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
+30 -31
View File
@@ -125,13 +125,12 @@ def smart_resize(
class KeyeImagePixelInputs(TensorSchema):
"""
Dimensions:
- bnp: Batch size * Number of patches
- c: Number of channels
- ps: Patch size
- ni: Number of images
- g: Grid dimensions (3 for t, h, w)
"""Dimensions:
- bnp: Batch size * Number of patches
- c: Number of channels
- ps: Patch size
- ni: Number of images
- g: Grid dimensions (3 for t, h, w)
"""
type: Literal["pixel_values"]
@@ -142,13 +141,12 @@ class KeyeImagePixelInputs(TensorSchema):
class KeyeImageEmbeddingInputs(TensorSchema):
"""
Dimensions:
- nf: Number of image features
- hs: Hidden size (must match the hidden size of language model
backbone)
- ni: Number of images
- g: Grid dimensions (3 for t, h, w)
"""Dimensions:
- nf: Number of image features
- hs: Hidden size (must match the hidden size of language model
backbone)
- ni: Number of images
- g: Grid dimensions (3 for t, h, w)
"""
type: Literal["image_embeds"]
@@ -160,13 +158,12 @@ KeyeImageInputs: TypeAlias = KeyeImagePixelInputs | KeyeImageEmbeddingInputs
class KeyeVideoPixelInputs(TensorSchema):
"""
Dimensions:
- bnp: Batch size * Number of patches
- c: Number of channels
- ps: Patch size
- ni: Number of images
- g: Grid dimensions (3 for t, h, w)
"""Dimensions:
- bnp: Batch size * Number of patches
- c: Number of channels
- ps: Patch size
- ni: Number of images
- g: Grid dimensions (3 for t, h, w)
"""
type: Literal["pixel_values_videos"]
@@ -177,13 +174,12 @@ class KeyeVideoPixelInputs(TensorSchema):
class KeyeVideoEmbeddingInputs(TensorSchema):
"""
Dimensions:
- nf: Number of video features
- hs: Hidden size (must match the hidden size of language model
backbone)
- nv: Number of videos
- g: Grid dimensions (3 for t, h, w)
"""Dimensions:
- nf: Number of video features
- hs: Hidden size (must match the hidden size of language model
backbone)
- nv: Number of videos
- g: Grid dimensions (3 for t, h, w)
"""
type: Literal["video_embeds"]
@@ -349,7 +345,8 @@ def apply_rotary_pos_emb_flashatt(
class KeyeSiglipAttention(nn.Module):
"""Multi-headed attention from 'Attention Is All You
Need' paper."""
Need' paper.
"""
def __init__(
self,
@@ -1465,6 +1462,9 @@ class BaseKeyeModule(nn.Module, SupportsMultiModal):
otherwise it will be `(seq_len,)`.
intermediate_tensors: Intermediate tensors from prior forward pass.
inputs_embeds: Optional tensor of input embeddings.
**kwargs: Multimodal inputs for this batch, forwarded to the
multimodal embedding path.
"""
if intermediate_tensors is not None:
inputs_embeds = None
@@ -1588,8 +1588,7 @@ class KeyeForConditionalGeneration(
def _split_video_grid_thw(
grid_thw: torch.Tensor | list[list[int]] | list[int],
) -> list[list[int]]:
"""
Split video grid_thw along the t dimension into per-frame rows.
"""Split video grid_thw along the t dimension into per-frame rows.
This preserves Keye's current M-RoPE behavior, where a video is emitted
as consecutive frame-level multimodal blocks rather than a single block