Remove the ability to choose pybinding when installing ExecuTorch (#11233)

### Summary
Fixes #11176

### Test plan
CI

```
rg "\-\-pybind" --hidden -g '!**/{third-party,.git}'
```

Author: jathu

.github/workflows/trunk.yml

@@ -305,7 +305,7 @@ jobs:
         # Install requirements
         ${CONDA_RUN} sh install_requirements.sh
         ${CONDA_RUN} sh backends/apple/coreml/scripts/install_requirements.sh
-        ${CONDA_RUN} python install_executorch.py --pybind coreml
+        ${CONDA_RUN} python install_executorch.py
         ${CONDA_RUN} sh examples/models/llama/install_requirements.sh
 
         # Test ANE llama

backends/apple/mps/setup.md

@@ -91,7 +91,7 @@ I 00:00:00.122615 executorch:mps_executor_runner.mm:501] Model verified successf
 ### [Optional] Run the generated model directly using pybind
 1. Make sure `pybind` MPS support was installed:
 ```bash
-./install_executorch.sh --pybind mps
+CMAKE_ARGS="-DEXECUTORCH_BUILD_MPS=ON" ./install_executorch.sh
 ```
 2. Run the `mps_example` script to trace the model and run it directly from python:
 ```bash

docs/source/backends-mps.md

@@ -91,7 +91,7 @@ I 00:00:00.122615 executorch:mps_executor_runner.mm:501] Model verified successf
 ### [Optional] Run the generated model directly using pybind
 1. Make sure `pybind` MPS support was installed:
 ```bash
-./install_executorch.sh --pybind mps
+CMAKE_ARGS="-DEXECUTORCH_BUILD_MPS=ON" ./install_executorch.sh
 ```
 2. Run the `mps_example` script to trace the model and run it directly from python:
 ```bash

docs/source/using-executorch-building-from-source.md

@@ -64,25 +64,15 @@ Or alternatively, [install conda on your machine](https://conda.io/projects/cond
    ./install_executorch.sh
    ```
 
-   Use the [`--pybind` flag](https://github.com/pytorch/executorch/blob/main/install_executorch.sh#L26-L29) to install with pybindings and dependencies for other backends.
-   ```bash
-   ./install_executorch.sh --pybind <coreml | mps | xnnpack>
-
-   # Example: pybindings with CoreML *only*
-   ./install_executorch.sh --pybind coreml
-
-   # Example: pybinds with CoreML *and* XNNPACK
-   ./install_executorch.sh --pybind coreml xnnpack
-   ```
+   Not all backends are built into the pip wheel by default. You can link these missing/experimental backends by turning on the corresponding cmake flag. For example, to include the MPS backend:
 
-   By default, `./install_executorch.sh` command installs pybindings for XNNPACK. To disable any pybindings altogether:
-   ```bash
-   ./install_executorch.sh --pybind off
-   ```
+  ```bash
+  CMAKE_ARGS="-DEXECUTORCH_BUILD_MPS=ON" ./install_executorch.sh
+  ```
 
    For development mode, run the command with `--editable`, which allows us to modify Python source code and see changes reflected immediately.
    ```bash
-   ./install_executorch.sh --editable [--pybind xnnpack]
+   ./install_executorch.sh --editable
 
    # Or you can directly do the following if dependencies are already installed
    # either via a previous invocation of `./install_executorch.sh` or by explicitly installing requirements via `./install_requirements.sh` first.

examples/demo-apps/react-native/rnllama/README.md

@@ -26,7 +26,7 @@ A React Native mobile application for running LLaMA language models using ExecuT
 
 3. Pull submodules: `git submodule sync && git submodule update --init`
 
-4. Install dependencies: `./install_executorch.sh --pybind xnnpack && ./examples/models/llama/install_requirements.sh`
+4. Install dependencies: `./install_executorch.sh && ./examples/models/llama/install_requirements.sh`
 
 5. Follow the instructions in the [README](https://github.com/pytorch/executorch/blob/main/examples/models/llama/README.md#option-a-download-and-export-llama32-1b3b-model) to export a model as `.pte`
 

examples/models/llama/README.md

@@ -148,7 +148,7 @@ Llama 3 8B performance was measured on the Samsung Galaxy S22, S24, and OnePlus
 ## Step 1: Setup
 > :warning: **double check your python environment**: make sure `conda activate <VENV>` is run before all the bash and python scripts.
 
