[Executorch][target recipes] Add target based recipes for lowering models to a target device #13791

abhinaykukkadapu · 2025-08-29T05:53:45Z

Stack from ghstack (oldest at bottom):

-> [Executorch][target recipes] Add target based recipes for lowering models to a target device #13791

This diff introduces multi backend/ target based recipes to lower a model with very less code. Target recipes provide pre-configured backend recipes to use them and retarget if needed.

See RFC: #13732

Usage

from executorch.export import export
from executorch.export.target_recipes import get_ios_recipe

# CoreML + XNNPACK (FP32)
recipe = get_ios_recipe() # default = "ios-arm64-coreml-fp16"
session = export(model, recipe, example_inputs)
session.save_pte_file("model.pte")

Advanced usage to combine specific recipes of one or two or backends

one can directly use ExportRecipe.combine_recipes() to combine specific backend recipes.

recipe1 = ExportRecipe.get_recipe(CoreMLRecipeType.FP32)
recipe2 = ExportRecipe.get_recipe(XNNPackRecipeType.FP32)
combined_recipe = ExportRecipe.combine(
                     [recipe1, recipe2],
                     recipe_name="multi_backend_coreml_xnnpack_fp32"
                  )
session = export(model, combined_recipe, example_inputs)

Additional changes:

Relaxed kwarg validation in the backend providers to just ignore them instead of erroring out.

Fixes: #13732

Differential Revision: D81297451

…dels to a target device This diff introduces multi backend/ target based recipes to lower a model with very less code. Target recipes provide pre-configured backend recipes to use them and retarget if needed. See RFC: #13732 ## Usage ``` from executorch.export import export, ExportRecipe, IOSTargetRecipeType # CoreML + XNNPACK coreml_xnnpack_recipe = ExportRecipe.get_recipe(IOSTargetRecipeType.IOS_ARM64_COREML_FP32) session = export(model, coreml_xnnpack_recipe, example_inputs) session.save_pte_file("model.pte") ``` ## Advanced usage one can directly use `ExportRecipe.combine_recipes()` to combine specific backend recipes. ``` recipe1 = ExportRecipe.get_recipe(AndroidRecipeType.XYZ) recipe2 = ExportRecipe.get_recipe(XNNPackRecipeType.FP32) combined_recipe = ExportRecipe.combine( [recipe1, recipe2], recipe_name="multi_backend_coreml_xnnpack_fp32" ) session = export(model, combined_recipe, example_inputs) ``` Fixes: #13732 Differential Revision: [D81297451](https://our.internmc.facebook.com/intern/diff/D81297451/) [ghstack-poisoned]

pytorch-bot · 2025-08-29T05:53:48Z

🔗 Helpful Links

🧪 See artifacts and rendered test results at hud.pytorch.org/pr/pytorch/executorch/13791

📄 Preview Python docs built from this PR

Note: Links to docs will display an error until the docs builds have been completed.

❌ 2 New Failures

As of commit ef5f1ad with merge base 19d22ab ():

NEW FAILURES - The following jobs have failed:

Build documentation / build (buck2) / Build doc (gh)
At least one of the pre-conditions you specified did not hold
pull / unittest-nxp-neutron / linux-job (gh)
RuntimeError: Command docker exec -t 175bcdc60b4acef3d0db0686c8d457be8276d6af22e12e49957eeba4d7467ae5 /exec failed with exit code 1

This comment was automatically generated by Dr. CI and updates every 15 minutes.

…dels to a target device This diff introduces multi backend/ target based recipes to lower a model with very less code. Target recipes provide pre-configured backend recipes to use them and retarget if needed. See RFC: #13732 ## Usage ``` from executorch.export import export, ExportRecipe, IOSTargetRecipeType # CoreML + XNNPACK coreml_xnnpack_recipe = ExportRecipe.get_recipe(IOSTargetRecipeType.IOS_ARM64_COREML_FP32) session = export(model, coreml_xnnpack_recipe, example_inputs) session.save_pte_file("model.pte") ``` ## Advanced usage one can directly use `ExportRecipe.combine_recipes()` to combine specific backend recipes. ``` recipe1 = ExportRecipe.get_recipe(AndroidRecipeType.XYZ) recipe2 = ExportRecipe.get_recipe(XNNPackRecipeType.FP32) combined_recipe = ExportRecipe.combine( [recipe1, recipe2], recipe_name="multi_backend_coreml_xnnpack_fp32" ) session = export(model, combined_recipe, example_inputs) ``` Fixes: #13732 Differential Revision: [D81297451](https://our.internmc.facebook.com/intern/diff/D81297451/) ghstack-source-id: 306434236 Pull Request resolved: #13791

facebook-github-bot · 2025-08-29T05:53:56Z

This pull request was exported from Phabricator. Differential Revision: D81297451

github-actions · 2025-08-29T05:54:25Z

This PR needs a `release notes:` label

If your change should be included in the release notes (i.e. would users of this library care about this change?), please use a label starting with release notes:. This helps us keep track and include your important work in the next release notes.

