Skip to content

Update urls for pytorch documentation #2521

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Open
wants to merge 3 commits into
base: main
Choose a base branch
from
Open

Update urls for pytorch documentation #2521

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
3 commits
Select commit Hold shift + click to select a range
c4f6e78
Update urls for pytorch documentation
xadupre Aug 27, 2025
7de7ea6
docs
xadupre Aug 27, 2025
dd22584
fix
xadupre Aug 27, 2025
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
12 changes: 6 additions & 6 deletions docs/articles/2023-06-20-introducing-onnx-script/index.md
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -326,12 +326,12 @@ Finally, a _huge thank you_ to the wonderful engineering team at Microsoft that
[onnx-expand-operator]: https://onnx.ai/onnx/operators/onnx__Expand.html#expand-13
[onnxscript-pypi]: https://pypi.org/project/onnxscript
[onnxscript-github]: https://github.com/microsoft/onnxscript
[torch-onnx]: https://pytorch.org/docs/stable/onnx.html
[torch-ir]: https://pytorch.org/docs/stable/ir.html
[torch-dynamo]: https://pytorch.org/docs/stable/dynamo/index.html
[torch-onnx-dynamoexport]: https://pytorch.org/docs/main/onnx.html#preview-torch-onnx-torchdynamo-exporter
[torch-onnx-customops]: https://pytorch.org/docs/stable/onnx.html#onnx-script-functions
[torch-chunk]: https://pytorch.org/docs/stable/generated/torch.chunk.html
[torch-onnx]: https://docs.pytorch.org/docs/stable/onnx.html
[torch-ir]: https://docs.pytorch.org/docs/stable/ir.html
[torch-dynamo]: https://docs.pytorch.org/docs/stable/dynamo/index.html
[torch-onnx-dynamoexport]: https://docs.pytorch.org/docs/stable/onnx.html#preview-torch-onnx-torchdynamo-exporter
[torch-onnx-customops]: https://docs.pytorch.org/docs/stable/onnx.html#onnx-script-functions
[torch-chunk]: https://docs.pytorch.org/docs/stable/generated/torch.chunk.html
[netron]: https://netron.app
[numpy]: https://numpy.org
[pdb]: https://docs.python.org/3/library/pdb.html
2 changes: 1 addition & 1 deletion docs/conf.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -92,7 +92,7 @@
"onnx_ir": ("https://onnx.ai/ir-py/", None),
"onnxruntime": ("https://onnxruntime.ai/docs/api/python/", None),
"scipy": ("https://docs.scipy.org/doc/scipy/", None),
"torch": ("https://pytorch.org/docs/main/", None),
"torch": ("https://docs.pytorch.org/docs/stable/", None),
}

# -- Options for Sphinx Gallery ----------------------------------------------
Expand Down
8 changes: 4 additions & 4 deletions onnxscript/function_libs/torch_lib/ops/core.py
View file
Open in desktop
Copy link
Collaborator

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

We can still keep the changes here and in the sphinx docs

Original file line number Diff line number Diff line change
Expand Up @@ -243,7 +243,7 @@ def aten_addmm(
alpha = float(alpha)
beta = float(beta)

# addmm only accepts 2d tensors: https://pytorch.org/docs/stable/generated/torch.addmm.html
# addmm only accepts 2d tensors: https://docs.pytorch.org/docs/stable/generated/torch.addmm.html
return op.Gemm(mat1, mat2, self, alpha=alpha, beta=beta)


Expand Down Expand Up @@ -3710,7 +3710,7 @@ def aten_frac(self: TFloat) -> TFloat:
Computes the fractional portion of each element in input.
"""

# https://pytorch.org/docs/stable/generated/torch.frac.html
# https://docs.pytorch.org/docs/stable/generated/torch.frac.html
return op.Sub(self, op.Mul(op.Floor(op.Abs(self)), op.Sign(self)))


Expand Down Expand Up @@ -6311,7 +6311,7 @@ def aten_native_layer_norm(
) -> Tuple[TReal, TReal, TReal]:
"""native_layer_norm(Tensor input, SymInt[] normalized_shape, Tensor? weight, Tensor? bias, float eps) -> (Tensor, Tensor, Tensor)"""

