Optimize JoinDims and SplitDims by canonicalizing to simpler operations (Partial fixes #1843) by mengxingbw · Pull Request #1847 · pymc-devs/pytensor

mengxingbw · 2026-01-13T22:22:55Z

Description

This PR implements the 3 out of 4 canonicalization rewrites suggested in #1843:

✅ join_dims(x, axis=axis, n_axes=1) → identity (no-op)
✅ join_dims(x, axis=axis, n_axes=0) → expand_dims(x, axis)
✅ split_dims(x, axis=axis, shape=()) → squeeze(x, axis)
⏳ split_dims(x, axis=axis, shape=(dim,)) → specify_shape(...) (see Block section)

Questions

I tried to work on the last requested change:

split_dims(x, axis=axis, shape=(dim,)) -> specify_shape(x, (*[None] * axis, dim, ...))

The issue: specify_shape preserves the input's known shape when it's already concrete, so it doesn't match SplitDims's output type. If the input already has a known shape at a dimension, it uses that shape; and it only uses the specified shape when the input shape is None. This has caused the function to fail.
For this rewrite to work even when the input shape is known, I'd need to use reshape instead of specify_shape, but that defeats the purpose of using specify_shape for shape assertion.

⚠️ Question: How should I proceed?

Related Issue

Closes Specialize (join|split)_dims rewrite #1843
Related to Specialize (join|split)_dims rewrite #1843

Checklist

Checked that the pre-commit linting/style checks pass
Included tests that prove the fix is effective or that the new feature works
Added necessary documentation (docstrings and/or example notebooks)
If you are a pro: each commit corresponds to a relevant logical change

Type of change

Optimize JoinDims and SplitDims by canonicalizing to simpler operations (identity, expand_dims, squeeze). Partial fixes pymc-devs#1843

jessegrabowski · 2026-01-14T03:45:14Z

My guess is that ricardo meant reshape, not literally specify_shape (which you're right, just adds metadata but doesn't do any computation)

ricardoV94 · 2026-01-14T16:27:02Z

I meant split dims, when the shape argument has just one entry shape.type.shape==(1,) is equivalent to specify shape on that axis.

That's what the syntax shape=(d,) was supposed to convey. As opposed to empy shape=() which is just a squeeze

ricardoV94 · 2026-01-14T16:28:15Z

pytensor/tensor/rewriting/reshape.py

+    x, shape = node.inputs
+    axis = node.op.axis
+
+    if isinstance(shape, Constant) and shape.data.size == 0:


Doesn't need to be constant just static shape of zero shape.type.shape == (0,)

Also I would merge this with the split-to-reshape rewrite so we don't accidentally run that before this

mbaldourw · 2026-01-14T19:40:44Z

Thank you @jessegrabowski and @ricardoV94 for clarifying - so it sounds like we don't need split_dims(x, axis=axis, shape=(dim,)) → specify_shape(...) this function since it will fall into reshape anyways?

I have made the changes according to the comment above.

ricardoV94 · 2026-01-14T19:47:54Z

reshape should be our last resort, everything we can avoid as reshape we should

ricardoV94 · 2026-01-14T19:53:32Z

To clarify, none of the changes in this PR were strictly needed, they are an improvement over simple reshape

mbaldourw · 2026-01-15T15:37:51Z

To clarify, none of the changes in this PR were strictly needed, they are an improvement over simple reshape

understood. is there anything else to do with the last function:
split_dims(x, axis=axis, shape=(dim,)) → specify_shape(...)?

ricardoV94

Making progress, needs a few more tweaks

pytensor/tensor/rewriting/reshape.py

ricardoV94 · 2026-01-16T12:48:18Z

pytensor/tensor/rewriting/reshape.py


+@register_canonicalize
+@node_rewriter([JoinDims])
+def local_join_dims_noop(fgraph, node):


merge these join dims rewrites in a single one, like we did with SplitDims

ricardoV94 · 2026-01-16T12:49:10Z

pytensor/tensor/rewriting/reshape.py

@@ -9,11 +11,24 @@
 def local_split_dims_to_reshape(fgraph, node):


Now that we don't do only reshape, we should have a more generic name. Same for the join_dims when we merge the special cases

