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
+14 -20
View File
@@ -41,13 +41,12 @@ from .utils import AutoWeightsLoader, WeightsMapper, maybe_prefix
class AriaImagePixelInputs(TensorSchema):
"""
Dimensions:
- b: Batch size
- n: Number of images
- c: Number of channels
- h: Height of each image
- w: Width of each image
"""Dimensions:
- b: Batch size
- n: Number of images
- c: Number of channels
- h: Height of each image
- w: Width of each image
"""
type: Literal["pixel_values"]
@@ -119,8 +118,7 @@ class AriaProjectorMLP(nn.Module):
class AriaProjector(nn.Module):
"""
A projection module with one cross attention layer and one FFN layer, which
"""A projection module with one cross attention layer and one FFN layer, which
projects ViT's outputs into MoE's inputs.
Args:
@@ -129,6 +127,7 @@ class AriaProjector(nn.Module):
Outputs:
A tensor with the shape of (batch_size, query_number, output_dim)
"""
def __init__(self, config: AriaConfig, prefix: str = "") -> None:
@@ -187,8 +186,7 @@ class AriaProjector(nn.Module):
class AriaTextMoELayer(nn.Module):
"""
Mixture of Experts (MoE) Layer for the AriaMoE model.
"""Mixture of Experts (MoE) Layer for the AriaMoE model.
This layer implements the MoE mechanism, which routes input tokens to
different experts based on a routing algorithm, processes them through the
@@ -227,8 +225,7 @@ class AriaTextMoELayer(nn.Module):
)
def forward(self, hidden_states: torch.Tensor) -> torch.Tensor:
"""
Forward pass of the MoE Layer.
"""Forward pass of the MoE Layer.
Args:
hidden_states: Input tensor of shape
@@ -236,16 +233,15 @@ class AriaTextMoELayer(nn.Module):
Returns:
torch.Tensor: Output tensor after passing through the MoE layer.
"""
"""
router_output = torch.nn.functional.linear(hidden_states, self.router_weight)
return self.experts(hidden_states, router_output)
class AriaTextDecoderLayer(LlamaDecoderLayer):
"""
Custom Decoder Layer for the AriaMoE model which modifies the standard
"""Custom Decoder Layer for the AriaMoE model which modifies the standard
`LlamaDecoderLayer` by replacing the traditional MLP with a Mixture of
Experts (MoE) Layer.
"""
@@ -262,8 +258,7 @@ class AriaTextDecoderLayer(LlamaDecoderLayer):
class AriaTextModel(LlamaModel, SupportsQuant):
"""
Custom LlamaModel for the AriaMoE model which modifies the standard
"""Custom LlamaModel for the AriaMoE model which modifies the standard
LlamaModel by replacing the `LlamaDecoderLayer` with `MoEDecoderLayer`.
"""
@@ -377,8 +372,7 @@ class AriaMultiModalProcessor(BaseMultiModalProcessor[AriaProcessingInfo]):
dummy_inputs=AriaDummyInputsBuilder,
)
class AriaForConditionalGeneration(nn.Module, SupportsMultiModal):
"""
Aria model for conditional generation tasks.
"""Aria model for conditional generation tasks.
This model combines a vision tower, a multi-modal projector, and a language
model to perform tasks that involve both image and text inputs.