To add a label, you can comment to pytorchbot, for example
@pytorchbot label "release notes: none"

For more information, see
https://github.com/pytorch/pytorch/wiki/PyTorch-AutoLabel-Bot#why-categorize-for-release-notes-and-how-does-it-work.

…lowering models to a target device" This diff introduces multi backend/ target based recipes to lower a model with very less code. Target recipes provide pre-configured backend recipes to use them and retarget if needed. See RFC: #13732 ## Usage ``` from executorch.export import export, ExportRecipe, IOSTargetRecipeType # CoreML + XNNPACK coreml_xnnpack_recipe = ExportRecipe.get_recipe(IOSTargetRecipeType.IOS_ARM64_COREML_FP32) session = export(model, coreml_xnnpack_recipe, example_inputs) session.save_pte_file("model.pte") ``` ## Advanced usage one can directly use `ExportRecipe.combine_recipes()` to combine specific backend recipes. ``` recipe1 = ExportRecipe.get_recipe(AndroidRecipeType.XYZ) recipe2 = ExportRecipe.get_recipe(XNNPackRecipeType.FP32) combined_recipe = ExportRecipe.combine( [recipe1, recipe2], recipe_name="multi_backend_coreml_xnnpack_fp32" ) session = export(model, combined_recipe, example_inputs) ``` Fixes: #13732 Differential Revision: [D81297451](https://our.internmc.facebook.com/intern/diff/D81297451/) [ghstack-poisoned]

…dels to a target device Pull Request resolved: #13791 This diff introduces multi backend/ target based recipes to lower a model with very less code. Target recipes provide pre-configured backend recipes to use them and retarget if needed. See RFC: #13732 ## Usage ``` from executorch.export import export, ExportRecipe, IOSTargetRecipeType # CoreML + XNNPACK coreml_xnnpack_recipe = ExportRecipe.get_recipe(IOSTargetRecipeType.IOS_ARM64_COREML_FP32) session = export(model, coreml_xnnpack_recipe, example_inputs) session.save_pte_file("model.pte") ``` ## Advanced usage one can directly use `ExportRecipe.combine_recipes()` to combine specific backend recipes. ``` recipe1 = ExportRecipe.get_recipe(AndroidRecipeType.XYZ) recipe2 = ExportRecipe.get_recipe(XNNPackRecipeType.FP32) combined_recipe = ExportRecipe.combine( [recipe1, recipe2], recipe_name="multi_backend_coreml_xnnpack_fp32" ) session = export(model, combined_recipe, example_inputs) ``` Fixes: #13732 ghstack-source-id: 306438497 @exported-using-ghexport Differential Revision: [D81297451](https://our.internmc.facebook.com/intern/diff/D81297451/)

facebook-github-bot · 2025-08-29T06:13:56Z

This pull request was exported from Phabricator. Differential Revision: D81297451

kimishpatel · 2025-08-29T18:14:54Z

I left a comment on the original issue. Not fully convinced why we need multibackend recipe as a separate concept

…lowering models to a target device" This diff introduces multi backend/ target based recipes to lower a model with very less code. Target recipes provide pre-configured backend recipes to use them and retarget if needed. See RFC: #13732 ## Usage ``` from executorch.export import export, ExportRecipe, IOSTargetRecipeType # CoreML + XNNPACK coreml_xnnpack_recipe = ExportRecipe.get_recipe(IOSTargetRecipeType.IOS_ARM64_COREML_FP32) session = export(model, coreml_xnnpack_recipe, example_inputs) session.save_pte_file("model.pte") ``` ## Advanced usage one can directly use `ExportRecipe.combine_recipes()` to combine specific backend recipes. ``` recipe1 = ExportRecipe.get_recipe(AndroidRecipeType.XYZ) recipe2 = ExportRecipe.get_recipe(XNNPackRecipeType.FP32) combined_recipe = ExportRecipe.combine( [recipe1, recipe2], recipe_name="multi_backend_coreml_xnnpack_fp32" ) session = export(model, combined_recipe, example_inputs) ``` Fixes: #13732 Differential Revision: [D81297451](https://our.internmc.facebook.com/intern/diff/D81297451/) [ghstack-poisoned]

