soifunc 0.11.2__tar.gz → 0.12.0__tar.gz

{soifunc-0.11.2 → soifunc-0.12.0}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.3
 Name: soifunc
-Version: 0.11.2
+Version: 0.12.0
 Summary: Soichiro's VapourSynth Functions Collection
 License: MIT
 Author: Josh Holmer
@@ -11,7 +11,7 @@ Classifier: Programming Language :: Python :: 3
 Classifier: Programming Language :: Python :: 3.12
 Classifier: Programming Language :: Python :: 3.13
 Requires-Dist: vapoursynth (>=68)
-Requires-Dist: vsjetpack (>=0.5.1,<0.6.0)
+Requires-Dist: vsjetpack (>=0.7.1,<0.8.0)
 Description-Content-Type: text/markdown
 
 ## soifunc

{soifunc-0.11.2 → soifunc-0.12.0}/pyproject.toml

@@ -1,6 +1,6 @@
 [tool.poetry]
 name = "soifunc"
-version = "0.11.2"
+version = "0.12.0"
 description = "Soichiro's VapourSynth Functions Collection"
 authors = ["Josh Holmer <jholmer.in@gmail.com>"]
 license = "MIT"
@@ -9,7 +9,7 @@ readme = "README.md"
 [tool.poetry.dependencies]
 python = ">=3.12,<4.0"
 vapoursynth = ">=68"
-vsjetpack = "^0.5.1"
+vsjetpack = "^0.7.1"
 
 [tool.poetry.group.dev.dependencies]
 black = "^25.1.0"

{soifunc-0.11.2 → soifunc-0.12.0}/soifunc/interpolate.py

@@ -4,6 +4,7 @@ import platform
 from typing import TYPE_CHECKING
 
 import vstools
+from vsscale import autoselect_backend
 from vstools import vs
 
 if TYPE_CHECKING:
@@ -40,15 +41,7 @@ def rate_doubler(
         model=vsmlrt.RIFEModel.v4_25_heavy,
         # Why these defaults? Because running ML stuff on AMD on Windows sucks hard.
         # Trial and error led me to finally find that ORT_DML works.
-        backend=(
-            backend
-            if backend
-            else (
-                vsmlrt.Backend.ORT_DML()
-                if platform.system() == "Windows"
-                else vsmlrt.Backend.TRT()
-            )
-        ),
+        backend=(backend if backend else autoselect_backend()),
     )
     # TODO: Handle other chroma samplings
     clip = clip.resize.Bicubic(
@@ -98,15 +91,7 @@ def decimation_fixer(
     doubled = vsmlrt.RIFE(
         clip,
         model=vsmlrt.RIFEModel.v4_25_heavy,
-        backend=(
-            backend
-            if backend
-            else (
-                vsmlrt.Backend.ORT_DML()
-                if platform.system() == "Windows"
-                else vsmlrt.Backend.TRT()
-            )
-        ),
+        backend=(backend if backend else autoselect_backend()),
     )
 
     out_clip = None

soifunc-0.12.0/soifunc/resize.py

@@ -0,0 +1,67 @@
+from __future__ import annotations
+
+from vsaa.deinterlacers import NNEDI3
+from vskernels import (
+    Hermite,
+    Spline36,
+)
+from vsscale import ArtCNN
+from vstools import check_variable_format, is_gpu_available, join, vs
+
+__all__ = [
+    "good_resize",
+]
+
+
+def good_resize(
+    clip: vs.VideoNode,
+    width: int,
+    height: int,
+    shift: tuple[float, float] = (0, 0),
+    gpu: bool | None = None,
+    anime: bool = False,
+) -> vs.VideoNode:
+    """High quality resizing filter
+
+    Parameters
+    ----------
+    clip: VideoNode
+        Video clip to apply resizing to.
+    width: int
+        Target width to resize to.
+    height: int
+        Target height to resize to.
+    shift: tuple[float, float], optional
+        Horizontal and vertical amount of shift to apply.
+    gpu: bool, optional
+        Whether to allow usage of GPU for ArtCNN.
+        Defaults to None, which will auto-select based on available mlrt and hardware.
+    anime: bool, optional
+        Enables scalers that are better tuned toward anime.
+        Defaults to False.
+    """
+
+    if gpu is None:
+        gpu = is_gpu_available()
+
+    is_upscale = clip.width < width or clip.height < height
+    chroma_scaler = Spline36()
+
+    # We've ended up where the only special case is anime + upscale + GPU enabled
+    if anime and is_upscale and gpu:
+        luma_scaler = ArtCNN(scaler=Hermite(sigmoid=True))
+    elif is_upscale:
+        luma_scaler = NNEDI3(scaler=Hermite(sigmoid=True))
+    else:
+        luma_scaler = Hermite(sigmoid=True)
+
+    assert check_variable_format(clip, "good_resize")
+
+    luma = luma_scaler.scale(clip, width, height, shift)
+
+    if clip.format.num_planes == 1:
+        return luma
+
+    chroma = chroma_scaler.scale(clip, width, height, shift)
+
+    return join(luma, chroma)

