graxpert 3.2.0a0.dev5__tar.gz → 3.2.0a0.dev7__tar.gz

{graxpert-3.2.0a0.dev5 → graxpert-3.2.0a0.dev7}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: graxpert
-Version: 3.2.0a0.dev5
+Version: 3.2.0a0.dev7
 Summary: GraXpert is an astronomical image processing program for extracting and removing gradients in the background of your astrophotos
 Home-page: https://graxpert.com
 Author: GraXpert Development Team
@@ -95,17 +95,29 @@ We do this via the the python [pypi.org](https://pypi.org/project/graxpert/) rep
 1. Install [python](https://www.python.org/downloads/) runtime for your OS.  The default install should automatically include the python package manager 'pip'.
 2. At a shell prompt type the correct flavor of graxpert to install for your needs.
 
-FIXME, recommend directml for windows, FIXME for linux and FIXME for mac
-
-* **pip install graxpert\[cpuonly\]** - Use this if you are on OS-X - it is poorly named, but it will still use the CoreML Apple GPU support
+* **brew install pipx python-tk@3.13; pipx install --python python3.13 graxpert\\\[cpuonly\\\]** - Use this if you are on OS-X - it is poorly named, but it will still use the CoreML Apple GPU support
 * **pip install graxpert\[cuda\]** - Use this if you have an Nvidia GPU
-* **pip install graxpert\[rocm\]** - Use this if you have an AMD GPU
+* **pip install graxpert\[rocm\]** --pip-args="-f https://repo.radeon.com/rocm/manylinux/rocm-rel-7.0/" - Use this if you have an AMD GPU
 * **pip install graxpert\[openvino\]** - Use this if you don't have an advanced GPU or don't know what you have (provides sizable acceleration on most modern Intel CPUs)
 * **pip install graxpert\[directml\]** - Use this if you want the Windows DirectML acceleration (not as fast as the options above but widely supported and easy to install)
 * **pip install graxpert\[cpuonly\]** - If you encounter bugs with any of our GPU versions, please file a bug and temporarily use this version (which excludes all GPU code)
 
 If you are an advanced python user and know about the 'pipx' tool, we recommend that you use that tool instead of pip for installs.
 
