rename bypass_export_some_patches into torch_export_patches, keep the old name

Author: xadupre

README.rst

@@ -30,13 +30,13 @@ it helps exporting **pytorch models into ONNX**, mostly designed for LLMs using
 
 .. code-block:: python
 
-  with bypass_export_some_errors(patch_transformers=True) as f:
+  with torch_export_patches(patch_transformers=True) as f:
       ep = torch.export.export(model, args, kwargs=kwargs, dynamic_shapes=dynamic_shapes)
       # ...
 
 It also implements tools to investigate, validate exported models (ExportedProgramm, ONNXProgram, ...).
 See `documentation of onnx-diagnostic <https://sdpython.github.io/doc/onnx-diagnostic/dev/>`_ and
-`bypass_export_some_errors <https://sdpython.github.io/doc/onnx-diagnostic/dev/api/torch_export_patches/index.html#onnx_diagnostic.torch_export_patches.bypass_export_some_errors>`_.
+`torch_export_patches <https://sdpython.github.io/doc/onnx-diagnostic/dev/api/torch_export_patches/index.html#onnx_diagnostic.torch_export_patches.torch_export_patches>`_.
 
 Getting started
 +++++++++++++++

_doc/api/torch_export_patches/index.rst

@@ -13,6 +13,6 @@ onnx_diagnostic.torch_export_patches
     :members:
     :no-undoc-members:
 
-.. autofunction:: onnx_diagnostic.torch_export_patches.bypass_export_some_errors
+.. autofunction:: onnx_diagnostic.torch_export_patches.torch_export_patches
 
 .. autofunction:: onnx_diagnostic.torch_export_patches.register_additional_serialization_functions

_doc/examples/plot_export_hub_codellama.py

@@ -29,7 +29,7 @@
     get_pretrained_config,
     task_from_id,
 )
-from onnx_diagnostic.torch_export_patches import bypass_export_some_errors
+from onnx_diagnostic.torch_export_patches import torch_export_patches
 from onnx_diagnostic.torch_export_patches.patch_inputs import use_dyn_not_str
 
 model_id = "codellama/CodeLlama-7b-Python-hf"
@@ -90,9 +90,9 @@
 #
 # The model uses :class:`transformers.cache_utils.DynamicCache`.
 # It still requires patches to be exportable (control flow).
-# See :func:`onnx_diagnostic.torch_export_patches.bypass_export_some_errors`
+# See :func:`onnx_diagnostic.torch_export_patches.torch_export_patches`
 
-with bypass_export_some_errors(patch_transformers=True) as f:
+with torch_export_patches(patch_transformers=True) as f:
     ep = torch.export.export(
         model,
         (),

_doc/examples/plot_export_locate_issue.py

@@ -26,7 +26,7 @@
 import traceback
 import torch
 from onnx_diagnostic import doc
-from onnx_diagnostic.torch_export_patches import bypass_export_some_errors
+from onnx_diagnostic.torch_export_patches import torch_export_patches
 
 
 class ModelWithIssue(torch.nn.Module):
@@ -80,12 +80,12 @@ def forward(self, x: torch.Tensor, ys: list[torch.Tensor]):
 # Stop when a dynamic dimension turns static
 # ==========================================
 #
-# We use :func:`bypass_export_some_errors
-# <onnx_diagnostic.torch_export_patches.bypass_export_some_errors>`
+# We use :func:`torch_export_patches
+# <onnx_diagnostic.torch_export_patches.torch_export_patches>`
 # to replace torch implementation by a new one raising the exception
 # mentioned in previous section.
 
-with bypass_export_some_errors(stop_if_static=1, verbose=1):
+with torch_export_patches(stop_if_static=1, verbose=1):
     try:
         torch.export.export(model, inputs, dynamic_shapes=dyn_shapes)
     except (AssertionError, torch._dynamo.exc.TorchRuntimeError) as e:

_doc/examples/plot_export_tiny_llm_patched.py

@@ -69,7 +69,7 @@
 from onnx_diagnostic import doc
 from onnx_diagnostic.helpers.cache_helper import is_cache_dynamic_registered
 from onnx_diagnostic.helpers import string_type
-from onnx_diagnostic.torch_export_patches import bypass_export_some_errors
+from onnx_diagnostic.torch_export_patches import torch_export_patches
 from onnx_diagnostic.torch_models.llms import get_tiny_llm
 
 
@@ -101,10 +101,10 @@
 
 # %%
 # If they are not registered, function
-# func:`onnx_diagnostic.torch_export_patches.bypass_export_some_errors`
+# func:`onnx_diagnostic.torch_export_patches.torch_export_patches`
 # should take care of it. Then we export.
 
-with bypass_export_some_errors(patch_transformers=True, verbose=10) as modificator:
+with torch_export_patches(patch_transformers=True, verbose=10) as modificator:
     assert is_cache_dynamic_registered()  # it must be true here
     ep = torch.export.export(
         untrained_model,
@@ -126,7 +126,7 @@
 
 cloned_inputs = copy.deepcopy(inputs)
 
-with bypass_export_some_errors(patch_transformers=True, verbose=10) as modificator:
+with torch_export_patches(patch_transformers=True, verbose=10) as modificator:
     ep = torch.export.export(
         model,
         (),

_doc/examples/plot_export_tiny_phi2.py

@@ -26,7 +26,7 @@
 from onnx_diagnostic.helpers import max_diff, string_diff, string_type
 from onnx_diagnostic.helpers.cache_helper import is_cache_dynamic_registered
 from onnx_diagnostic.helpers.rt_helper import make_feeds
-from onnx_diagnostic.torch_export_patches import bypass_export_some_errors
+from onnx_diagnostic.torch_export_patches import torch_export_patches
 from onnx_diagnostic.torch_export_patches.patch_inputs import use_dyn_not_str
 from onnx_diagnostic.torch_models.hghub import (
     get_untrained_model_with_inputs,
@@ -76,7 +76,7 @@
 # ++++++
 
 
-with bypass_export_some_errors(patch_transformers=True) as modificator:
+with torch_export_patches(patch_transformers=True) as modificator:
 
     # Unnecessary steps but useful in case of an error
     # We check the cache is registered.
@@ -110,7 +110,7 @@
 # applies :meth:`torch.export.ExportedProgram.run_decompositions`
 # may export local pieces of the model again.
 
