feat: upgrade to ONNX 1.20.1, PaddlePaddle 3.3.0, and modernize tooling

.ai-rulez/config.yaml

@@ -0,0 +1,22 @@
+$schema: https://raw.githubusercontent.com/Goldziher/ai-rulez/main/schema/ai-rules-v3.schema.json
+description: Convert PaddlePaddle models to ONNX format
+gitignore: true
+name: Paddle2ONNX
+
+builtins:
+  - python
+  - security
+  - git-workflow
+  - code-quality
+  - testing
+  - token-efficiency
+  - default-commands
+
+presets:
+  - claude
+  - gemini
+  - codex
+
+compression:
+  level: moderate
+version: "3.0"

.gitmodules

@@ -1,15 +1,15 @@
 [submodule "third_party/onnx"]
 	path = third_party/onnx
 	url = https://github.com/onnx/onnx.git
-	branch = 990217f043af7222348ca8f0301e17fa7b841781
+	branch = d3f6b795aedb48eaecc881bf5e8f5dd6efbe25b3
 [submodule "third_party/optimizer"]
 	path = third_party/optimizer
 	url = https://github.com/onnx/optimizer.git
-	branch = b3a4611861734e0731bbcc2bed1f080139e4988b
+	branch = 94d238d96e3fb3a7ba34f03c284b9ad3516163be
 [submodule "third_party/pybind11"]
 	path = third_party/pybind11
 	url = https://github.com/pybind/pybind11.git
-	branch = 3e9dfa2866941655c56877882565e7577de6fc7b
+	branch = 45fab4087eaaff234227a10cf7845e8b07f28a98
 [submodule "third_party/glog"]
 	path = third_party/glog
 	url = https://github.com/google/glog.git