+## AMD GPU library installation
+
+If you are using an AMD GPU and the python packages above, you'll probably need to install the latest AMD GPU software.
+
+For Ubuntu linux the instructions run:
+```
+wget https://repo.radeon.com/amdgpu-install/7.0.1/ubuntu/noble/amdgpu-install_7.0.1.70001-1_all.deb
+sudo apt install ./amdgpu-install_7.0.1.70001-1_all.deb
+sudo apt update
+sudo apt install rocm python3-tk
+sudo usermod -a -G render,video $LOGNAME # Add the current user to the render and video groups
+```
+For other OSes see the instructions [here](https://rocm.docs.amd.com/projects/install-on-linux/en/latest/install/quick-start.html).
+
 # Legacy Installation
 You can download the latest official release of GraXpert [here](https://github.com/Steffenhir/GraXpert/releases/latest). Select the correct version for your operating system. For macOS, we provide different versions
 for Intel processors (x86_64) and for apple silicon (arm64).
@@ -200,7 +212,7 @@ For macOS, we have to install tkinter separately.
 We use the version provided by brew because it is newer
 and solves issues with macOS Sonoma. Please use the version matching with your Python version.
 """
-brew install python-tk@3.10
+brew install python-tk
 ```
 
 ## Option 2: Automatic setup with devcontainers

{graxpert-3.2.0a0.dev5 → graxpert-3.2.0a0.dev7}/README.md

@@ -29,17 +29,29 @@ We do this via the the python [pypi.org](https://pypi.org/project/graxpert/) rep
 1. Install [python](https://www.python.org/downloads/) runtime for your OS.  The default install should automatically include the python package manager 'pip'.
 2. At a shell prompt type the correct flavor of graxpert to install for your needs.
 
-FIXME, recommend directml for windows, FIXME for linux and FIXME for mac
-
-* **pip install graxpert\[cpuonly\]** - Use this if you are on OS-X - it is poorly named, but it will still use the CoreML Apple GPU support
+* **brew install pipx python-tk@3.13; pipx install --python python3.13 graxpert\\\[cpuonly\\\]** - Use this if you are on OS-X - it is poorly named, but it will still use the CoreML Apple GPU support
 * **pip install graxpert\[cuda\]** - Use this if you have an Nvidia GPU
-* **pip install graxpert\[rocm\]** - Use this if you have an AMD GPU
+* **pip install graxpert\[rocm\]** --pip-args="-f https://repo.radeon.com/rocm/manylinux/rocm-rel-7.0/" - Use this if you have an AMD GPU
 * **pip install graxpert\[openvino\]** - Use this if you don't have an advanced GPU or don't know what you have (provides sizable acceleration on most modern Intel CPUs)
 * **pip install graxpert\[directml\]** - Use this if you want the Windows DirectML acceleration (not as fast as the options above but widely supported and easy to install)
 * **pip install graxpert\[cpuonly\]** - If you encounter bugs with any of our GPU versions, please file a bug and temporarily use this version (which excludes all GPU code)
 
 If you are an advanced python user and know about the 'pipx' tool, we recommend that you use that tool instead of pip for installs.
 
+## AMD GPU library installation
+
+If you are using an AMD GPU and the python packages above, you'll probably need to install the latest AMD GPU software.
+
+For Ubuntu linux the instructions run:
+```
+wget https://repo.radeon.com/amdgpu-install/7.0.1/ubuntu/noble/amdgpu-install_7.0.1.70001-1_all.deb
+sudo apt install ./amdgpu-install_7.0.1.70001-1_all.deb
+sudo apt update
+sudo apt install rocm python3-tk
+sudo usermod -a -G render,video $LOGNAME # Add the current user to the render and video groups
+```
+For other OSes see the instructions [here](https://rocm.docs.amd.com/projects/install-on-linux/en/latest/install/quick-start.html).
+
 # Legacy Installation
 You can download the latest official release of GraXpert [here](https://github.com/Steffenhir/GraXpert/releases/latest). Select the correct version for your operating system. For macOS, we provide different versions
 for Intel processors (x86_64) and for apple silicon (arm64).
@@ -134,7 +146,7 @@ For macOS, we have to install tkinter separately.
 We use the version provided by brew because it is newer
 and solves issues with macOS Sonoma. Please use the version matching with your Python version.
 """
-brew install python-tk@3.10
+brew install python-tk
 ```
 
 ## Option 2: Automatic setup with devcontainers

{graxpert-3.2.0a0.dev5 → graxpert-3.2.0a0.dev7}/graxpert/ai_model_handling.py

@@ -213,3 +213,40 @@ def get_execution_providers_ordered(gpu_acceleration=True):
         "If you are using one of our prebuilt executables, please file a bug with the following information:\n"
         "{}".format(e))
         sys.exit(1)
+
+# Provides a mutable session context that can swap out different sessions if needed
+class SessionContext:
+    model_name: str
+    session: any # onnxruntime.InferenceSession
+
+    def __init__(self, model_name: str,  gpu_acceleration: bool = True):
+        """Initialize the ONNX model session with the specified execution provider."""
+
+        providers = get_execution_providers_ordered(gpu_acceleration)
+        logging.info(f"Available inference providers : {providers}")
+
+        import onnxruntime as ort # Must be after get_execution_providers_ordered (so it can check for missing libs)
+        self.session = ort.InferenceSession(model_name, providers=providers)
+        self.model_name = model_name
+
+        logging.info(f"Used inference providers for {model_name}: {self.session.get_providers()}")
+
+
+    def run(self, model_args: dict) -> any:
+        """Run the ONNX model using the specified execution provider."""
+
+        providers = self.session.get_providers()
+        gpu_acceleration = len(providers) > 1 or providers[0] != "CPUExecutionProvider"
+        try:
+            result = self.session.run(None, model_args)
+        except Exception as e:
+            if not gpu_acceleration:
+                raise  # Rethrow, the failure occurred with the regular CPU model also - show error dialog
+
+            logging.warning(f"Error running model, falling back to CPU only: {e}")
+            import onnxruntime as ort
+            self.session = ort.InferenceSession(self.model_name, providers=get_execution_providers_ordered(False))
+            result = self.session.run(None, model_args)
+
+        # all graxpert results are in the first array index
+        return result[0]