…dels to a target device Pull Request resolved: #13791 This diff introduces multi backend/ target based recipes to lower a model with very less code. Target recipes provide pre-configured backend recipes to use them and retarget if needed. See RFC: #13732 ## Usage ``` from executorch.export import export from executorch.export.target_recipes import get_ios_recipe # CoreML + XNNPACK (FP32) recipe = get_ios_recipe() # default = "ios-arm64-coreml-fp32" session = export(model, recipe, example_inputs) session.save_pte_file("model.pte") ``` ## Advanced usage to combine specific recipes of two or one backend one can directly use `ExportRecipe.combine_recipes()` to combine specific backend recipes. ``` recipe1 = ExportRecipe.get_recipe(CoreMLRecipeType.FP32) recipe2 = ExportRecipe.get_recipe(XNNPackRecipeType.FP32) combined_recipe = ExportRecipe.combine( [recipe1, recipe2], recipe_name="multi_backend_coreml_xnnpack_fp32" ) session = export(model, combined_recipe, example_inputs) ``` Fixes: #13732 ghstack-source-id: 307058764 @exported-using-ghexport Differential Revision: [D81297451](https://our.internmc.facebook.com/intern/diff/D81297451/)

facebook-github-bot · 2025-09-02T20:38:14Z

This pull request was exported from Phabricator. Differential Revision: D81297451

…lowering models to a target device" This diff introduces multi backend/ target based recipes to lower a model with very less code. Target recipes provide pre-configured backend recipes to use them and retarget if needed. See RFC: #13732 ## Usage ``` from executorch.export import export, ExportRecipe, IOSTargetRecipeType # CoreML + XNNPACK coreml_xnnpack_recipe = ExportRecipe.get_recipe(IOSTargetRecipeType.IOS_ARM64_COREML_FP32) session = export(model, coreml_xnnpack_recipe, example_inputs) session.save_pte_file("model.pte") ``` ## Advanced usage one can directly use `ExportRecipe.combine_recipes()` to combine specific backend recipes. ``` recipe1 = ExportRecipe.get_recipe(AndroidRecipeType.XYZ) recipe2 = ExportRecipe.get_recipe(XNNPackRecipeType.FP32) combined_recipe = ExportRecipe.combine( [recipe1, recipe2], recipe_name="multi_backend_coreml_xnnpack_fp32" ) session = export(model, combined_recipe, example_inputs) ``` Fixes: #13732 Differential Revision: [D81297451](https://our.internmc.facebook.com/intern/diff/D81297451/) [ghstack-poisoned]

…dels to a target device Pull Request resolved: #13791 This diff introduces multi backend/ target based recipes to lower a model with very less code. Target recipes provide pre-configured backend recipes to use them and retarget if needed. See RFC: #13732 ## Usage ``` from executorch.export import export from executorch.export.target_recipes import get_ios_recipe # CoreML + XNNPACK (FP32) recipe = get_ios_recipe() # default = "ios-arm64-coreml-fp32" session = export(model, recipe, example_inputs) session.save_pte_file("model.pte") ``` ## Advanced usage to combine specific recipes of one or two or backends one can directly use `ExportRecipe.combine_recipes()` to combine specific backend recipes. ``` recipe1 = ExportRecipe.get_recipe(CoreMLRecipeType.FP32) recipe2 = ExportRecipe.get_recipe(XNNPackRecipeType.FP32) combined_recipe = ExportRecipe.combine( [recipe1, recipe2], recipe_name="multi_backend_coreml_xnnpack_fp32" ) session = export(model, combined_recipe, example_inputs) ``` Additional changes: 1. Relaxed kwarg validation in the backend providers to just ignore them instead of erroring out. Fixes: #13732 ghstack-source-id: 307121550 @exported-using-ghexport Differential Revision: [D81297451](https://our.internmc.facebook.com/intern/diff/D81297451/)

facebook-github-bot · 2025-09-03T00:00:20Z

This pull request was exported from Phabricator. Differential Revision: D81297451

abhinaykukkadapu · 2025-09-03T00:00:56Z

I left a comment on the original issue. Not fully convinced why we need multibackend recipe as a separate concept

@kimishpatel Removed the multi backend provider and went with simple funciton calls.

