Ethan/dev

.github/workflows/build-wheels-cpu-macos.yaml

@@ -36,6 +36,7 @@ jobs:
         env:
           # disable repair
           CIBW_REPAIR_WHEEL_COMMAND: ""
+          CIBW_BUILD_FRONTEND: "build"
           CIBW_ENVIRONMENT: CMAKE_ARGS="-DCMAKE_OSX_ARCHITECTURES=arm64;x86_64 "
           CIBW_BUILD: "cp310-* cp311-* cp312-*"
           MACOSX_DEPLOYMENT_TARGET: ${{ matrix.os }}

.github/workflows/build-wheels-cpu.yaml

@@ -35,6 +35,7 @@ jobs:
         env:
           # disable repair
           CIBW_REPAIR_WHEEL_COMMAND: ""
+          CIBW_BUILD_FRONTEND: "build"
           CIBW_SKIP: "*musllinux*"
           CIBW_BUILD: "cp310-* cp311-* cp312-*"
           CMAKE_BUILD_PARALLEL_LEVEL: 4
@@ -77,6 +78,7 @@ jobs:
         env:
           # disable repair
           CIBW_REPAIR_WHEEL_COMMAND: ""
+          CIBW_BUILD_FRONTEND: "build"
           CIBW_BUILD: "cp310-* cp311-* cp312-*"
           CMAKE_BUILD_PARALLEL_LEVEL: 16
         with:
@@ -104,8 +106,9 @@ jobs:
       - name: Build wheels
         uses: pypa/[email protected]
         env:
-          CIBW_SKIP: "*musllinux* pp*"
           CIBW_REPAIR_WHEEL_COMMAND: ""
+          CIBW_BUILD_FRONTEND: "build"
+          CIBW_SKIP: "*musllinux* pp*"
           CIBW_ARCHS: "aarch64"
           CIBW_BUILD: "cp310-* cp311-* cp312-*"
           CMAKE_BUILD_PARALLEL_LEVEL: $(nproc)

.github/workflows/build-wheels-metal.yaml

@@ -35,6 +35,7 @@ jobs:
         env:
           # disable repair
           CIBW_REPAIR_WHEEL_COMMAND: ""
+          CIBW_BUILD_FRONTEND: "build"
           CIBW_ARCHS: "arm64"
           CIBW_BUILD: "cp310-* cp311-* cp312-*"
           CIBW_ENVIRONMENT: CMAKE_ARGS="-DCMAKE_OSX_ARCHITECTURES=arm64 -DCMAKE_APPLE_SILICON_PROCESSOR=arm64 -DGGML_METAL=ON -DSD_METAL=ON"

.github/workflows/ci.yaml

@@ -25,7 +25,7 @@ jobs:
         fetch-depth: 0  # This fetches all history so you can access any version of the submodules
 
     - name: Set up Python
-      uses: actions/setup-python@v4
+      uses: actions/setup-python@v5
       with:
         python-version: ${{ matrix.python-version }}
 

nexa/__init__.py

@@ -1 +1 @@
-__version__ = "0.0.1"
+__version__ = "0.0.5"

nexa/cli/entry.py

@@ -1,6 +1,8 @@
 import argparse
 import uvicorn
 
+from nexa import __version__
+
 def run_onnx_inference(args):
     kwargs = {k: v for k, v in vars(args).items() if v is not None}
     model_path = kwargs.pop("model_path")
@@ -73,6 +75,15 @@ def main():
     parser = argparse.ArgumentParser(
         description="Nexa CLI tool for handling various model operations."
     )
+
+    parser.add_argument(
+        "-V",
+        "--version",
+        action="version",
+        version=__version__,
+        help="Show the version of the Nexa SDK.",
+    )
+
     subparsers = parser.add_subparsers(dest="command", help="sub-command help")
 
     # ONNX subparsers

nexa/gguf/lib_utils.py

@@ -8,11 +8,6 @@
 from pathlib import Path
 from typing import List
 
-from nexa.utils import (
-    is_nexa_cuda_installed,
-    is_nexa_metal_installed,
-)
-
 
 def is_gpu_available():
     current_dir = os.path.dirname(os.path.abspath(__file__))
@@ -72,17 +67,19 @@ def load_library(lib_base_name: str):
 
 
 def _add_windows_dll_directories(base_path: Path) -> None:
-    try_add_cuda_lib_path()
     os.add_dll_directory(str(base_path))
     os.environ["PATH"] = str(base_path) + os.pathsep + os.environ["PATH"]
 
-    if sys.version_info >= (3, 8):
-        if "CUDA_PATH" in os.environ:
-            os.add_dll_directory(os.path.join(os.environ["CUDA_PATH"], "bin"))
-            os.add_dll_directory(os.path.join(os.environ["CUDA_PATH"], "lib"))
-        if "HIP_PATH" in os.environ:
-            os.add_dll_directory(os.path.join(os.environ["HIP_PATH"], "bin"))
-            os.add_dll_directory(os.path.join(os.environ["HIP_PATH"], "lib"))
+    if is_gpu_available():
+        try_add_cuda_lib_path()
+
+        if sys.version_info >= (3, 8):
+            if "CUDA_PATH" in os.environ:
+                os.add_dll_directory(os.path.join(os.environ["CUDA_PATH"], "bin"))
+                os.add_dll_directory(os.path.join(os.environ["CUDA_PATH"], "lib"))
+            if "HIP_PATH" in os.environ:
+                os.add_dll_directory(os.path.join(os.environ["HIP_PATH"], "bin"))
+                os.add_dll_directory(os.path.join(os.environ["HIP_PATH"], "lib"))
 
 
 def try_add_cuda_lib_path():

pyproject.toml

@@ -4,14 +4,14 @@ build-backend = "scikit_build_core.build"
 