soifunc-0.11.2/soifunc/resize.py

@@ -1,112 +0,0 @@
-from __future__ import annotations
-
-from dataclasses import dataclass
-from inspect import getfullargspec
-from typing import Any
-
-from vsaa.deinterlacers import NNEDI3
-from vskernels import (
-    Hermite,
-    Scaler,
-    ScalerLike,
-    Spline36,
-)
-from vsscale import ArtCNN, GenericScaler
-from vstools import check_variable_format, inject_self, is_gpu_available, join, vs
-
-__all__ = [
-    "good_resize",
-    "HybridScaler",
-]
-
-
-def good_resize(
-    clip: vs.VideoNode,
-    width: int,
-    height: int,
-    shift: tuple[float, float] = (0, 0),
-    gpu: bool | None = None,
-    anime: bool = False,
-) -> vs.VideoNode:
-    """High quality resizing filter
-
-    Parameters
-    ----------
-    clip: VideoNode
-        Video clip to apply resizing to.
-    width: int
-        Target width to resize to.
-    height: int
-        Target height to resize to.
-    shift: tuple[float, float], optional
-        Horizontal and vertical amount of shift to apply.
-    gpu: bool, optional
-        Whether to allow usage of GPU for ArtCNN.
-        Defaults to None, which will auto-select based on available mlrt and hardware.
-    anime: bool, optional
-        Enables scalers that are better tuned toward anime.
-        Defaults to False.
-    """
-
-    if gpu is None:
-        gpu = is_gpu_available()
-
-    is_upscale = clip.width < width or clip.height < height
-    chroma_scaler = Spline36()
-
-    # We've ended up where the only special case is anime + upscale + GPU enabled
-    if anime and is_upscale and gpu:
-        luma_scaler = ArtCNN(scaler=Hermite(sigmoid=True))
-    elif is_upscale:
-        luma_scaler = NNEDI3(scaler=Hermite(sigmoid=True))
-    else:
-        luma_scaler = Hermite(sigmoid=True)
-
-    return HybridScaler(luma_scaler, chroma_scaler).scale(
-        clip, width, height, shift=shift
-    )
-
-
-@dataclass
-class HybridScaler(GenericScaler):
-    luma_scaler: ScalerLike
-    chroma_scaler: ScalerLike
-
-    def __post_init__(self) -> None:
-        super().__post_init__()
-
-        self._luma = Scaler.ensure_obj(self.luma_scaler)
-        self._chroma = Scaler.ensure_obj(self.chroma_scaler)
-
-    @Scaler.cached_property
-    def kernel_radius(self) -> int:
-        return self._luma.kernel_radius
-
-    def scale(  # type:ignore
-        self,
-        clip: vs.VideoNode,
-        width: int,
-        height: int,
-        shift: tuple[float, float] = (0, 0),
-        **kwargs: Any,
-    ) -> vs.VideoNode:
-        assert check_variable_format(clip, self.__class__)
-
-        luma = self._luma.scale(clip, width, height, shift, **kwargs)
-
-        if clip.format.num_planes == 1:
-            return luma
-
-        chroma = self._chroma.scale(clip, width, height, shift, **kwargs)
-
-        return join(luma, chroma)
-
-
-def _get_scaler(scaler: ScalerLike, **kwargs: Any) -> Scaler:
-    scaler_cls = Scaler.from_param(scaler, _get_scaler)
-
-    args = getfullargspec(scaler_cls).args
-
-    clean_kwargs = {key: value for key, value in kwargs.items() if key in args}
-
-    return scaler_cls(**clean_kwargs)