…lowering models to a target device" This diff introduces multi backend/ target based recipes to lower a model with very less code. Target recipes provide pre-configured backend recipes to use them and retarget if needed. See RFC: #13732 ## Usage ``` from executorch.export import export, ExportRecipe, IOSTargetRecipeType # CoreML + XNNPACK coreml_xnnpack_recipe = ExportRecipe.get_recipe(IOSTargetRecipeType.IOS_ARM64_COREML_FP32) session = export(model, coreml_xnnpack_recipe, example_inputs) session.save_pte_file("model.pte") ``` ## Advanced usage one can directly use `ExportRecipe.combine_recipes()` to combine specific backend recipes. ``` recipe1 = ExportRecipe.get_recipe(AndroidRecipeType.XYZ) recipe2 = ExportRecipe.get_recipe(XNNPackRecipeType.FP32) combined_recipe = ExportRecipe.combine( [recipe1, recipe2], recipe_name="multi_backend_coreml_xnnpack_fp32" ) session = export(model, combined_recipe, example_inputs) ``` Fixes: #13732 Differential Revision: [D81297451](https://our.internmc.facebook.com/intern/diff/D81297451/) [ghstack-poisoned]

…dels to a target device Pull Request resolved: #13791 This diff introduces multi backend/ target based recipes to lower a model with very less code. Target recipes provide pre-configured backend recipes to use them and retarget if needed. See RFC: #13732 ## Usage ``` from executorch.export import export from executorch.export.target_recipes import get_ios_recipe # CoreML + XNNPACK (FP32) recipe = get_ios_recipe() # default = "ios-arm64-coreml-fp32" session = export(model, recipe, example_inputs) session.save_pte_file("model.pte") ``` ## Advanced usage to combine specific recipes of one or two or backends one can directly use `ExportRecipe.combine_recipes()` to combine specific backend recipes. ``` recipe1 = ExportRecipe.get_recipe(CoreMLRecipeType.FP32) recipe2 = ExportRecipe.get_recipe(XNNPackRecipeType.FP32) combined_recipe = ExportRecipe.combine( [recipe1, recipe2], recipe_name="multi_backend_coreml_xnnpack_fp32" ) session = export(model, combined_recipe, example_inputs) ``` Additional changes: 1. Relaxed kwarg validation in the backend providers to just ignore them instead of erroring out. Fixes: #13732 ghstack-source-id: 307139141 @exported-using-ghexport Differential Revision: [D81297451](https://our.internmc.facebook.com/intern/diff/D81297451/)

facebook-github-bot · 2025-09-03T01:23:24Z

This pull request was exported from Phabricator. Differential Revision: D81297451

kimishpatel · 2025-09-03T01:24:01Z

backends/apple/coreml/recipes/coreml_recipe_provider.py

        if unexpected:
-            raise ValueError(
-                f"Recipe '{recipe_type.value}' received unexpected parameters: {list(unexpected)}"
+            logging.warning(


Why this turned into warning now? to combine multiple recipes? and fallback to the other backend when it fails?

Correct, to combine multiple recipes, for example, get_coreml_recipe(**kwargs) this will have to forward them to both coreml and xnnpack. Hard failure like above would restrict the creation of multi backend recipes.

export/recipe.py

kimishpatel

looks ok to me. some nits plus a comment on quantizaiton.

you probably want to add recipes plus test for android

abhinaykukkadapu · 2025-09-03T05:03:43Z

looks ok to me. some nits plus a comment on quantizaiton.

you probably want to add recipes plus test for android

Thanks for the review, will add android specific things after adding QNN recipes, which i'm working on right now.