{graxpert-3.2.0a0.dev5 → graxpert-3.2.0a0.dev7}/graxpert/background_extraction.py

@@ -12,7 +12,7 @@ from astropy.stats import sigma_clipped_stats
 from pykrige.ok import OrdinaryKriging
 from scipy import interpolate, linalg
 
-from graxpert.ai_model_handling import get_execution_providers_ordered
+from graxpert.ai_model_handling import SessionContext
 from graxpert.mp_logging import get_logging_queue, worker_configurer
 from graxpert.parallel_processing import executor
 from graxpert.radialbasisinterpolation import RadialBasisInterpolation
@@ -69,14 +69,9 @@ def extract_background(in_imarray, background_points, interpolation_type, smooth
         if progress is not None:
             progress.update(8)
 
-        providers = get_execution_providers_ordered(ai_gpu_acceleration)
-        import onnxruntime as ort # Must be after get_execution_providers_ordered
-        session = ort.InferenceSession(ai_path, providers=providers)
-
-        logging.info(f"Providers : {providers}")
-        logging.info(f"Used providers : {session.get_providers()}")
-
-        background = session.run(None, {"gen_input_image": np.expand_dims(imarray_shrink, axis=0)})[0][0]
+        # allow persistent changes to gpu_accel in case of failure
+        context = SessionContext(ai_path, gpu_acceleration=ai_gpu_acceleration)
+        background = context.run({"gen_input_image": np.expand_dims(imarray_shrink, axis=0)})[0]
 
         background = background / 0.04 * mad + median
 

{graxpert-3.2.0a0.dev5 → graxpert-3.2.0a0.dev7}/graxpert/deconvolution.py

@@ -3,7 +3,7 @@ import logging
 
 import numpy as np
 
-from graxpert.ai_model_handling import get_execution_providers_ordered
+from graxpert.ai_model_handling import SessionContext
 from graxpert.application.app_events import AppEvents
 from graxpert.application.eventbus import eventbus
 
@@ -63,13 +63,6 @@ def deconvolve(image, ai_path, strength, psfsize, batch_size=4, window_size=512,
 
     output = copy.deepcopy(image)
 
-    providers = get_execution_providers_ordered(ai_gpu_acceleration)
-    import onnxruntime as ort # Must be after get_execution_providers_ordered
-    session = ort.InferenceSession(ai_path, providers=providers)
-
-    logging.info(f"Available inference providers : {providers}")
-    logging.info(f"Used inference providers : {session.get_providers()}")
-
     cancel_flag = False
 
     def cancel_listener(event):
@@ -78,6 +71,8 @@ def deconvolve(image, ai_path, strength, psfsize, batch_size=4, window_size=512,
 
     eventbus.add_listener(AppEvents.CANCEL_PROCESSING, cancel_listener)
 
+    # allow persistent changes to gpu_accel in case of failure
+    context = SessionContext(ai_path, gpu_acceleration=ai_gpu_acceleration)
     last_progress = 0
     for b in range(0, ith * itw + batch_size, batch_size):
 
@@ -129,9 +124,10 @@ def deconvolve(image, ai_path, strength, psfsize, batch_size=4, window_size=512,
         strenght_p = np.full(shape=(input_tiles.shape[0], 1), fill_value=strength, dtype=np.float32)
         conds = np.concatenate([sigma, strenght_p], axis=-1)
         if type == "Obj" and "1.0.0" in ai_path:
-            session_result = session.run(None, {"gen_input_image": input_tiles, "sigma": sigma, "strenght": strenght_p})[0]
+            model_args = {"gen_input_image": input_tiles, "sigma": sigma, "strenght": strenght_p}
         else:
-            session_result = session.run(None, {"gen_input_image": input_tiles, "params": conds})[0]
+            model_args = {"gen_input_image": input_tiles, "params": conds}
+        session_result = context.run(model_args)
         for e in session_result:
             output_tiles.append(e)
 

