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
@@ -166,8 +166,8 @@ def _init_kv_cache_quant(
layer: The attention layer instance to initialize.
quant_config: Optional quantization configuration.
prefix: Layer name prefix for quantization method lookup.
"""
"""
# Note [Register q/k/v/prob scales in state dict]
# When calling model.to(device), only parameters/buffers in state dict are
# moved. If not registering q/k/v/prob scales in state dict, there would
@@ -254,8 +254,7 @@ class Attention(nn.Module, AttentionLayerBase):
head_size_v: int | None = None,
**extra_impl_args,
) -> None:
"""
The KV cache is stored inside this class and is accessed via
"""The KV cache is stored inside this class and is accessed via
`self.kv_cache`.
"""
super().__init__()
@@ -493,8 +492,7 @@ class Attention(nn.Module, AttentionLayerBase):
output_shape: torch.Size | None = None,
output_dtype: torch.dtype | None = None,
) -> torch.Tensor:
"""
The KV cache is stored inside this class and is accessed via
"""The KV cache is stored inside this class and is accessed via
`self.kv_cache`.
Attention metadata (`attn_metadata`) is set using a context manager in
@@ -691,6 +689,7 @@ def get_attention_context(
Note: attn_metadata may be None, but attn_layer and kv_cache are always
extracted from the forward context.
"""
forward_context: ForwardContext = get_forward_context()
attn_metadata_raw = forward_context.attn_metadata
@@ -718,8 +717,7 @@ def unified_kv_cache_update(
value: torch.Tensor,
layer_name: LayerNameType,
) -> torch.Tensor:
"""
Returns a dummy that is passed to unified_attention to signal a side effect and
"""Returns a dummy that is passed to unified_attention to signal a side effect and
the data dependency between them to ensure torch.compile preserves ordering.
"""
layer_name = _resolve_layer_name(layer_name)