…lowering models to a target device" This diff introduces multi backend/ target based recipes to lower a model with very less code. Target recipes provide pre-configured backend recipes to use them and retarget if needed. See RFC: #13732 ## Usage ``` from executorch.export import export, ExportRecipe, IOSTargetRecipeType # CoreML + XNNPACK coreml_xnnpack_recipe = ExportRecipe.get_recipe(IOSTargetRecipeType.IOS_ARM64_COREML_FP32) session = export(model, coreml_xnnpack_recipe, example_inputs) session.save_pte_file("model.pte") ``` ## Advanced usage one can directly use `ExportRecipe.combine_recipes()` to combine specific backend recipes. ``` recipe1 = ExportRecipe.get_recipe(AndroidRecipeType.XYZ) recipe2 = ExportRecipe.get_recipe(XNNPackRecipeType.FP32) combined_recipe = ExportRecipe.combine( [recipe1, recipe2], recipe_name="multi_backend_coreml_xnnpack_fp32" ) session = export(model, combined_recipe, example_inputs) ``` Fixes: #13732 Differential Revision: [D81297451](https://our.internmc.facebook.com/intern/diff/D81297451/) [ghstack-poisoned]

…dels to a target device Pull Request resolved: #13791 This diff introduces multi backend/ target based recipes to lower a model with very less code. Target recipes provide pre-configured backend recipes to use them and retarget if needed. See RFC: #13732 ## Usage ``` from executorch.export import export from executorch.export.target_recipes import get_ios_recipe # CoreML + XNNPACK (FP32) recipe = get_ios_recipe() # default = "ios-arm64-coreml-fp32" session = export(model, recipe, example_inputs) session.save_pte_file("model.pte") ``` ## Advanced usage to combine specific recipes of one or two or backends one can directly use `ExportRecipe.combine_recipes()` to combine specific backend recipes. ``` recipe1 = ExportRecipe.get_recipe(CoreMLRecipeType.FP32) recipe2 = ExportRecipe.get_recipe(XNNPackRecipeType.FP32) combined_recipe = ExportRecipe.combine( [recipe1, recipe2], recipe_name="multi_backend_coreml_xnnpack_fp32" ) session = export(model, combined_recipe, example_inputs) ``` Additional changes: 1. Relaxed kwarg validation in the backend providers to just ignore them instead of erroring out. Fixes: #13732 ghstack-source-id: 307321089 @exported-using-ghexport Differential Revision: [D81297451](https://our.internmc.facebook.com/intern/diff/D81297451/)

facebook-github-bot · 2025-09-03T18:58:53Z

This pull request was exported from Phabricator. Differential Revision: D81297451

…lowering models to a target device" This diff introduces multi backend/ target based recipes to lower a model with very less code. Target recipes provide pre-configured backend recipes to use them and retarget if needed. See RFC: #13732 ## Usage ``` from executorch.export import export from executorch.export.target_recipes import get_ios_recipe # CoreML + XNNPACK (FP32) recipe = get_ios_recipe() # default = "ios-arm64-coreml-fp16" session = export(model, recipe, example_inputs) session.save_pte_file("model.pte") ``` ## Advanced usage to combine specific recipes of one or two or backends one can directly use `ExportRecipe.combine_recipes()` to combine specific backend recipes. ``` recipe1 = ExportRecipe.get_recipe(CoreMLRecipeType.FP32) recipe2 = ExportRecipe.get_recipe(XNNPackRecipeType.FP32) combined_recipe = ExportRecipe.combine( [recipe1, recipe2], recipe_name="multi_backend_coreml_xnnpack_fp32" ) session = export(model, combined_recipe, example_inputs) ``` Additional changes: 1. Relaxed kwarg validation in the backend providers to just ignore them instead of erroring out. Fixes: #13732 Differential Revision: [D81297451](https://our.internmc.facebook.com/intern/diff/D81297451/) [ghstack-poisoned]

…dels to a target device Pull Request resolved: #13791 This diff introduces multi backend/ target based recipes to lower a model with very less code. Target recipes provide pre-configured backend recipes to use them and retarget if needed. See RFC: #13732 ## Usage ``` from executorch.export import export from executorch.export.target_recipes import get_ios_recipe # CoreML + XNNPACK (FP32) recipe = get_ios_recipe() # default = "ios-arm64-coreml-fp16" session = export(model, recipe, example_inputs) session.save_pte_file("model.pte") ``` ## Advanced usage to combine specific recipes of one or two or backends one can directly use `ExportRecipe.combine_recipes()` to combine specific backend recipes. ``` recipe1 = ExportRecipe.get_recipe(CoreMLRecipeType.FP32) recipe2 = ExportRecipe.get_recipe(XNNPackRecipeType.FP32) combined_recipe = ExportRecipe.combine( [recipe1, recipe2], recipe_name="multi_backend_coreml_xnnpack_fp32" ) session = export(model, combined_recipe, example_inputs) ``` Additional changes: 1. Relaxed kwarg validation in the backend providers to just ignore them instead of erroring out. Fixes: #13732 ghstack-source-id: 307327146 @exported-using-ghexport Differential Revision: [D81297451](https://our.internmc.facebook.com/intern/diff/D81297451/)

facebook-github-bot · 2025-09-03T19:20:57Z

This pull request was exported from Phabricator. Differential Revision: D81297451