Suggested change

def local_split_dims_to_reshape(fgraph, node):

def local_lower_split_dims(fgraph, node):

ricardoV94 · 2026-01-16T12:50:23Z

tests/tensor/rewriting/test_reshape.py

+    # After rewrite: should have 0 JoinDims nodes
+    assert sum([1 for node in fg.toposort() if isinstance(node.op, JoinDims)]) == 0
+    # Output should be equivalent to input (identity rewrite)
+    # The rewrite returns the input variable, so output should match input shape/type
+    assert fg.outputs[0].type.shape == x.type.shape
+    assert fg.outputs[0].type.dtype == x.type.dtype
+    assert fg.outputs[0].type.ndim == x.type.ndim


Use utt.assert_equal_computations to check we have the specific graph that we expect, not just anything without JoinDims

This recommendation applies to all new tests

This recommendation applies to all new tests

I can't seem to get it to pass for the first 2 tests. when i looked it up, i got "assert_equal_computations is better suited for cases where the canonical form is a specific operation (like expand_dims, squeeze, or identity) where graph structures match. For basic reshape cases, the rewrite produces a different but equivalent graph structure, so structural checks are sufficient"
Please let me know how to proceed!

Suggest you look at the generated graph, the utility prints it when the assert fails. It shouldn't have anything too strange in it

after excluding:

--------------------------------------- Captured stdout call ---------------------------------------- rewriting: rewrite local_split_dims replaces SplitDims{axis=1}.0 of SplitDims{axis=1}(x, [2 5 1]) with Reshape{5}.0 of Reshape{5}(x, MakeVector{dtype='int64'}.0) rewriting: rewrite MergeOptimizer replaces 2 of None with 2 of None rewriting: rewrite MergeOptimizer replaces 0 of None with 0 of None rewriting: rewrite MergeOptimizer replaces 1 of None with 1 of None rewriting: rewrite MergeOptimizer replaces Shape.0 of Shape(x) with Shape.0 of Shape(x) rewriting: rewrite MergeOptimizer replaces 0 of None with 0 of None rewriting: rewrite MergeOptimizer replaces 2 of None with 2 of None rewriting: rewrite local_subtensor_remove_broadcastable_index replaces Subtensor{i}.0 of Subtensor{i}(Subtensor{:stop}.0, 0) with Squeeze{axis=0}.0 of Squeeze{axis=0}(Subtensor{:stop}.0) rewriting: rewrite local_subtensor_remove_broadcastable_index replaces Subtensor{i}.0 of Subtensor{i}(Subtensor{start:}.0, 0) with Squeeze{axis=0}.0 of Squeeze{axis=0}(Subtensor{start:}.0) rewriting: rewrite constant_folding replaces Subtensor{i}.0 of Subtensor{i}([2 5 1], 2) with 1 of None rewriting: rewrite constant_folding replaces Subtensor{i}.0 of Subtensor{i}([2 5 1], 1) with 5 of None rewriting: rewrite constant_folding replaces Subtensor{i}.0 of Subtensor{i}([2 5 1], 0) with 2 of None rewriting: rewrite local_reshape_to_dimshuffle replaces Reshape{5}.0 of Reshape{5}(x, MakeVector{dtype='int64'}.0) with ExpandDims{axis=3}.0 of ExpandDims{axis=3}(Reshape{4}.0)```

And how does the rewritten graph look like now (vs the expected)?