# https://pytorch.org/docs/stable/generated/torch.nn.LayerNorm.html#torch.nn.LayerNorm
# https://docs.pytorch.org/docs/stable/generated/torch.nn.LayerNorm.html#torch.nn.LayerNorm
# The mean and standard-deviation are calculated over the last D dimensions,
# where D is the dimension of normalized_shape. For example, if normalized_shape is
# (3, 5) (a 2-dimensional shape), the mean and standard-deviation are computed
Expand Down Expand Up @@ -7901,7 +7901,7 @@ def aten_slice_scatter(
# And, 'end' also must be specified, and end-start must be equal to the size of 'src'
# Assert(end-start == shape(src) > 0)
# Try torch sample to get more information:
# https://pytorch.org/docs/master/generated/torch.slice_scatter.html?highlight=slice_scatter#torch.slice_scatter
# https://docs.pytorch.org/master/generated/torch.slice_scatter.html?highlight=slice_scatter#torch.slice_scatter
# Take (torch.zeros(8, 8), torch.ones(2, 8), 0, 6, 64, 1) as example:
# Step 1: get 1D tensor from 0 to dim_size-1, then Slice it using start, end and step.
# We cannot use Range(start, end, step) directly as start or end may out of range.
Expand Down
8 changes: 4 additions & 4 deletions onnxscript/function_libs/torch_lib/ops/nn.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -1754,7 +1754,7 @@ def aten_scaled_dot_product_attention(
) -> TFloat:
"""scaled_dot_product_attention(Tensor query, Tensor key, Tensor value, Tensor? attn_mask=None, float dropout_p=0.0, bool is_causal=False, *, float? scale=None, bool enable_gqa=False) -> Tensor

Reference: https://pytorch.org/docs/stable/generated/torch.nn.functional.scaled_dot_product_attention.html
Reference: https://docs.pytorch.org/docs/stable/generated/torch.nn.functional.scaled_dot_product_attention.html

Equivalent to the PyTorch code::
scale_factor = 1 / math.sqrt(Q.size(-1)) if scale is None else scale
Expand All @@ -1776,7 +1776,7 @@ def aten_scaled_dot_product_attention(
"conversion of scaled_dot_product_attention not implemented if enable_gqa is True"
)

# Reference: https://pytorch.org/docs/stable/generated/torch.nn.functional.scaled_dot_product_attention.html
# Reference: https://docs.pytorch.org/docs/stable/generated/torch.nn.functional.scaled_dot_product_attention.html
if scale is None:
scale = _attention_scale(query)
scale = op.CastLike(scale, query)
Expand Down Expand Up @@ -1825,7 +1825,7 @@ def aten__scaled_dot_product_flash_attention(
"""_scaled_dot_product_flash_attention(Tensor query, Tensor key, Tensor value, float dropout_p=0.0, bool is_causal=False, bool return_debug_mask=False, *, float? scale=None) -> (Tensor output, Tensor logsumexp, Tensor cum_seq_q, Tensor cum_seq_k, int max_q, int max_k, Tensor philox_seed, Tensor philox_offset, Tensor debug_attn_mask)

One of the implementations of scaled_dot_product_attention.
Reference: https://pytorch.org/docs/stable/generated/torch.nn.functional.scaled_dot_product_attention.html
Reference: https://docs.pytorch.org/docs/stable/generated/torch.nn.functional.scaled_dot_product_attention.html

NOTE: Currently, there are three implementations of nn.scaled_dot_product_attention in PyTorch due to optimization.
However, it's the same implementation from ONNX perspective.
Expand Down Expand Up @@ -1964,7 +1964,7 @@ def aten_scaled_dot_product_attention_bool_mask(
) -> TFloat:
"""scaled_dot_product_attention(Tensor query, Tensor key, Tensor value, Tensor? attn_mask=None, float dropout_p=0.0, bool is_causal=False, *, float? scale=None, bool enable_gqa=False) -> Tensor

Reference: https://pytorch.org/docs/stable/generated/torch.nn.functional.scaled_dot_product_attention.html
Reference: https://docs.pytorch.org/docs/stable/generated/torch.nn.functional.scaled_dot_product_attention.html

Equivalent to the PyTorch code::
scale_factor = 1 / math.sqrt(Q.size(-1)) if scale is None else scale
Expand Down
4 changes: 2 additions & 2 deletions onnxscript/function_libs/torch_lib/ops/special.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -343,7 +343,7 @@ def aten_special_sinc(self: TFloat) -> TFloat:
"""special_sinc(Tensor self) -> Tensor"""