-1. Follow the [tutorial](https://pytorch.org/executorch/main/getting-started-setup) to set up ExecuTorch. For installation run `./install_executorch.sh --pybind xnnpack`
+1. Follow the [tutorial](https://pytorch.org/executorch/main/getting-started-setup) to set up ExecuTorch. For installation run `./install_executorch.sh`
 2. Run `examples/models/llama/install_requirements.sh` to install a few dependencies.
 
 
@@ -528,7 +528,7 @@ This example tries to reuse the Python code, with minimal modifications to make
 git clean -xfd
 pip uninstall executorch
 ./install_executorch.sh --clean
-./install_executorch.sh --pybind xnnpack
+./install_executorch.sh
 ```
 - If you encounter `pthread` related issues during link time, add `pthread` in `target_link_libraries` in `CMakeLists.txt`
 - On Mac, if there is linking error in Step 4 with error message like

examples/models/phi-3-mini/README.md

@@ -3,7 +3,7 @@ This example demonstrates how to run a [Phi-3-mini](https://huggingface.co/micro
 
 # Instructions
 ## Step 1: Setup
-1. Follow the [tutorial](https://pytorch.org/executorch/main/getting-started-setup) to set up ExecuTorch. For installation run `./install_executorch.sh --pybind xnnpack`
+1. Follow the [tutorial](https://pytorch.org/executorch/main/getting-started-setup) to set up ExecuTorch. For installation run `./install_executorch.sh`
 2. Currently, we support transformers v4.44.2. Install transformers with the following command:
 ```
 pip uninstall -y transformers ; pip install transformers==4.44.2

extension/pybindings/README.md

@@ -2,28 +2,18 @@
 This Python module, named `portable_lib`, provides a set of functions and classes for loading and executing bundled programs. To install it, run the fullowing command:
 
 ```bash
-CMAKE_ARGS="-DEXECUTORCH_BUILD_XNNPACK=ON" pip install . --no-build-isolation
-```
-
-Or when installing the rest of dependencies:
+./install_executorch.sh
 
-```bash
-install_executorch.sh --pybind
+# ...or use pip directly
+pip install . --no-build-isolation
 ```
 
 # Link Backends
 
-You can link the runtime against some backends to make sure a delegated or partitioned model can still run by Python module successfully:
-
-```bash
-CMAKE_ARGS="-DEXECUTORCH_BUILD_XNNPACK=ON -DEXECUTORCH_BUILD_COREML=ON -DEXECUTORCH_BUILD_MPS=ON" \
-  pip install . --no-build-isolation
-```
-
-Similarly, when installing the rest of dependencies:
+Not all backends are built into the pip wheel by default. You can link these missing/experimental backends by turning on the corresponding cmake flag. For example, to include the MPS backend:
 
 ```bash
-install_executorch.sh --pybind xnnpack coreml mps
+CMAKE_ARGS="-DEXECUTORCH_BUILD_MPS=ON" ./install_executorch.sh
 ```
 
 ## Functions

install_executorch.py

@@ -8,14 +8,12 @@
 
 import argparse
 import glob
-import itertools
 import logging
 import os
 import shutil
 import subprocess
 import sys
 from contextlib import contextmanager
-from typing import List, Tuple
 
 from install_requirements import (
     install_requirements,
@@ -52,10 +50,6 @@ def clean():
     print("Done cleaning build artifacts.")
 
 
-# Please keep this insync with `ShouldBuild.pybindings` in setup.py.
-VALID_PYBINDS = ["coreml", "mps", "xnnpack", "training", "openvino"]
-
-
 ################################################################################
 # Git submodules
 ################################################################################
@@ -139,14 +133,9 @@ def check_folder(folder: str, file: str) -> bool:
     logger.info("All required submodules are present.")
 