E E Rewritten: E SpecifyShape [id A] <Tensor5(float64, shape=(2, 2, 5, 1, 3))> E ├─ ExpandDims{axis=3} [id B] <Tensor5(float64, shape=(?, 2, 5, 1, ?))> E │ └─ Reshape{4} [id C] <Tensor4(float64, shape=(?, 2, 5, ?))> E │ ├─ x [id D] <Tensor3(float64, shape=(2, 10, 3))> E │ └─ MakeVector{dtype='int64'} [id E] <Vector(int64, shape=(4,))> E │ ├─ Squeeze{axis=0} [id F] <Scalar(int64, shape=())> E │ │ └─ Subtensor{:stop} [id G] <Vector(int64, shape=(1,))> E │ │ ├─ Shape [id H] <Vector(int64, shape=(3,))> E │ │ │ └─ x [id D] <Tensor3(float64, shape=(2, 10, 3))> E │ │ └─ 1 [id I] <int64> E │ ├─ 2 [id J] <Scalar(int64, shape=())> E │ ├─ 5 [id K] <Scalar(int64, shape=())> E │ └─ Squeeze{axis=0} [id L] <Scalar(int64, shape=())> E │ └─ Subtensor{start:} [id M] <Vector(int64, shape=(1,))> E │ ├─ Shape [id H] <Vector(int64, shape=(3,))> E │ │ └─ ··· E │ └─ 2 [id N] <int64> E ├─ 2 [id O] <Scalar(int8, shape=())> E ├─ 2 [id O] <Scalar(int8, shape=())> E ├─ 5 [id P] <Scalar(int8, shape=())> E ├─ 1 [id Q] <Scalar(int8, shape=())> E └─ 3 [id R] <Scalar(int8, shape=())> E E Expected: E ExpandDims{axis=3} [id A] <Tensor5(float64, shape=(2, 2, 5, 1, 3))> E └─ Reshape{4} [id B] <Tensor4(float64, shape=(2, 2, 5, 3))> E ├─ x [id C] <Tensor3(float64, shape=(2, 10, 3))> E └─ MakeVector{dtype='int64'} [id D] <Vector(int64, shape=(4,))> E ├─ Subtensor{i} [id E] <Scalar(int64, shape=())> E │ ├─ Shape [id F] <Vector(int64, shape=(3,))> E │ │ └─ x [id C] <Tensor3(float64, shape=(2, 10, 3))> E │ └─ 0 [id G] <int64> E ├─ Cast{int64} [id H] <Scalar(int64, shape=())> E │ └─ 2 [id I] <Scalar(int8, shape=())> E ├─ Cast{int64} [id J] <Scalar(int64, shape=())> E │ └─ 5 [id K] <Scalar(int8, shape=())> E └─ Subtensor{i} [id L] <Scalar(int64, shape=())> E ├─ Shape [id F] <Vector(int64, shape=(3,))> E │ └─ ··· E └─ 2 [id M] <int64>```

removing local_subtensor_remove_broadcastable_index should bring you closer, and using np.int64(2|5) for the expected shape. That will get rid of the Cast thing, which comes from #1073

E Rewritten: E SpecifyShape [id A] <Tensor5(float64, shape=(2, 2, 5, 1, 3))> E ├─ ExpandDims{axis=3} [id B] <Tensor5(float64, shape=(?, 2, 5, 1, ?))> E │ └─ Reshape{4} [id C] <Tensor4(float64, shape=(?, 2, 5, ?))> E │ ├─ x [id D] <Tensor3(float64, shape=(2, 10, 3))> E │ └─ MakeVector{dtype='int64'} [id E] <Vector(int64, shape=(4,))> E │ ├─ Subtensor{i} [id F] <Scalar(int64, shape=())> E │ │ ├─ Subtensor{:stop} [id G] <Vector(int64, shape=(1,))> E │ │ │ ├─ Shape [id H] <Vector(int64, shape=(3,))> E │ │ │ │ └─ x [id D] <Tensor3(float64, shape=(2, 10, 3))> E │ │ │ └─ 1 [id I] <int64> E │ │ └─ 0 [id J] <int64> E │ ├─ 2 [id K] <Scalar(int64, shape=())> E │ ├─ 5 [id L] <Scalar(int64, shape=())> E │ └─ Subtensor{i} [id M] <Scalar(int64, shape=())> E │ ├─ Subtensor{start:} [id N] <Vector(int64, shape=(1,))> E │ │ ├─ Shape [id H] <Vector(int64, shape=(3,))> E │ │ │ └─ ··· E │ │ └─ 2 [id O] <int64> E │ └─ 0 [id J] <int64> E ├─ 2 [id P] <Scalar(int8, shape=())> E ├─ 2 [id P] <Scalar(int8, shape=())> E ├─ 5 [id Q] <Scalar(int8, shape=())> E ├─ 1 [id R] <Scalar(int8, shape=())> E └─ 3 [id S] <Scalar(int8, shape=())> E E Expected: E ExpandDims{axis=3} [id A] <Tensor5(float64, shape=(2, 2, 5, 1, 3))> E └─ Reshape{4} [id B] <Tensor4(float64, shape=(2, 2, 5, 3))> E ├─ x [id C] <Tensor3(float64, shape=(2, 10, 3))> E └─ MakeVector{dtype='int64'} [id D] <Vector(int64, shape=(4,))> E ├─ Subtensor{i} [id E] <Scalar(int64, shape=())> E │ ├─ Shape [id F] <Vector(int64, shape=(3,))> E │ │ └─ x [id C] <Tensor3(float64, shape=(2, 10, 3))> E │ └─ 0 [id G] <int64> E ├─ 2 [id H] <Scalar(int64, shape=())> E ├─ 5 [id I] <Scalar(int64, shape=())> E └─ Subtensor{i} [id J] <Scalar(int64, shape=())> E ├─ Shape [id F] <Vector(int64, shape=(3,))> E │ └─ ··· E └─ 2 [id K] <int64>