.pre-commit-config.yaml

@@ -1,111 +1,54 @@
-# Exclude all third-party libraries
-exclude: |
-    (?x)^(
-        patches/.+|
-        third_party/.+
-    )$
+default_install_hook_types:
+  - pre-commit
+  - commit-msg
+exclude: ^third_party/|vendor/|node_modules/|dist/|\.setuptools-cmake-build/|patches/
+
 repos:
-# Common hooks
--   repo: https://github.com/pre-commit/pre-commit-hooks
-    rev: v4.4.0
-    hooks:
-    -   id: check-added-large-files
-    -   id: check-merge-conflict
-    -   id: check-symlinks
-    -   id: detect-private-key
-    -   id: end-of-file-fixer
-    -   id: trailing-whitespace
--   repo: https://github.com/Lucas-C/pre-commit-hooks.git
-    rev: v1.5.1
-    hooks:
-    -   id: remove-crlf
-    -   id: remove-tabs
-        name: Tabs remover (C++)
-        files: \.(c|cc|cxx|cpp|cu|h|hpp|hxx|xpu|kps)$
-        args: [--whitespaces-count, '2']
-    -   id: remove-tabs
-        name: Tabs remover (Python)
-        files: (.*\.(py|bzl)|BUILD|.*\.BUILD|WORKSPACE)$
-        args: [--whitespaces-count, '4']
-        # Exclude some unit test files that require tabs.
--   repo: local
-    hooks:
-    -   id: copyright_checker
-        name: copyright_checker
-        entry: python ./tools/codestyle/copyright.py
-        language: system
-        files: \.(c|cc|cxx|cpp|cu|h|hpp|hxx|proto|xpu|kps|py|pyi|sh)$
-# For Python files
--   repo: https://github.com/psf/black-pre-commit-mirror
-    rev: 24.8.0
-    hooks:
-    -   id: black
--   repo: https://github.com/astral-sh/ruff-pre-commit
-    rev: v0.5.0
-    hooks:
-    -   id: ruff
-        args: [--fix, --exit-non-zero-on-fix, --no-cache]
-# For C++ files
--   repo: local
-    hooks:
-    -   id: clang-format
-        name: clang-format
-        description: Format files with ClangFormat.
-        entry: bash ./tools/codestyle/clang_format.sh -i
-        language: system
-        files: \.(c|cc|cxx|cpp|cu|h|hpp|hxx|xpu|kps)$
--   repo: local
-    hooks:
-    -   id: cpplint-cpp-source
-        name: cpplint
-        description: Check C++ code style using cpplint.py.
-        entry: bash ./tools/codestyle/cpplint_pre_commit.sh
-        language: system
-        files: \.(cc|cxx|cpp|cu|h|hpp|hxx)$
-        args:
-            - --extensions=cc,cxx,cpp,cu,cuh,h,hpp,hxx,kps
-            - --filter=-readability/fn_size,-build/include_what_you_use,-build/c++11,-whitespace/parens
-            - --quiet
--   repo: local
-    hooks:
-    -   id: clang-tidy
-        name: clang-tidy
-        description: Parallel clang-tidy runner.
-        entry: python ./tools/codestyle/clang-tidy.py
-        language: system
-        files: \.(c|cc|cxx|cpp|h|hpp|hxx)$
-        args:
-            - -p=build/
-            - -extra-arg=-Wno-unknown-warning-option
-            - -extra-arg=-Wno-pessimizing-move
-            - -extra-arg=-Wno-braced-scalar-init
-            - -extra-arg=-Wno-dangling-gsl
-            - -extra-arg=-Wno-deprecated-copy
-            - -extra-arg=-Wno-final-dtor-non-final-class
-            - -extra-arg=-Wno-implicit-int-float-conversion
-            - -extra-arg=-Wno-inconsistent-missing-override
-            - -extra-arg=-Wno-infinite-recursion
-            - -extra-arg=-Wno-mismatched-tags
-            - -extra-arg=-Wno-self-assign
-            - -extra-arg=-Wno-sign-compare
-            - -extra-arg=-Wno-sometimes-uninitialized
-            - -extra-arg=-Wno-tautological-overlap-compare
-            - -extra-arg=-Wno-unused-const-variable
-            - -extra-arg=-Wno-unused-lambda-capture
-            - -extra-arg=-Wno-unused-private-field
-            - -extra-arg=-Wno-unused-value
-            - -extra-arg=-Wno-unused-variable
-            - -extra-arg=-Wno-overloaded-virtual
-            - -extra-arg=-Wno-defaulted-function-deleted
-            - -extra-arg=-Wno-delete-non-abstract-non-virtual-dtor
-            - -extra-arg=-Wno-return-type-c-linkage
-# For CMake files
--   repo: https://github.com/cheshirekow/cmake-format-precommit
-    rev: v0.6.13
-    hooks:
-    -   id: cmake-format
--   repo: https://github.com/PFCCLab/cmake-lint-paddle
-    rev: v1.5.1
-    hooks:
-    -   id: cmakelint
-        args: [--config=./tools/codestyle/.cmakelintrc]
+  # Commit message linting
+  - repo: https://github.com/Goldziher/gitfluff
+    rev: v0.8.0
+    hooks:
+      - id: gitfluff-lint
+        args: ["--write"]
+        stages: [commit-msg]
+
+  # AI-Rulez: auto-generate AI assistant configuration files
+  - repo: https://github.com/Goldziher/ai-rulez
+    rev: v3.8.3
+    hooks:
+      - id: ai-rulez-generate
+
+  # General file checks
+  - repo: https://github.com/pre-commit/pre-commit-hooks
+    rev: v6.0.0
+    hooks:
+      - id: trailing-whitespace
+      - id: end-of-file-fixer
+      - id: check-merge-conflict
+      - id: check-added-large-files
+      - id: detect-private-key
+      - id: check-json
+      - id: check-yaml
+        args: ["--allow-multiple-documents", "--unsafe"]
+      - id: check-toml
+      - id: check-case-conflict
+
+  # TOML formatting
+  - repo: https://github.com/tox-dev/pyproject-fmt
+    rev: "v2.16.2"
+    hooks:
+      - id: pyproject-fmt
+
+  - repo: https://github.com/ComPWA/taplo-pre-commit
+    rev: v0.9.3
+    hooks:
+      - id: taplo-format
+        exclude: "pyproject.toml"
+
+  # Python: ruff (linting + formatting)
+  - repo: https://github.com/astral-sh/ruff-pre-commit
+    rev: v0.15.5
+    hooks:
+      - id: ruff
+        args: [--fix]
+      - id: ruff-format

.python-version

@@ -0,0 +1 @@
+3.13

CMakeLists.txt

@@ -1,6 +1,6 @@
 cmake_minimum_required(VERSION 3.16)
 project(paddle2onnx C CXX)
-# ONNX 1.16 requires C++ 17
+# ONNX 1.20 requires C++ 17
 set(CMAKE_CXX_STANDARD 17)
 set(CMAKE_MODULE_PATH ${CMAKE_MODULE_PATH} "${CMAKE_CURRENT_SOURCE_DIR}/cmake")
 # Build the libraries with - fPIC
@@ -24,7 +24,7 @@ endif()
 
 # Set max opset version for onnx if you build from other version of onnx this
 # should be modified.
-add_definitions(-DMAX_ONNX_OPSET_VERSION=23)
+add_definitions(-DMAX_ONNX_OPSET_VERSION=25)
 add_definitions(-DPADDLE2ONNX_LIB)
 
 # Internal flags for convert.h.in
@@ -57,10 +57,8 @@ include_directories(${CMAKE_CURRENT_BINARY_DIR})
 include_directories(${PROJECT_SOURCE_DIR}/third_party/optimizer)
 
 file(GLOB_RECURSE ALL_SRCS ${PROJECT_SOURCE_DIR}/paddle2onnx/*.cc
-     ${PROJECT_SOURCE_DIR}/third_party/optimizer/onnxoptimizer/*.cc)
+     ${PROJECT_SOURCE_DIR}/third_party/optimizer/onnx/optimizer/*.cc)
 list(REMOVE_ITEM ALL_SRCS ${PROJECT_SOURCE_DIR}/paddle2onnx/cpp2py_export.cc)
-list(REMOVE_ITEM ALL_SRCS
-     ${PROJECT_SOURCE_DIR}/third_party/optimizer/onnxoptimizer/cpp2py_export.cc)
 
 file(READ "${PROJECT_SOURCE_DIR}/VERSION_NUMBER" PADDLE2ONNX_VERSION)
 string(STRIP "${PADDLE2ONNX_VERSION}" PADDLE2ONNX_VERSION)

debug/p2o_infer_debugger.py

@@ -13,20 +13,22 @@
 # limitations under the License.
 
