ns for fx: add F.conv2d and F.conv3d weight extraction

[vkuzo]

Stack from ghstack:

• ns for fx: fill out coverage for node I/O types #55918 ns for fx: fill out coverage for node I/O types
• ns for fx: add category for ops which accept fp32 or int8 input #55859 ns for fx: add category for ops which accept fp32 or int8 input
• ns for fx: move op dtype category mapping to separate file #55858 ns for fx: move op dtype category mapping to separate file
• ns for fx: add test for op relationship coverage #55837 ns for fx: add test for op relationship coverage
• ns for fx: move pattern utils to separate file #55805 ns for fx: move pattern utils to separate file
• ns for fx: make get_reversed_fusions reuse quantization fusions #55803 ns for fx: make get_reversed_fusions reuse quantization fusions
• ns for fx: add qat tests cases for shadowed activations #55614 ns for fx: add qat tests cases for shadowed activations
• ns for fx: qat test cases for unshadowed activations #55508 ns for fx: qat test cases for unshadowed activations
• ns for fx: extend functional weight extraction testing to QAT #55507 ns for fx: extend functional weight extraction testing to QAT
• ns for fx: add qat handling for weight extraction #55506 ns for fx: add qat handling for weight extraction
• ns for fx: make NSTracer inherit from QuantizationTracer #55505 ns for fx: make NSTracer inherit from QuantizationTracer
• ns for fx: change node I/O determination to strict allowlist #55434 ns for fx: change node I/O determination to strict allowlist
• ns for fx: add functional conv-relu fusion support #55433 ns for fx: add functional conv-relu fusion support
• ns for fx: add test case for linear dynamic #55432 ns for fx: add test case for linear dynamic
• ns for fx: fix bug in weight extraction testing #55431 ns for fx: fix bug in weight extraction testing
• ns for fx: add conv1d weight extraction #55327 ns for fx: add conv1d weight extraction
• fx quant: fix prepacking for F.conv1d #55311 fx quant: fix prepacking for F.conv1d
• ns for fx: move more weight matching logic to weight_utils.py #55288 ns for fx: move more weight matching logic to weight_utils.py
• ns for fx: add F.conv2d and F.conv3d weight extraction #55287 ns for fx: add F.conv2d and F.conv3d weight extraction
• ns for fx: rename SugraphTypeRelationship to SubgraphTypeRelationship #55155 ns for fx: rename SugraphTypeRelationship to SubgraphTypeRelationship
• ns for fx: add allowlist for ops with same signature across dtypes #55154 ns for fx: add allowlist for ops with same signature across dtypes
• ns for fx: add linear-relu mod weight extraction #55080 ns for fx: add linear-relu mod weight extraction

Summary:

Adds support for extracting weights from F.conv2d and F.conv3d.
F.conv1d and the fused variants are saved for future PRs.

Test Plan:

python test/test_quantization.py TestFXNumericSuiteCoreAPIs.test_extract_weights_conv_fun

Reviewers:

Subscribers:

Tasks:

Tags:

Differential Revision: D27575424

[facebook-github-bot]

💊 CI failures summary and remediations

As of commit 952981a (more details on the Dr. CI page):

---

• 2/2 failures possibly* introduced in this PR
  • 1/2 non-scanned failure(s)

---

1 failure not recognized by patterns:

Job	Step	Action
pytorch_linux_bionic_rocm3_9_py3_6_build	Build	🔁 rerun

---

1 job timed out:

• pytorch_linux_bionic_rocm3_9_py3_6_build

---

This comment was automatically generated by Dr. CI (expand for details).

Follow this link to opt-out of these comments for your Pull Requests.

Please report bugs/suggestions to the (internal) Dr. CI Users group.

[facebook-github-bot]

This pull request has been merged in 13d7b40.