ricardoV94

Looks good.

However you are basically doing the same checks twice in the tests, the assert_equal_computations already checks the specific Ops are in the graph (and none unexpected ones are in it), so the redundant checks can be removed.

ricardoV94 · 2026-02-05T11:08:01Z

tests/tensor/rewriting/test_reshape.py

+    # After rewrite: should have 0 JoinDims nodes
+    assert sum([1 for node in fg.toposort() if isinstance(node.op, JoinDims)]) == 0
+    # Should have 1 DimShuffle node with is_expand_dims=True
+    expand_nodes = [
+        node
+        for node in fg.toposort()
+        if isinstance(node.op, DimShuffle) and node.op.is_expand_dims
+    ]
+    assert len(expand_nodes) == 1


This is all already covered by the assert_equal_computations

Suggested change

# After rewrite: should have 0 JoinDims nodes

assert sum([1 for node in fg.toposort() if isinstance(node.op, JoinDims)]) == 0

# Should have 1 DimShuffle node with is_expand_dims=True

expand_nodes = [

node

for node in fg.toposort()

if isinstance(node.op, DimShuffle) and node.op.is_expand_dims

]

assert len(expand_nodes) == 1

ricardoV94 · 2026-02-05T11:08:53Z

tests/tensor/rewriting/test_reshape.py

+    # After rewrite: should have 0 JoinDims nodes
+    assert sum([1 for node in fg.toposort() if isinstance(node.op, JoinDims)]) == 0


Already covered by assert_equal_computations

Suggested change

# After rewrite: should have 0 JoinDims nodes

assert sum([1 for node in fg.toposort() if isinstance(node.op, JoinDims)]) == 0

still missing this

ricardoV94 · 2026-02-05T11:09:13Z

tests/tensor/rewriting/test_reshape.py

    assert sum([1 for node in fg.toposort() if isinstance(node.op, Reshape)]) == 1
    assert fg.outputs[0].type.shape == (2, 10, 3)


Already covered by assert_equal_computations

Suggested change

assert sum([1 for node in fg.toposort() if isinstance(node.op, Reshape)]) == 1

assert fg.outputs[0].type.shape == (2, 10, 3)

ricardoV94 · 2026-02-05T11:09:31Z

tests/tensor/rewriting/test_reshape.py

    assert sum([1 for node in fg.toposort() if isinstance(node.op, SplitDims)]) == 0
    assert sum([1 for node in fg.toposort() if isinstance(node.op, Reshape)]) == 1


Already covered by assert_equal_computations

Suggested change

assert sum([1 for node in fg.toposort() if isinstance(node.op, SplitDims)]) == 0

assert sum([1 for node in fg.toposort() if isinstance(node.op, Reshape)]) == 1

ricardoV94 · 2026-02-05T11:10:03Z

tests/tensor/rewriting/test_reshape.py

+    # After rewrite: should have 0 SplitDims nodes
+    assert sum([1 for node in fg.toposort() if isinstance(node.op, SplitDims)]) == 0
+    # Should have 1 DimShuffle node with is_squeeze=True (not Reshape)
+    squeeze_nodes = [
+        node
+        for node in fg.toposort()
+        if isinstance(node.op, DimShuffle) and node.op.is_squeeze
+    ]
+    assert len(squeeze_nodes) == 1
+    # Should NOT have a Reshape node
+    assert sum([1 for node in fg.toposort() if isinstance(node.op, Reshape)]) == 0


