py2edg 0.1.0__py3-none-any.whl

py2edg/__init__.py

@@ -0,0 +1,80 @@
+"""
+╔══════════════════════════════════════════════════════════════════════╗
+║                        Py2Edg v0.1.0                              ║
+║   One-line edge deployment for Computer Vision models              ║
+║   by Mouissat Rabah Abderrahmane                                   ║
+╚══════════════════════════════════════════════════════════════════════╝
+
+Convert, quantize, optimize, and benchmark CV models for edge devices
+with a single function call.
+
+Quick Start:
+    >>> import py2edg as rcv
+    >>>
+    >>> # Convert PyTorch model to optimized ONNX
+    >>> result = rcv.convert("model.pt", target="onnx", quantize="fp16")
+    >>>
+    >>> # Full deployment pipeline: convert + optimize + benchmark
+    >>> report = rcv.deploy("model.pt", device="rpi4", quantize="int8")
+    >>>
+    >>> # Benchmark any model
+    >>> stats = rcv.benchmark("model.onnx", input_shape=(1, 3, 640, 640))
+    >>>
+    >>> # Compare original vs optimized
+    >>> rcv.compare("model.pt", "model_opt.onnx", input_shape=(1, 3, 640, 640))
+"""
+
+__version__ = "0.1.0"
+__author__ = "Mouissat Rabah Abderrahmane"
+__license__ = "MIT"
+
+# ── High-level API (the magic) ──
+from py2edg.api import (
+    convert,
+    deploy,
+    benchmark,
+    compare,
+    profile,
+    validate,
+    inspect_model,
+)
+
+# ── Device Profiles ──
+from py2edg.devices import (
+    DeviceProfile,
+    get_device,
+    list_devices,
+    register_device,
+)
+
+# ── Deployment Recipe ──
+from py2edg.recipe import (
+    DeployRecipe,
+    load_recipe,
+    save_recipe,
+)
+
+# ── Report ──
+from py2edg.report import DeployReport
+
+__all__ = [
+    # API
+    "convert",
+    "deploy",
+    "benchmark",
+    "compare",
+    "profile",
+    "validate",
+    "inspect_model",
+    # Devices
+    "DeviceProfile",
+    "get_device",
+    "list_devices",
+    "register_device",
+    # Recipe
+    "DeployRecipe",
+    "load_recipe",
+    "save_recipe",
+    # Report
+    "DeployReport",
+]

py2edg/_imports.py

@@ -0,0 +1,60 @@
+"""Lazy import helpers for optional dependencies."""
+
+from __future__ import annotations
+
+import importlib
+from typing import Any
+
+
+class _LazyModule:
+    """Delay ImportError until the module is actually used."""
+
+    def __init__(self, name: str, install_hint: str):
+        self._name = name
+        self._hint = install_hint
+        self._mod = None
+
+    def _load(self):
+        if self._mod is None:
+            try:
+                self._mod = importlib.import_module(self._name)
+            except ImportError:
+                raise ImportError(
+                    f"'{self._name}' is required for this operation.\n"
+                    f"Install it with:  pip install {self._hint}"
+                )
+        return self._mod
+
+    def __getattr__(self, item: str) -> Any:
+        return getattr(self._load(), item)
+
+
+def require(module_name: str, install_hint: str | None = None):
+    """Import a module or raise a helpful error."""
+    hint = install_hint or module_name
+    try:
+        return importlib.import_module(module_name)
+    except ImportError:
+        raise ImportError(
+            f"'{module_name}' is required for this operation.\n"
+            f"Install it with:  pip install {hint}"
+        )
+
+
+def is_available(module_name: str) -> bool:
+    """Check if a module is importable."""
+    try:
+        importlib.import_module(module_name)
+        return True
+    except ImportError:
+        return False
+
+
+# ── Pre-built lazy imports ──
+torch = _LazyModule("torch", "torch")
+torchvision = _LazyModule("torchvision", "torchvision")
+onnx = _LazyModule("onnx", "onnx")
+onnxruntime = _LazyModule("onnxruntime", "onnxruntime")
+tensorflow = _LazyModule("tensorflow", "tensorflow")
+openvino = _LazyModule("openvino", "openvino")
+onnxsim = _LazyModule("onnxsim", "onnxsim")