# This computes the normalized sinc, where the input is multiplied by pi.
# https://pytorch.org/docs/stable/special.html#torch.special.sinc
# https://docs.pytorch.org/docs/stable/special.html#torch.special.sinc
pi_self = self * _MATH_PI

return op.Where(self == 0.0, op.CastLike(1, self), op.Sin(pi_self) / pi_self)
Expand All @@ -365,7 +365,7 @@ def aten_special_xlog1py(self: TensorType, other: TensorType) -> TensorType:
def aten_special_xlogy(self: TFloat, other: TFloat) -> TFloat:
"""special_xlogy(Tensor self, Tensor other) -> Tensor"""

# https://pytorch.org/docs/stable/special.html#torch.special.xlogy
# https://docs.pytorch.org/docs/stable/special.html#torch.special.xlogy
# out := {
# NaN if other == NaN
# 0 if self == 0
Expand Down
2 changes: 1 addition & 1 deletion onnxscript/function_libs/torch_lib/tensor_typing.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -26,7 +26,7 @@
)

# NOTE: We do not care about unsigned types beyond UINT8 because PyTorch does not us them.
# More detail can be found: https://pytorch.org/docs/stable/tensors.html
# More detail can be found: https://docs.pytorch.org/docs/stable/tensors.html

_TensorType = Union[
BFLOAT16,
Expand Down
7 changes: 3 additions & 4 deletions onnxscript/onnx_opset/_impl/opset12.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -7,8 +7,7 @@
# --------------------------------------------------------------------------
# pylint: disable=W0221,W0222,R0901,W0237
# mypy: disable-error-code=override
# ruff: noqa: N801,E741
# ruff: noqa: D214,D402,D405,D411,D412,D416,D417
# ruff: noqa: D402
# --------------------------------------------------------------------------

from __future__ import annotations
Expand Down Expand Up @@ -666,7 +665,7 @@ def MaxPool(
subset of the input tensor according to the kernel size and downsampling the
data into the output tensor Y for further processing. The output spatial shape is calculated differently
depending on whether explicit padding is used, where pads is employed, or auto padding is used, where auto_pad is utilized.
With explicit padding (https://pytorch.org/docs/stable/generated/torch.nn.MaxPool2d.html?highlight=maxpool#torch.nn.MaxPool2d):
With explicit padding (https://docs.pytorch.org/docs/stable/generated/torch.nn.MaxPool2d.html?highlight=maxpool#torch.nn.MaxPool2d):
```
output_spatial_shape[i] = floor((input_spatial_shape[i] + pad_shape[i] - dilation[i] * (kernel_shape[i] - 1) - 1) / strides_spatial_shape[i] + 1)
```
Expand Down Expand Up @@ -820,7 +819,7 @@ def NegativeLogLikelihoodLoss(
sum(loss) / sum(weight[target[n][d_1][d_2]...[d_k]]]), for all samples.
If "reduction" attribute is set to "sum", the output is a scalar:
sum(loss).
See also https://pytorch.org/docs/stable/nn.html#torch.nn.NLLLoss.
See also https://docs.pytorch.org/docs/stable/nn.html#torch.nn.NLLLoss.
Example 1:
// negative log likelihood loss, "none" reduction
N, C, d1 = 2, 3, 2
Expand Down
5 changes: 2 additions & 3 deletions onnxscript/onnx_opset/_impl/opset13.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -7,8 +7,7 @@
# --------------------------------------------------------------------------
# pylint: disable=W0221,W0222,R0901,W0237
# mypy: disable-error-code=override
# ruff: noqa: N801,E741
# ruff: noqa: D214,D402,D405,D411,D412,D416,D417
# ruff: noqa: D214, D402, D405, D411, D416, D417
# --------------------------------------------------------------------------

from __future__ import annotations
Expand Down Expand Up @@ -2058,7 +2057,7 @@ def NegativeLogLikelihoodLoss(

If "reduction" attribute is set to "sum", the output is a scalar: `sum(loss)`.

See also https://pytorch.org/docs/stable/nn.html#torch.nn.NLLLoss.
See also https://docs.pytorch.org/docs/stable/nn.html#torch.nn.NLLLoss.

Example 1:

Expand Down
38 changes: 3 additions & 35 deletions onnxscript/onnx_opset/_impl/opset16.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -7,8 +7,7 @@
# --------------------------------------------------------------------------
# pylint: disable=W0221,W0222,R0901,W0237
# mypy: disable-error-code=override
# ruff: noqa: N801,E741
# ruff: noqa: D214,D402,D405,D411,D412,D416,D417
# ruff: noqa: D214, D402, D405, D411, D416
# --------------------------------------------------------------------------

from __future__ import annotations
Expand Down Expand Up @@ -128,7 +127,7 @@
They are used to interpolate output values of `Y[N, C, H_out, W_out]`.