Already covered by assert_equal_computations

Suggested change

# After rewrite: should have 0 SplitDims nodes

assert sum([1 for node in fg.toposort() if isinstance(node.op, SplitDims)]) == 0

# Should have 1 DimShuffle node with is_squeeze=True (not Reshape)

squeeze_nodes = [

node

for node in fg.toposort()

if isinstance(node.op, DimShuffle) and node.op.is_squeeze

]

assert len(squeeze_nodes) == 1

# Should NOT have a Reshape node

assert sum([1 for node in fg.toposort() if isinstance(node.op, Reshape)]) == 0

ricardoV94 · 2026-02-05T11:10:33Z

tests/tensor/rewriting/test_reshape.py

    assert sum([1 for node in fg.toposort() if isinstance(node.op, SplitDims)]) == 1

-    rewrite_graph(fg, include=("canonicalize",))
+    with pt.config.change_flags(optimizer_verbose=True):


Just for debug, should be removed. You'll have to dedent the inner code

Suggested change

with pt.config.change_flags(optimizer_verbose=True):

ricardoV94 · 2026-02-05T11:11:25Z

pytensor/tensor/rewriting/reshape.py

+        *[x.shape[i] for i in range(axis)],
        *shape,
-        *x.shape[axis + 1 :],
+        *[x.shape[i] for i in range(axis + 1, x.type.ndim)],


The previous code is cleaner, this creates repeated Shape Ops in the list comprehenstion that will have to be removed.

The previous code is cleaner, this creates repeated Shape Ops in the list comprehenstion that will have to be removed.

That was to fix a test fail. For example, when I switched to the previous code, this happened:

E AssertionError: equal_computations failed E E Rewritten: E SpecifyShape [id A] <Tensor5(float64, shape=(2, 2, 5, 1, 3))> E ├─ ExpandDims{axis=3} [id B] <Tensor5(float64, shape=(?, 2, 5, 1, ?))> E │ └─ Reshape{4} [id C] <Tensor4(float64, shape=(?, 2, 5, ?))> E │ ├─ x [id D] <Tensor3(float64, shape=(2, 10, 3))> E │ └─ MakeVector{dtype='int64'} [id E] <Vector(int64, shape=(4,))> E │ ├─ Subtensor{i} [id F] <Scalar(int64, shape=())> E │ │ ├─ Subtensor{:stop} [id G] <Vector(int64, shape=(1,))> E │ │ │ ├─ Shape [id H] <Vector(int64, shape=(3,))> E │ │ │ │ └─ x [id D] <Tensor3(float64, shape=(2, 10, 3))> E │ │ │ └─ 1 [id I] <int64> E │ │ └─ 0 [id J] <int64> E │ ├─ 2 [id K] <Scalar(int64, shape=())> E │ ├─ 5 [id L] <Scalar(int64, shape=())> E │ └─ Subtensor{i} [id M] <Scalar(int64, shape=())> E │ ├─ Subtensor{start:} [id N] <Vector(int64, shape=(1,))> E │ │ ├─ Shape [id H] <Vector(int64, shape=(3,))> E │ │ │ └─ ··· E │ │ └─ 2 [id O] <int64> E │ └─ 0 [id J] <int64> E ├─ 2 [id P] <Scalar(int8, shape=())> E ├─ 2 [id P] <Scalar(int8, shape=())> E ├─ 5 [id Q] <Scalar(int8, shape=())> E ├─ 1 [id R] <Scalar(int8, shape=())> E └─ 3 [id S] <Scalar(int8, shape=())> E E Expected: E ExpandDims{axis=3} [id A] <Tensor5(float64, shape=(2, 2, 5, 1, 3))> E └─ Reshape{4} [id B] <Tensor4(float64, shape=(2, 2, 5, 3))> E ├─ x [id C] <Tensor3(float64, shape=(2, 10, 3))> E └─ MakeVector{dtype='int64'} [id D] <Vector(int64, shape=(4,))> E ├─ Subtensor{i} [id E] <Scalar(int64, shape=())> E │ ├─ Shape [id F] <Vector(int64, shape=(3,))> E │ │ └─ x [id C] <Tensor3(float64, shape=(2, 10, 3))> E │ └─ 0 [id G] <int64> E ├─ 2 [id H] <Scalar(int64, shape=())> E ├─ 5 [id I] <Scalar(int64, shape=())> E └─ Subtensor{i} [id J] <Scalar(int64, shape=())> E ├─ Shape [id F] <Vector(int64, shape=(3,))> E │ └─ ··· E └─ 2 [id K] <int64>