…lowering models to a target device" This diff introduces multi backend/ target based recipes to lower a model with very less code. Target recipes provide pre-configured backend recipes to use them and retarget if needed. See RFC: #13732 ## Usage ``` from executorch.export import export from executorch.export.target_recipes import get_ios_recipe # CoreML + XNNPACK (FP32) recipe = get_ios_recipe() # default = "ios-arm64-coreml-fp16" session = export(model, recipe, example_inputs) session.save_pte_file("model.pte") ``` ## Advanced usage to combine specific recipes of one or two or backends one can directly use `ExportRecipe.combine_recipes()` to combine specific backend recipes. ``` recipe1 = ExportRecipe.get_recipe(CoreMLRecipeType.FP32) recipe2 = ExportRecipe.get_recipe(XNNPackRecipeType.FP32) combined_recipe = ExportRecipe.combine( [recipe1, recipe2], recipe_name="multi_backend_coreml_xnnpack_fp32" ) session = export(model, combined_recipe, example_inputs) ``` Additional changes: 1. Relaxed kwarg validation in the backend providers to just ignore them instead of erroring out. Fixes: #13732 Differential Revision: [D81297451](https://our.internmc.facebook.com/intern/diff/D81297451/) [ghstack-poisoned]

…dels to a target device Pull Request resolved: #13791 This diff introduces multi backend/ target based recipes to lower a model with very less code. Target recipes provide pre-configured backend recipes to use them and retarget if needed. See RFC: #13732 ## Usage ``` from executorch.export import export from executorch.export.target_recipes import get_ios_recipe # CoreML + XNNPACK (FP32) recipe = get_ios_recipe() # default = "ios-arm64-coreml-fp16" session = export(model, recipe, example_inputs) session.save_pte_file("model.pte") ``` ## Advanced usage to combine specific recipes of one or two or backends one can directly use `ExportRecipe.combine_recipes()` to combine specific backend recipes. ``` recipe1 = ExportRecipe.get_recipe(CoreMLRecipeType.FP32) recipe2 = ExportRecipe.get_recipe(XNNPackRecipeType.FP32) combined_recipe = ExportRecipe.combine( [recipe1, recipe2], recipe_name="multi_backend_coreml_xnnpack_fp32" ) session = export(model, combined_recipe, example_inputs) ``` Additional changes: 1. Relaxed kwarg validation in the backend providers to just ignore them instead of erroring out. Fixes: #13732 ghstack-source-id: 307388517 @exported-using-ghexport Differential Revision: [D81297451](https://our.internmc.facebook.com/intern/diff/D81297451/)

facebook-github-bot · 2025-09-03T22:51:22Z

This pull request was exported from Phabricator. Differential Revision: D81297451

…lowering models to a target device" This diff introduces multi backend/ target based recipes to lower a model with very less code. Target recipes provide pre-configured backend recipes to use them and retarget if needed. See RFC: #13732 ## Usage ``` from executorch.export import export from executorch.export.target_recipes import get_ios_recipe # CoreML + XNNPACK (FP32) recipe = get_ios_recipe() # default = "ios-arm64-coreml-fp16" session = export(model, recipe, example_inputs) session.save_pte_file("model.pte") ``` ## Advanced usage to combine specific recipes of one or two or backends one can directly use `ExportRecipe.combine_recipes()` to combine specific backend recipes. ``` recipe1 = ExportRecipe.get_recipe(CoreMLRecipeType.FP32) recipe2 = ExportRecipe.get_recipe(XNNPackRecipeType.FP32) combined_recipe = ExportRecipe.combine( [recipe1, recipe2], recipe_name="multi_backend_coreml_xnnpack_fp32" ) session = export(model, combined_recipe, example_inputs) ``` Additional changes: 1. Relaxed kwarg validation in the backend providers to just ignore them instead of erroring out. Fixes: #13732 Differential Revision: [D81297451](https://our.internmc.facebook.com/intern/diff/D81297451/) [ghstack-poisoned]