{graxpert-3.2.0a0.dev5 → graxpert-3.2.0a0.dev7}/graxpert/denoising.py

@@ -4,7 +4,7 @@ import time
 
 import numpy as np
 
-from graxpert.ai_model_handling import get_execution_providers_ordered
+from graxpert.ai_model_handling import SessionContext
 from graxpert.application.app_events import AppEvents
 from graxpert.application.eventbus import eventbus
 from graxpert.ui.ui_events import UiEvents
@@ -64,13 +64,6 @@ def denoise(image, ai_path, strength, batch_size=4, window_size=256, stride=128,
 
     output = copy.deepcopy(image)
 
-    providers = get_execution_providers_ordered(ai_gpu_acceleration)
-    import onnxruntime as ort # Must be after get_execution_providers_ordered
-    session = ort.InferenceSession(ai_path, providers=providers)
-
-    logging.info(f"Available inference providers : {providers}")
-    logging.info(f"Used inference providers : {session.get_providers()}")
-
     cancel_flag = False
 
     def cancel_listener(event):
@@ -79,6 +72,8 @@ def denoise(image, ai_path, strength, batch_size=4, window_size=256, stride=128,
 
     eventbus.add_listener(AppEvents.CANCEL_PROCESSING, cancel_listener)
 
+    # allow persistent changes to gpu_accel in case of failure
+    context = SessionContext(ai_path, gpu_acceleration=ai_gpu_acceleration)
     last_progress = 0
     for b in range(0, ith * itw + batch_size, batch_size):
 
@@ -114,7 +109,7 @@ def denoise(image, ai_path, strength, batch_size=4, window_size=256, stride=128,
         input_tiles = np.array(input_tiles)
 
         output_tiles = []
-        session_result = session.run(None, {"gen_input_image": input_tiles})[0]
+        session_result = context.run({ "gen_input_image": input_tiles})
         for e in session_result:
             output_tiles.append(e)
 

{graxpert-3.2.0a0.dev5 → graxpert-3.2.0a0.dev7}/graxpert/main.py

@@ -28,7 +28,8 @@ except ImportError:
     print("  For Debian/Ubuntu: sudo apt-get install python3-tk", file=sys.stderr)
     print("  For Fedora:        sudo dnf install python3-tkinter", file=sys.stderr)
     print("  For Arch Linux:    sudo pacman -S tk", file=sys.stderr)
-    print("  For Windows/OS-X:  It should be included with your Python installation by default.", file=sys.stderr)
+    print("  For OS-X:          brew install python-tk", file=sys.stderr)
+    print("  For Windows:       It should be included with your Python installation by default.", file=sys.stderr)
     sys.exit(1) # Exit with error
 
 

graxpert-3.2.0a0.dev7/graxpert/version.py

@@ -0,0 +1,2 @@
+release = "the github CI DMG builds are not yet reliable, don't require them before pypi publish"
+version = "3.2.0.a0.dev7"

graxpert-3.2.0a0.dev7/graxpert/version.py.bak

@@ -0,0 +1,2 @@
+release = "the github CI DMG builds are not yet reliable, don't require them before pypi publish"
+version = "SNAPSHOT"

{graxpert-3.2.0a0.dev5 → graxpert-3.2.0a0.dev7}/graxpert.egg-info/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: graxpert
-Version: 3.2.0a0.dev5
+Version: 3.2.0a0.dev7
 Summary: GraXpert is an astronomical image processing program for extracting and removing gradients in the background of your astrophotos
 Home-page: https://graxpert.com
 Author: GraXpert Development Team
@@ -95,17 +95,29 @@ We do this via the the python [pypi.org](https://pypi.org/project/graxpert/) rep
 1. Install [python](https://www.python.org/downloads/) runtime for your OS.  The default install should automatically include the python package manager 'pip'.
 2. At a shell prompt type the correct flavor of graxpert to install for your needs.
 