 import argparse
+import logging
 import os
+import queue
 import re
-import sys
-import logging
 import shutil
+import sys
+import tempfile
+import traceback
+from contextlib import contextmanager
+
 import numpy as np
 from onnxruntime import InferenceSession
+from prune_onnx_model import prune_onnx_model
+
 import paddle
 import paddle2onnx
-from prune_onnx_model import prune_onnx_model
-from contextlib import contextmanager
-import traceback
-import queue
-import tempfile
 
 current_dir = os.path.dirname(os.path.abspath(__file__))
 tests_dir = os.path.join(current_dir, "..", "tests")
@@ -202,8 +204,8 @@ def check_operator_with_print(
     def _redirect_paddle_output_to_file(
         paddle_model_file, log_file, inputs_data: tuple
     ):
-        import subprocess
         import pickle
+        import subprocess
         import tempfile
 
         temp_filename = None
@@ -266,7 +268,7 @@ def _compare_results(paddle_model_path, onnx_model_path, inputs_data: tuple):
         # TODO(wangmingkai02): adjust n according to the number of print op
         n = 8
         shape_list, dtype, data_list = [], None, []
-        with open(log_file, "r", encoding="utf-8") as f:
+        with open(log_file, encoding="utf-8") as f:
             lines = []
             for _ in range(n):
                 line = f.readline()
@@ -291,7 +293,7 @@ def _compare_results(paddle_model_path, onnx_model_path, inputs_data: tuple):
             providers=["CPUExecutionProvider"],
         )
         input_names = session.get_inputs()
-        input_feed = dict()
+        input_feed = {}
         for idx, input_name in enumerate(input_names):
             input_feed[input_name.name] = inputs_data[idx]
         result = session.run(output_names=None, input_feed=input_feed)
@@ -359,10 +361,7 @@ def _binary_search(program, block):
             idx = (left + right) // 2 + offset
             op = block.ops[idx]
             op_name = op.name()
-            if idx < left:
-                left = idx - offset + 1
-                offset = 0
-            elif op_name in SKIP_FORWARD_OP_LIST:
+            if idx < left or op_name in SKIP_FORWARD_OP_LIST:
                 left = idx - offset + 1
                 offset = 0
             elif op_name in SKIP_BACKWARD_OP_LIST:
@@ -533,7 +532,7 @@ def _compare_results(
             providers=["CPUExecutionProvider"],
         )
         input_names = session.get_inputs()
-        input_feed = dict()
+        input_feed = {}
         for idx, input_name in enumerate(input_names):
             input_feed[input_name.name] = inputs_data[idx]
         result = session.run(output_names=None, input_feed=input_feed)
@@ -606,10 +605,7 @@ def _check_operator(program, model_file, idx, input_shapes, input_dtypes):
             clone_program = program.clone()
             idx = (left + right) // 2 + offset
             op = clone_program.blocks[0].ops[idx]
-            if idx < left:
-                left = idx - offset + 1
-                offset = 0
-            elif op.name() in SKIP_FORWARD_OP_LIST:
+            if idx < left or op.name() in SKIP_FORWARD_OP_LIST:
                 left = idx - offset + 1
                 offset = 0
             elif op.name() in SKIP_BACKWARD_OP_LIST:
@@ -654,7 +650,7 @@ def locate_issue(
     model_file,
     input_shapes,
     input_dtypes,
-    candidates: list[int] = None,
+    candidates: list[int] | None = None,
     has_cf=False,
     binary_search=False,
     output_num=1,
@@ -705,7 +701,7 @@ def _get_mapping_and_uniq_set(program):
         index_mapping = {}
         ops = set()
         global_ops = set()
-        global_res = list()
+        global_res = []
         shadow_output_op_num = 0
         for block in program.blocks:
             ops |= _dfs(block, index_mapping)

debug/prune_onnx_model.py

@@ -12,21 +12,24 @@
 # See the License for the specific language governing permissions and
 # limitations under the License.
 
-import onnx
-from onnx import helper, TensorProto
-import os
-import math
 import logging
+import math
+import os
+
+import onnx
+from onnx import TensorProto, helper
 
 logger = logging.getLogger("p2o-logger")
 
 
 def prune_onnx_model(
     onnx_model_file,
     target_node_name="p2o.print",
-    target_dims=[1],
+    target_dims=None,
     target_dtype="float32",
 ):
+    if target_dims is None:
+        target_dims = [1]
     dtype_map = {
         "bool": (TensorProto.BOOL, bool),
         "float32": (TensorProto.FLOAT, float),