…dels to a target device Pull Request resolved: #13791 This diff introduces multi backend/ target based recipes to lower a model with very less code. Target recipes provide pre-configured backend recipes to use them and retarget if needed. See RFC: #13732 ## Usage ``` from executorch.export import export from executorch.export.target_recipes import get_ios_recipe # CoreML + XNNPACK (FP32) recipe = get_ios_recipe() # default = "ios-arm64-coreml-fp16" session = export(model, recipe, example_inputs) session.save_pte_file("model.pte") ``` ## Advanced usage to combine specific recipes of one or two or backends one can directly use `ExportRecipe.combine_recipes()` to combine specific backend recipes. ``` recipe1 = ExportRecipe.get_recipe(CoreMLRecipeType.FP32) recipe2 = ExportRecipe.get_recipe(XNNPackRecipeType.FP32) combined_recipe = ExportRecipe.combine( [recipe1, recipe2], recipe_name="multi_backend_coreml_xnnpack_fp32" ) session = export(model, combined_recipe, example_inputs) ``` Additional changes: 1. Relaxed kwarg validation in the backend providers to just ignore them instead of erroring out. Fixes: #13732 ghstack-source-id: 307424597 @exported-using-ghexport Differential Revision: [D81297451](https://our.internmc.facebook.com/intern/diff/D81297451/)

facebook-github-bot · 2025-09-04T02:02:10Z

This pull request was exported from Phabricator. Differential Revision: D81297451

…lowering models to a target device" This diff introduces multi backend/ target based recipes to lower a model with very less code. Target recipes provide pre-configured backend recipes to use them and retarget if needed. See RFC: #13732 ## Usage ``` from executorch.export import export from executorch.export.target_recipes import get_ios_recipe # CoreML + XNNPACK (FP32) recipe = get_ios_recipe() # default = "ios-arm64-coreml-fp16" session = export(model, recipe, example_inputs) session.save_pte_file("model.pte") ``` ## Advanced usage to combine specific recipes of one or two or backends one can directly use `ExportRecipe.combine_recipes()` to combine specific backend recipes. ``` recipe1 = ExportRecipe.get_recipe(CoreMLRecipeType.FP32) recipe2 = ExportRecipe.get_recipe(XNNPackRecipeType.FP32) combined_recipe = ExportRecipe.combine( [recipe1, recipe2], recipe_name="multi_backend_coreml_xnnpack_fp32" ) session = export(model, combined_recipe, example_inputs) ``` Additional changes: 1. Relaxed kwarg validation in the backend providers to just ignore them instead of erroring out. Fixes: #13732 Differential Revision: [D81297451](https://our.internmc.facebook.com/intern/diff/D81297451/) [ghstack-poisoned]

…dels to a target device Pull Request resolved: #13791 This diff introduces multi backend/ target based recipes to lower a model with very less code. Target recipes provide pre-configured backend recipes to use them and retarget if needed. See RFC: #13732 ## Usage ``` from executorch.export import export from executorch.export.target_recipes import get_ios_recipe # CoreML + XNNPACK (FP32) recipe = get_ios_recipe() # default = "ios-arm64-coreml-fp16" session = export(model, recipe, example_inputs) session.save_pte_file("model.pte") ``` ## Advanced usage to combine specific recipes of one or two or backends one can directly use `ExportRecipe.combine_recipes()` to combine specific backend recipes. ``` recipe1 = ExportRecipe.get_recipe(CoreMLRecipeType.FP32) recipe2 = ExportRecipe.get_recipe(XNNPackRecipeType.FP32) combined_recipe = ExportRecipe.combine( [recipe1, recipe2], recipe_name="multi_backend_coreml_xnnpack_fp32" ) session = export(model, combined_recipe, example_inputs) ``` Additional changes: 1. Relaxed kwarg validation in the backend providers to just ignore them instead of erroring out. Fixes: #13732 ghstack-source-id: 307671542 @exported-using-ghexport Differential Revision: [D81297451](https://our.internmc.facebook.com/intern/diff/D81297451/)

facebook-github-bot · 2025-09-04T23:17:04Z

This pull request was exported from Phabricator. Differential Revision: D81297451