The GridSample operator is often used in doing grid generator and sampler in the [Spatial Transformer Networks](https://arxiv.org/abs/1506.02025).
See also in [torch.nn.functional.grid_sample](https://pytorch.org/docs/master/generated/torch.nn.functional.grid_sample.html#torch-nn-functional-grid-sample).
See also in [torch.nn.functional.grid_sample](https://docs.pytorch.org/master/generated/torch.nn.functional.grid_sample.html#torch-nn-functional-grid-sample).


Args:
Expand Down Expand Up @@ -253,38 +252,7 @@
B_If: TypeAlias = BOOL

V_If: TypeAlias = Union[
Optional[Sequence[BFLOAT16]],
Optional[Sequence[BOOL]],
Optional[Sequence[COMPLEX128]],
Optional[Sequence[COMPLEX64]],
Optional[Sequence[DOUBLE]],
Optional[Sequence[FLOAT]],
Optional[Sequence[FLOAT16]],
Optional[Sequence[INT16]],
Optional[Sequence[INT32]],
Optional[Sequence[INT64]],
Optional[Sequence[INT8]],
Optional[Sequence[STRING]],
Optional[Sequence[UINT16]],
Optional[Sequence[UINT32]],
Optional[Sequence[UINT64]],
Optional[Sequence[UINT8]],
Optional[BFLOAT16],
Optional[BOOL],
Optional[COMPLEX128],
Optional[COMPLEX64],
Optional[DOUBLE],
Optional[FLOAT],
Optional[FLOAT16],
Optional[INT16],
Optional[INT32],
Optional[INT64],
Optional[INT8],
Optional[STRING],
Optional[UINT16],
Optional[UINT32],
Optional[UINT64],
Optional[UINT8],
None,

Check warning

Code scanning / lintrunner

RUFF/RUF036 Warning

None not at the end of the type annotation..
See https://docs.astral.sh/ruff/rules/none-not-at-end-of-union
Sequence[BFLOAT16],
Sequence[BOOL],
Sequence[COMPLEX128],
Expand Down
5 changes: 2 additions & 3 deletions onnxscript/onnx_opset/_impl/opset18.py
View file
Open in desktop
Copy link
Collaborator

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

onnxscript/onnx_opset/_impl is auto generated - we cannot create manual changes to them. I would suggest reverting

Original file line number Diff line number Diff line change
Expand Up @@ -7,8 +7,7 @@
# --------------------------------------------------------------------------
# pylint: disable=W0221,W0222,R0901,W0237
# mypy: disable-error-code=override
# ruff: noqa: N801,E741
# ruff: noqa: D214,D402,D405,D411,D412,D416,D417
# ruff: noqa: D402, D405
# --------------------------------------------------------------------------

from __future__ import annotations
Expand Down Expand Up @@ -235,7 +234,7 @@ def Col2Im(

The operator rearranges column blocks back into a multidimensional image

Col2Im behaves similarly to PyTorch's fold https://pytorch.org/docs/stable/generated/torch.nn.Fold.html,
Col2Im behaves similarly to PyTorch's fold https://docs.pytorch.org/docs/stable/generated/torch.nn.Fold.html,
but it only supports *batched* multi-dimensional image tensors.
Another implementation in Python with N-dimension support can be found at https://github.com/f-dangel/unfoldNd/.

Expand Down
50 changes: 7 additions & 43 deletions onnxscript/onnx_opset/_impl/opset19.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -7,8 +7,7 @@
# --------------------------------------------------------------------------
# pylint: disable=W0221,W0222,R0901,W0237
# mypy: disable-error-code=override
# ruff: noqa: N801,E741
# ruff: noqa: D214,D402,D405,D411,D412,D416,D417
# ruff: noqa: D214, D402, D405, D411, D412, D416
# --------------------------------------------------------------------------

from __future__ import annotations
Expand Down Expand Up @@ -72,7 +71,7 @@
subset of the input tensor according to the kernel size and downsampling the
data into the output tensor Y for further processing. The output spatial shape is calculated differently
depending on whether explicit padding is used, where pads is employed, or auto padding is used, where auto_pad is utilized.
With explicit padding (https://pytorch.org/docs/stable/generated/torch.nn.MaxPool2d.html?highlight=maxpool#torch.nn.MaxPool2d):
With explicit padding (https://docs.pytorch.org/docs/stable/generated/torch.nn.MaxPool2d.html?highlight=maxpool#torch.nn.MaxPool2d):
```
output_spatial_shape[i] = floor((input_spatial_shape[i] + pad_shape[i] - dilation[i] * (kernel_shape[i] - 1) - 1) / strides_spatial_shape[i] + 1)
```
Expand Down Expand Up @@ -701,53 +700,14 @@
B_If: TypeAlias = BOOL

