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:
co-authored by
Claude Opus 5
parent
804e5377aa
commit
8c1557a79c
@@ -53,17 +53,16 @@ _MAX_FRAMES_PER_VIDEO = 16
|
||||
|
||||
|
||||
class LlavaOnevisionVideoPixelInputs(TensorSchema):
|
||||
"""
|
||||
Dimensions:
|
||||
- bn: Batch size * number of videos
|
||||
- f: Number of frames
|
||||
- c: Number of channels (3)
|
||||
- h: Height
|
||||
- w: Width
|
||||
"""Dimensions:
|
||||
- bn: Batch size * number of videos
|
||||
- f: Number of frames
|
||||
- c: Number of channels (3)
|
||||
- h: Height
|
||||
- w: Width
|
||||
|
||||
Note that `f` may be different for each batch, and 'num_frames'
|
||||
may be different for each video, in which case the data is passed as a
|
||||
list instead of a batched tensor.
|
||||
Note that `f` may be different for each batch, and 'num_frames'
|
||||
may be different for each video, in which case the data is passed as a
|
||||
list instead of a batched tensor.
|
||||
"""
|
||||
|
||||
type: Literal["pixel_values_videos"] = "pixel_values_videos"
|
||||
@@ -75,16 +74,15 @@ class LlavaOnevisionVideoPixelInputs(TensorSchema):
|
||||
|
||||
|
||||
class LlavaOnevisionImagePixelInputs(TensorSchema):
|
||||
"""
|
||||
Dimensions:
|
||||
- bn: Batch size * number of images
|
||||
- np: Number of patches (1 + num_patches)
|
||||
- c: Number of channels (3)
|
||||
- h: Height
|
||||
- w: Width
|
||||
"""Dimensions:
|
||||
- bn: Batch size * number of images
|
||||
- np: Number of patches (1 + num_patches)
|
||||
- c: Number of channels (3)
|
||||
- h: Height
|
||||
- w: Width
|
||||
|
||||
Note that `num_patches` may be different per batch and image,
|
||||
in which case the data is passed as a list instead of a batched tensor.
|
||||
Note that `num_patches` may be different per batch and image,
|
||||
in which case the data is passed as a list instead of a batched tensor.
|
||||
"""
|
||||
|
||||
type: Literal["pixel_values"] = "pixel_values"
|
||||
@@ -98,11 +96,10 @@ class LlavaOnevisionImagePixelInputs(TensorSchema):
|
||||
|
||||
|
||||
class LlavaOnevisionImageEmbeddingInputs(TensorSchema):
|
||||
"""
|
||||
Dimensions:
|
||||
- bn: Batch size * number of images
|
||||
- ifs: Image feature size
|
||||
- hs: Hidden size (must match language model backbone)
|
||||
"""Dimensions:
|
||||
- bn: Batch size * number of images
|
||||
- ifs: Image feature size
|
||||
- hs: Hidden size (must match language model backbone)
|
||||
"""
|
||||
|
||||
type: Literal["image_embeds"] = "image_embeds"
|
||||
@@ -557,8 +554,7 @@ class LlavaOnevisionForConditionalGeneration(nn.Module, SupportsMultiModal, Supp
|
||||
def _parse_and_validate_video_input(
|
||||
self, **kwargs: object
|
||||
) -> LlavaOnevisionVideoPixelInputs | None:
|
||||
"""
|
||||
A legal video input should have the following dimensions:
|
||||
"""A legal video input should have the following dimensions:
|
||||
{
|
||||
"pixel_values_videos" :
|
||||
list[b, Tensor(nb_frames, nb_channels, height, width)]
|
||||
@@ -891,10 +887,17 @@ class LlavaOnevisionForConditionalGeneration(nn.Module, SupportsMultiModal, Supp
|
||||
**kwargs: object,
|
||||
) -> torch.Tensor | IntermediateTensors:
|
||||
"""Run forward pass for LlaVA-Onevision.
|
||||
|
||||
Args:
|
||||
input_ids: Flattened (concatenated) input_ids corresponding to a
|
||||
batch.
|
||||
pixel_values_videos: Pixels in each frames for each input videos.
|
||||
positions: Flattened (concatenated) position ids corresponding to a
|
||||
batch.
|
||||
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
|
||||
|
||||
Reference in New Issue
Block a user