-FIXME, recommend directml for windows, FIXME for linux and FIXME for mac
-
-* **pip install graxpert\[cpuonly\]** - Use this if you are on OS-X - it is poorly named, but it will still use the CoreML Apple GPU support
+* **brew install pipx python-tk@3.13; pipx install --python python3.13 graxpert\\\[cpuonly\\\]** - Use this if you are on OS-X - it is poorly named, but it will still use the CoreML Apple GPU support
 * **pip install graxpert\[cuda\]** - Use this if you have an Nvidia GPU
-* **pip install graxpert\[rocm\]** - Use this if you have an AMD GPU
+* **pip install graxpert\[rocm\]** --pip-args="-f https://repo.radeon.com/rocm/manylinux/rocm-rel-7.0/" - Use this if you have an AMD GPU
 * **pip install graxpert\[openvino\]** - Use this if you don't have an advanced GPU or don't know what you have (provides sizable acceleration on most modern Intel CPUs)
 * **pip install graxpert\[directml\]** - Use this if you want the Windows DirectML acceleration (not as fast as the options above but widely supported and easy to install)
 * **pip install graxpert\[cpuonly\]** - If you encounter bugs with any of our GPU versions, please file a bug and temporarily use this version (which excludes all GPU code)
 
 If you are an advanced python user and know about the 'pipx' tool, we recommend that you use that tool instead of pip for installs.
 
+## AMD GPU library installation
+
+If you are using an AMD GPU and the python packages above, you'll probably need to install the latest AMD GPU software.
+
+For Ubuntu linux the instructions run:
+```
+wget https://repo.radeon.com/amdgpu-install/7.0.1/ubuntu/noble/amdgpu-install_7.0.1.70001-1_all.deb
+sudo apt install ./amdgpu-install_7.0.1.70001-1_all.deb
+sudo apt update
+sudo apt install rocm python3-tk
+sudo usermod -a -G render,video $LOGNAME # Add the current user to the render and video groups
+```
+For other OSes see the instructions [here](https://rocm.docs.amd.com/projects/install-on-linux/en/latest/install/quick-start.html).
+
 # Legacy Installation
 You can download the latest official release of GraXpert [here](https://github.com/Steffenhir/GraXpert/releases/latest). Select the correct version for your operating system. For macOS, we provide different versions
 for Intel processors (x86_64) and for apple silicon (arm64).
@@ -200,7 +212,7 @@ For macOS, we have to install tkinter separately.
 We use the version provided by brew because it is newer
 and solves issues with macOS Sonoma. Please use the version matching with your Python version.
 """
-brew install python-tk@3.10
+brew install python-tk
 ```
 
 ## Option 2: Automatic setup with devcontainers

{graxpert-3.2.0a0.dev5 → graxpert-3.2.0a0.dev7}/graxpert.egg-info/top_level.txt

@@ -1 +1,2 @@
 graxpert
+tests

{graxpert-3.2.0a0.dev5 → graxpert-3.2.0a0.dev7}/setup.py

@@ -128,7 +128,9 @@ setup_options = {
 
 # We are careful to manually select which onnxruntime version we want based on target architecture
 # also exclude build time only tools
-packages = setuptools.find_packages(exclude=["tests"])
+# we can't exclude tests because that apparently breaks runtime numpy for the windows msi build
+# exclude=["tests"]
+packages = setuptools.find_packages()
 print(f"Including the following packages: {packages}")
 
 # import cx_Freeze only when needed

graxpert-3.2.0a0.dev5/graxpert/version.py

@@ -1,2 +0,0 @@
-release = "test pypi based builds on release generation"
-version = "3.2.0.a0.dev5"

graxpert-3.2.0a0.dev5/graxpert/version.py.bak

@@ -1,2 +0,0 @@
-release = "test pypi based builds on release generation"
-version = "SNAPSHOT"