 
-def build_args_parser() -> argparse.ArgumentParser:
-    # Parse options.
-    parser = argparse.ArgumentParser()
-    parser.add_argument(
-        "--pybind",
-        action="append",
-        nargs="+",
-        help="one or more of coreml/mps/xnnpack, or off",
+def _parse_args() -> argparse.Namespace:
+    parser = argparse.ArgumentParser(
+        description="Install executorch in your Python environment."
     )
     parser.add_argument(
         "--clean",
@@ -166,83 +155,34 @@ def build_args_parser() -> argparse.ArgumentParser:
         "picked up without rebuilding the wheel. Extension libraries will be "
         "installed inside the source tree.",
     )
-    return parser
-
-
-# Returns (wants_off, wanted_pybindings)
-def _list_pybind_defines(args) -> Tuple[bool, List[str]]:
-    if args.pybind is None:
-        return False, []
-
-    # Flatten list of lists.
-    args.pybind = list(itertools.chain(*args.pybind))
-    if "off" in args.pybind:
-        if len(args.pybind) != 1:
-            raise Exception(f"Cannot combine `off` with other pybinds: {args.pybind}")
-        return True, []
-
-    cmake_args = []
-    for pybind_arg in args.pybind:
-        if pybind_arg not in VALID_PYBINDS:
-            raise Exception(
-                f"Unrecognized pybind argument {pybind_arg}; valid options are: {', '.join(VALID_PYBINDS)}"
-            )
-        if pybind_arg == "training":
-            cmake_args.append("-DEXECUTORCH_BUILD_EXTENSION_TRAINING=ON")
-        else:
-            cmake_args.append(f"-DEXECUTORCH_BUILD_{pybind_arg.upper()}=ON")
-
-    return False, cmake_args
+    return parser.parse_args()
 
 
 def main(args):
     if not python_is_compatible():
         sys.exit(1)
 
-    parser = build_args_parser()
-    args = parser.parse_args()
-
-    cmake_args = [os.getenv("CMAKE_ARGS", "")]
-    use_pytorch_nightly = True
-
-    wants_pybindings_off, pybind_defines = _list_pybind_defines(args)
-    if wants_pybindings_off:
-        cmake_args.append("-DEXECUTORCH_BUILD_PYBIND=OFF")
-    else:
-        cmake_args += pybind_defines
+    args = _parse_args()
 
     if args.clean:
         clean()
         return
 
-    if args.use_pt_pinned_commit:
-        # This option is used in CI to make sure that PyTorch build from the pinned commit
-        # is used instead of nightly. CI jobs wouldn't be able to catch regression from the
-        # latest PT commit otherwise
-        use_pytorch_nightly = False
-
+    cmake_args = [os.getenv("CMAKE_ARGS", "")]
     # Use ClangCL on Windows.
     # ClangCL is an alias to Clang that configures it to work in an MSVC-compatible
     # mode. Using it on Windows to avoid compiler compatibility issues for MSVC.
     if os.name == "nt":
         cmake_args.append("-T ClangCL")
-
-    #
-    # Install executorch pip package. This also makes `flatc` available on the path.
-    # The --extra-index-url may be necessary if pyproject.toml has a dependency on a
-    # pre-release or nightly version of a torch package.
-    #
-
-    # Set environment variables
     os.environ["CMAKE_ARGS"] = " ".join(cmake_args)
 
-    # Check if the required submodules are present and update them if not
     check_and_update_submodules()
-
-    install_requirements(use_pytorch_nightly)
-
-    # Run the pip install command
-    subprocess.run(
+    # This option is used in CI to make sure that PyTorch build from the pinned commit
+    # is used instead of nightly. CI jobs wouldn't be able to catch regression from the
+    # latest PT commit otherwise
+    install_requirements(use_pytorch_nightly=not args.use_pt_pinned_commit)
+    os.execvp(
+        sys.executable,
         [
             sys.executable,
             "-m",
@@ -257,14 +197,10 @@ def main(args):
             "--extra-index-url",
             TORCH_NIGHTLY_URL,
         ],
-        check=True,
     )
 
 
 if __name__ == "__main__":
     # Before doing anything, cd to the directory containing this script.
     os.chdir(os.path.dirname(os.path.abspath(__file__)))
-    if not python_is_compatible():
-        sys.exit(1)
-
     main(sys.argv[1:])