We shouldn't change the actual code for the sake of the tests (but the other way around)

We shouldn't change the actual code for the sake of the tests (but the other way around)

that makes sense!! just updated the test

ricardoV94 · 2026-02-08T02:26:20Z

tests/tensor/rewriting/test_reshape.py

+    )
+
+
+def test_local_join_dims():


Now that the dust has settled, we can revert this test to how it looked before (including the name). The ones with Reshape are a bit too messy for the assert_equal_computations approach. This isn't affected by your PR anyway

ricardoV94 · 2026-02-08T02:26:41Z

tests/tensor/rewriting/test_reshape.py



-def test_local_split_dims_to_reshape():
+def test_local_split_dims():


same here, revert test changes and name. Keep only the new tests you added as they are

Add canonicalization rewrites for JoinDims/SplitDims

04e391e

Optimize JoinDims and SplitDims by canonicalizing to simpler operations (identity, expand_dims, squeeze). Partial fixes pymc-devs#1843

mengxingbw changed the title ~~Add canonicalization rewrites for JoinDims/SplitDims~~ Optimize JoinDims and SplitDims by canonicalizing to simpler operations (Partial fixes #1843) Jan 13, 2026

ricardoV94 reviewed Jan 14, 2026

View reviewed changes

mbaldourw added 2 commits January 14, 2026 14:33

removed constant; merged func

9a5e536

removed specify shape

273ecc3

ricardoV94 requested changes Jan 16, 2026

View reviewed changes

mbaldourw added 4 commits January 16, 2026 14:28

merged functions; updated test

0037dfa

updated tests

cacda87

fixed splitdims test; updated function def

69943d5

fixed join dims def and test

1c77c0e

ricardoV94 reviewed Feb 5, 2026

View reviewed changes

mbaldourw added 2 commits February 6, 2026 10:09

removed redundant assertion in test

104ab1a

restored changes to reshape and updated test

e26249c

ricardoV94 reviewed Feb 8, 2026

View reviewed changes

		@@ -9,11 +11,24 @@
		def local_split_dims_to_reshape(fgraph, node):

	def local_split_dims_to_reshape(fgraph, node):
	def local_lower_split_dims(fgraph, node):

		# After rewrite: should have 0 JoinDims nodes
		assert sum([1 for node in fg.toposort() if isinstance(node.op, JoinDims)]) == 0

		assert sum([1 for node in fg.toposort() if isinstance(node.op, Reshape)]) == 1
		assert fg.outputs[0].type.shape == (2, 10, 3)



		def test_local_split_dims_to_reshape():
		def test_local_split_dims():

Conversation

mengxingbw commented Jan 13, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Description

Questions

⚠️ Question: How should I proceed?

Related Issue

Checklist

Type of change

Uh oh!

jessegrabowski commented Jan 14, 2026

Uh oh!

ricardoV94 commented Jan 14, 2026

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

mbaldourw commented Jan 14, 2026

Uh oh!

ricardoV94 commented Jan 14, 2026

Uh oh!

ricardoV94 commented Jan 14, 2026

Uh oh!

mbaldourw commented Jan 15, 2026

Uh oh!

ricardoV94 left a comment

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

ricardoV94 left a comment

Choose a reason for hiding this comment

Uh oh!

ricardoV94 Feb 5, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

mengxingbw commented Jan 13, 2026 •

edited

Loading

ricardoV94 Feb 5, 2026 •

edited

Loading

ricardoV94 Feb 8, 2026 •

edited

Loading