V_If: TypeAlias = Union[
Optional[Sequence[BFLOAT16]],
Optional[Sequence[BOOL]],
Optional[Sequence[COMPLEX128]],
Optional[Sequence[COMPLEX64]],
Optional[Sequence[DOUBLE]],
Optional[Sequence[FLOAT]],
Optional[Sequence[FLOAT16]],
Optional[Sequence[INT16]],
Optional[Sequence[INT32]],
Optional[Sequence[INT64]],
Optional[Sequence[INT8]],
Optional[Sequence[STRING]],
Optional[Sequence[UINT16]],
Optional[Sequence[UINT32]],
Optional[Sequence[UINT64]],
Optional[Sequence[UINT8]],
Optional[BFLOAT16],
Optional[BOOL],
Optional[COMPLEX128],
Optional[COMPLEX64],
Optional[DOUBLE],
Optional[FLOAT],
Optional[FLOAT16],
Optional[FLOAT8E4M3FN],
Optional[FLOAT8E4M3FNUZ],
Optional[FLOAT8E5M2],
Optional[FLOAT8E5M2FNUZ],
Optional[INT16],
Optional[INT32],
Optional[INT64],
Optional[INT8],
Optional[STRING],
Optional[UINT16],
Optional[UINT32],
Optional[UINT64],
Optional[UINT8],
None,

Check warning

Code scanning / lintrunner

RUFF/RUF036 Warning

None not at the end of the type annotation..
See https://docs.astral.sh/ruff/rules/none-not-at-end-of-union
Sequence[BFLOAT16],
Sequence[BOOL],
Sequence[COMPLEX128],
Sequence[COMPLEX64],
Sequence[DOUBLE],
Sequence[FLOAT],
Sequence[FLOAT16],
Sequence[FLOAT8E4M3FN],
Sequence[FLOAT8E4M3FNUZ],
Sequence[FLOAT8E5M2],
Sequence[FLOAT8E5M2FNUZ],
Sequence[INT16],
Sequence[INT32],
Sequence[INT64],
Expand Down Expand Up @@ -777,6 +737,10 @@
UINT32,
UINT64,
UINT8,
Sequence[FLOAT8E4M3FN],
Sequence[FLOAT8E4M3FNUZ],
Sequence[FLOAT8E5M2],
Sequence[FLOAT8E5M2FNUZ],
]

def If(self, cond: B_If, *, else_branch: GraphProto, then_branch: GraphProto) -> V_If:
Expand Down
7 changes: 3 additions & 4 deletions onnxscript/onnx_opset/_impl/opset20.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -7,8 +7,7 @@
# --------------------------------------------------------------------------
# pylint: disable=W0221,W0222,R0901,W0237
# mypy: disable-error-code=override
# ruff: noqa: N801,E741
# ruff: noqa: D214,D402,D405,D411,D412,D416,D417
# ruff: noqa: D402
# --------------------------------------------------------------------------

from __future__ import annotations
Expand Down Expand Up @@ -60,7 +59,7 @@ def AffineGrid(


Generates a 2D or 3D flow field (sampling grid), given a batch of affine matrices theta
(https://pytorch.org/docs/stable/generated/torch.nn.functional.affine_grid.html).
(https://docs.pytorch.org/docs/stable/generated/torch.nn.functional.affine_grid.html).
An affine matrix `theta` is applied to a position tensor represented in its homogeneous expression. Here is an example in 3D:
::

Expand Down Expand Up @@ -304,7 +303,7 @@ def GridSample(

The GridSample operator is often used in doing grid generator and sampler in the
[Spatial Transformer Networks](https://arxiv.org/abs/1506.02025).
See also in [torch.nn.functional.grid_sample](https://pytorch.org/docs/stable/generated/torch.nn.functional.grid_sample.html).
See also in [torch.nn.functional.grid_sample](https://docs.pytorch.org/docs/stable/generated/torch.nn.functional.grid_sample.html).


Args:
Expand Down
Loading
Loading