer."""
return current_platform.get_vit_attn_backend(
head_size,
dtype,
@@ -102,9 +100,7 @@ def get_vit_attn_backend(
head_size: int,
dtype: torch.dtype,
) -> AttentionBackendEnum:
"""
Get the attention backend for Vision Transformer.
"""
"""Get the attention backend for Vision Transformer."""
mm_cfg = get_multimodal_config()
attn_backend_override = (
mm_cfg.mm_encoder_attn_backend if mm_cfg is not None else None
@@ -120,7 +116,8 @@ def get_multimodal_config() -> MultiModalConfig | None:
"""Return the current ``MultiModalConfig``, or ``None`` when no engine
config context is active (e.g., during unit tests) or when the current
``model_config`` does not carry a ``multimodal_config`` (e.g., minimal
stubs used in tests)."""
stubs used in tests).
"""
vllm_config = get_current_vllm_config_or_none()
if vllm_config is None or vllm_config.model_config is None:
return None
@@ -142,9 +139,7 @@ def get_fp8_padded_hidden_size(num_heads: int, head_dim: int) -> int | None:
def is_vit_use_data_parallel(num_heads: int | None = None) -> bool:
"""
Get the tensor parallel type for Vision Transformer.
"""
"""Get the tensor parallel type for Vision Transformer."""
mm_cfg = get_multimodal_config()
can_split = (
num_heads % get_tensor_model_parallel_world_size() == 0
@@ -235,6 +230,7 @@ def resolve_visual_encoder_outputs(
concatenated with the other select_layers along the last dimension.
feature_select_strategy: Defines how to select the hidden states
from each layer.
"""
if select_layers is None:
if not isinstance(encoder_outputs, torch.Tensor):
@@ -302,10 +298,11 @@ def run_dp_sharded_vision_model(
Args:
image_input (torch.Tensor): Image input tensor.
vision_model (torch.nn.Module): Vision model.
Returns:
torch.Tensor: Output image embeddings
"""
"""
num_chunks = image_input.shape[0]
mp_world_size = get_tensor_model_parallel_world_size()
num_chunks_per_rank = (num_chunks + mp_world_size - 1) // mp_world_size
@@ -329,8 +326,7 @@ def get_load_balance_assignment(
sizes: list[int],
num_gpus: int = 2,
) -> tuple[list[int], list[int], list[int]]:
"""
Generate load balancing assignment and metadata
"""Generate load balancing assignment and metadata
for distributing data across GPUs.
The load is determined by the total image sizes,
not the number of images.
@@ -354,7 +350,6 @@ def get_load_balance_assignment(
```
"""
n_samples = len(sizes)
# Handle edge cases
@@ -414,6 +409,7 @@ def run_dp_sharded_mrope_vision_model(
Different rope types have different dimension to do ViT.
"rope_3d" for 3D rope (e.g., Qwen2.5-VL)
"rope_2d" for 2D rope (e.g., Kimi-VL)
Returns:
torch.Tensor: Output image embeddings
@@ -585,8 +581,7 @@ def run_dp_sharded_mrope_vision_model(
class FusedInputNorm(nn.Module):
"""
Module that applies rescaling and normalization to input images.
"""Module that applies rescaling and normalization to input images.
Equivalent to: output = (input * rescale_factor - mean) / std
"""