…dels to a target device Pull Request resolved: #13791 This diff introduces multi backend/ target based recipes to lower a model with very less code. Target recipes provide pre-configured backend recipes to use them and retarget if needed. See RFC: #13732 ## Usage ``` from executorch.export import export from executorch.export.target_recipes import get_ios_recipe # CoreML + XNNPACK (FP32) recipe = get_ios_recipe() # default = "ios-arm64-coreml-fp16" session = export(model, recipe, example_inputs) session.save_pte_file("model.pte") ``` ## Advanced usage to combine specific recipes of one or two or backends one can directly use `ExportRecipe.combine_recipes()` to combine specific backend recipes. ``` recipe1 = ExportRecipe.get_recipe(CoreMLRecipeType.FP32) recipe2 = ExportRecipe.get_recipe(XNNPackRecipeType.FP32) combined_recipe = ExportRecipe.combine( [recipe1, recipe2], recipe_name="multi_backend_coreml_xnnpack_fp32" ) session = export(model, combined_recipe, example_inputs) ``` Additional changes: 1. Relaxed kwarg validation in the backend providers to just ignore them instead of erroring out. Fixes: #13732 ghstack-source-id: 307671542 @exported-using-ghexport Differential Revision: [D81297451](https://our.internmc.facebook.com/intern/diff/D81297451/)

abhinaykukkadapu requested review from cccclai, digantdesai and shoumikhin as code owners August 29, 2025 05:53

meta-cla bot added the CLA Signed This label is managed by the Facebook bot. Authors need to sign the CLA before a PR can be reviewed. label Aug 29, 2025

facebook-github-bot added the fb-exported label Aug 29, 2025

abhinaykukkadapu requested review from JacobSzwejbka, kimishpatel, mergennachin and metascroy and removed request for shoumikhin August 29, 2025 05:54

kimishpatel reviewed Sep 3, 2025

View reviewed changes

export/recipe.py Show resolved Hide resolved

kimishpatel reviewed Sep 3, 2025

View reviewed changes

export/recipe.py Show resolved Hide resolved

kimishpatel reviewed Sep 3, 2025

View reviewed changes

kimishpatel approved these changes Sep 4, 2025

View reviewed changes

abhinaykukkadapu merged commit e4bf93e into gh/abhinaykukkadapu/8/base Sep 5, 2025
121 of 125 checks passed

abhinaykukkadapu deleted the gh/abhinaykukkadapu/8/head branch September 5, 2025 01:25

abhinaykukkadapu temporarily deployed to cherry-pick-bot September 5, 2025 01:25 — with GitHub Actions Inactive

pytorchbot mentioned this pull request Sep 5, 2025

[Executorch][target recipes] Add target based recipes for lowering models to a target device #13983

Merged

[Executorch][target recipes] Add target based recipes for lowering models to a target device #13791

[Executorch][target recipes] Add target based recipes for lowering models to a target device #13791

Uh oh!

Conversation

abhinaykukkadapu commented Aug 29, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Usage

Advanced usage to combine specific recipes of one or two or backends

Uh oh!

pytorch-bot bot commented Aug 29, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

🔗 Helpful Links

🧪 See artifacts and rendered test results at hud.pytorch.org/pr/pytorch/executorch/13791

❌ 2 New Failures

Uh oh!

facebook-github-bot commented Aug 29, 2025

Uh oh!

github-actions bot commented Aug 29, 2025

This PR needs a release notes: label

Uh oh!

facebook-github-bot commented Aug 29, 2025

Uh oh!

kimishpatel commented Aug 29, 2025

Uh oh!

facebook-github-bot commented Sep 2, 2025

Uh oh!

facebook-github-bot commented Sep 3, 2025

Uh oh!

abhinaykukkadapu commented Sep 3, 2025

Uh oh!

facebook-github-bot commented Sep 3, 2025

Uh oh!

kimishpatel Sep 3, 2025

Choose a reason for hiding this comment

Uh oh!

abhinaykukkadapu Sep 3, 2025

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Uh oh!

kimishpatel left a comment

Choose a reason for hiding this comment

Uh oh!

abhinaykukkadapu commented Sep 3, 2025

Uh oh!

facebook-github-bot commented Sep 3, 2025

Uh oh!

facebook-github-bot commented Sep 3, 2025

Uh oh!

facebook-github-bot commented Sep 3, 2025

Uh oh!

facebook-github-bot commented Sep 4, 2025

Uh oh!

facebook-github-bot commented Sep 4, 2025

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

3 participants

abhinaykukkadapu commented Aug 29, 2025 •

edited

Loading

pytorch-bot bot commented Aug 29, 2025 •

edited

Loading

This PR needs a `release notes:` label