-with bypass_export_some_errors(patch_transformers=True):
+with torch_export_patches(patch_transformers=True):
     epo = torch.onnx.export(
         ep, (), kwargs=copy.deepcopy(inputs), dynamic_shapes=dynamic_shapes, dynamo=True
     )

_doc/examples/plot_export_with_dynamic_cache.py

@@ -30,7 +30,7 @@
     make_dynamic_cache,
 )
 from onnx_diagnostic.export import ModelInputs
-from onnx_diagnostic.torch_export_patches import bypass_export_some_errors
+from onnx_diagnostic.torch_export_patches import torch_export_patches
 
 
 class Model(torch.nn.Module):
@@ -99,14 +99,14 @@ def forward(self, cache, z):
 # And finally the export.
 # The export is simple if ``transformers>=4.50``, otherwise,
 # transformers needs to be patched.
-# :func:`onnx_diagnostic.torch_export_patches.bypass_export_some_errors`
+# :func:`onnx_diagnostic.torch_export_patches.torch_export_patches`
 # registers functions to serialize ``DynamicCache``. This one is modified to make
 # the shape inference implemented in :epkg:`torch` happy.
 
 if has_transformers("4.50"):
     ep = torch.export.export(model, inputs[0], dynamic_shapes=ds[0], strict=False)
 else:
-    with bypass_export_some_errors(patch_transformers=True) as modificator:
+    with torch_export_patches(patch_transformers=True) as modificator:
         ep = torch.export.export(
             model, modificator(inputs[0]), dynamic_shapes=ds[0], strict=False
         )

_doc/index.rst

@@ -24,12 +24,12 @@ Sources available at `github/onnx-diagnostic <https://github.com/sdpython/onnx-d
 
 .. code-block:: python
 
-  with bypass_export_some_errors(patch_transformers=True) as f:
+  with torch_export_patches(patch_transformers=True) as f:
       ep = torch.export.export(model, args, kwargs=kwargs, dynamic_shapes=dynamic_shapes)
       # ...
 
 It also implements tools to investigate, validate exported models (ExportedProgramm, ONNXProgram, ...).
-:func:`onnx_diagnostic.torch_export_patches.bypass_export_some_errors`.
+:func:`onnx_diagnostic.torch_export_patches.torch_export_patches`.
 
 .. toctree::
     :maxdepth: 1

_doc/recipes/plot_dynamic_shapes_python_int.py

@@ -14,7 +14,7 @@
 import math
 import torch
 from onnx_diagnostic import doc
-from onnx_diagnostic.torch_export_patches import bypass_export_some_errors
+from onnx_diagnostic.torch_export_patches import torch_export_patches
 
 
 class Model(torch.nn.Module):
@@ -73,12 +73,12 @@ def forward(self, x):
 # Find the error
 # ++++++++++++++
 #
-# Function :func:`onnx_diagnostic.torch_export_patches.bypass_export_some_errors`
+# Function :func:`onnx_diagnostic.torch_export_patches.torch_export_patches`
 # has a parameter ``stop_if_static`` which patches torch to raise exception
 # when something like that is happening.
 
 
-with bypass_export_some_errors(stop_if_static=True):
+with torch_export_patches(stop_if_static=True):
     ep = torch.export.export(model, (x,), dynamic_shapes=({0: DYN, 1: DYN},))
     print(ep)
 

_unittests/ut_export/test_dynamic_shapes.py

@@ -5,7 +5,7 @@
 from onnx_diagnostic.helpers import string_type
 from onnx_diagnostic.helpers.cache_helper import make_dynamic_cache
 from onnx_diagnostic.export import ModelInputs, CoupleInputsDynamicShapes
-from onnx_diagnostic.torch_export_patches import bypass_export_some_errors
+from onnx_diagnostic.torch_export_patches import torch_export_patches
 
 
 class TestDynamicShapes(ExtTestCase):
@@ -571,7 +571,7 @@ def test_couple_input_ds_cache(self):
 
         kwargs = {"A": T3x4, "B": (T3x1, cache)}
         Cls = CoupleInputsDynamicShapes
-        with bypass_export_some_errors(patch_transformers=True):
+        with torch_export_patches(patch_transformers=True):
             self.assertEqual(
                 None,
                 Cls(
@@ -749,7 +749,7 @@ def test_couple_input_ds_change_dynamic_dimensions_dynamic_cache(self):
             {"A": make_dynamic_cache([(torch.ones((2, 2, 2, 2)), torch.ones((2, 2, 2, 2)))])},
             {"A": [[{0: "batch", 2: "last"}], [{0: "batch", 2: "last"}]]},
         )
-        with bypass_export_some_errors(patch_transformers=True):
+        with torch_export_patches(patch_transformers=True):
             new_inputs = inst.change_dynamic_dimensions()
         self.assertIsInstance(new_inputs["A"], transformers.cache_utils.DynamicCache)
         self.assertEqual((3, 2, 3, 2), new_inputs["A"].key_cache[0].shape)