 [project]
 name = "nexaai"
-version = "0.0.5.1"
+dynamic = ["version"]
 description = "Nexa AI SDK"
 readme = "README.md"
 license = { text = "MIT" }
 authors = [{ name = "Nexa AI", email = "[email protected]" }]
 dependencies = [
     "faster_whisper",
-    "typing-extensions>=4.5.0",  # For ggml
+    "typing-extensions>=4.5.0", # For ggml
     "numpy>=1.20.0",
     "diskcache>=5.6.1",
     "jinja2>=2.11.3",
@@ -21,9 +21,9 @@ dependencies = [
     "pydantic",
     "pillow",
     "prompt_toolkit",
-    "tqdm",  # Shared dependencies
+    "tqdm",                     # Shared dependencies
     "tabulate",
-    "streamlit"
+    "streamlit",
 ]
 classifiers = [
     "Programming Language :: Python :: 3",
@@ -36,8 +36,8 @@ classifiers = [
 [project.optional-dependencies]
 onnx = [
     "librosa",
-    "optimum[onnxruntime]",  # for CPU version
-    "diffusers",  # required for image generation
+    "optimum[onnxruntime]", # for CPU version
+    "diffusers",            # required for image generation
     "optuna",
     "pydantic",
     "PyYAML",
@@ -46,7 +46,7 @@ onnx = [
     "soundfile",
     "streamlit_audiorec",
     "transformers",
-    "ttstokenizer"
+    "ttstokenizer",
 ]
 
 [project.urls]
@@ -71,15 +71,30 @@ wheel.packages = [
     "nexa.gguf.server",
     "nexa.onnx",
     "nexa.onnx.streamlit",
-    "nexa.onnx.server"
+    "nexa.onnx.server",
+]
+sdist.include = [
+    "CMakeLists.txt",
+    "dependency/llama.cpp/*",
+    "dependency/stable-diffusion.cpp/*",
+]
+sdist.exclude = [
+    ".github",
+    "build",
+    "dist",
+    "nexa.egg-info",
+    "dependency/llama.cpp/build",
+    "dependency/stable-diffusion.cpp/build",
 ]
-sdist.include = ["CMakeLists.txt", "dependency/llama.cpp/*", "dependency/stable-diffusion.cpp/*"]
-sdist.exclude = [".github", "build", "dist", "nexa.egg-info", "dependency/llama.cpp/build", "dependency/stable-diffusion.cpp/build"]
 build.verbose = true
 cmake.build-type = "Release"
 cmake.version = ">=3.16"
 # cmake.args = ["-DCMAKE_CXX_FLAGS=-fopenmp"] # for macOS with Intel Silicon
 cmake.args = ["-DCMAKE_BUILD_PARALLEL_LEVEL=16"]
 
+[tool.scikit-build.metadata.version]
+provider = "scikit_build_core.metadata.regex"
+input = "nexa/__init__.py"
+
 [tool.pytest.ini_options]
-testpaths = ["tests"]
+testpaths = ["tests"]

scripts/releases-to-pep-503.sh

@@ -79,9 +79,9 @@ if [ "$version" == "cpu" ]; then
     echo "    <h1>NEXAAI SDK Python Wheels</h1>" >> "$root_dir/index.html"
     echo "    <a href=\"cpu/\">CPU</a><br>" >> "$root_dir/index.html"
     echo "    <a href=\"metal/\">Metal</a><br>" >> "$root_dir/index.html"
-    echo "    <a href=\"cu121/\">CUDA 12.1</a><br>" >> "$root_dir/index.html"
-    echo "    <a href=\"cu122/\">CUDA 12.2</a><br>" >> "$root_dir/index.html"
-    echo "    <a href=\"cu123/\">CUDA 12.3</a><br>" >> "$root_dir/index.html"
+    # echo "    <a href=\"cu121/\">CUDA 12.1</a><br>" >> "$root_dir/index.html"
+    # echo "    <a href=\"cu122/\">CUDA 12.2</a><br>" >> "$root_dir/index.html"
+    # echo "    <a href=\"cu123/\">CUDA 12.3</a><br>" >> "$root_dir/index.html"
     echo "    <a href=\"cu124/\">CUDA 12.4</a><br>" >> "$root_dir/index.html"
     echo "  </body>" >> "$root_dir/index.html"
     echo "</html>" >> "$root_dir/index.html"

tests/utils.py

@@ -3,27 +3,27 @@
 
 def download_model(url, output_dir):
     """
-    Download a file from a given URL using wget, if it doesn't already exist.
-    
+    Download a file from a given URL using curl, if it doesn't already exist.
+
     Args:
     - url: str, the URL of the file to download.
     - output_dir: str, the directory where the file should be saved.
-    
+
     Returns:
     - str: The path to the downloaded file.
     """
     file_name = url.split("/")[-1]
     output_path = os.path.join(output_dir, file_name)
-    
+
     if os.path.exists(output_path):
         print(f"File {file_name} already exists in {output_dir}. Skipping download.")
         return output_path
-    
+
     try:
-        subprocess.run(["wget", url, "-O", output_path], check=True)
+        subprocess.run(["curl", url, "--output", output_path], check=True)
         print(f"Downloaded {file_name} to {output_dir}")
     except subprocess.CalledProcessError as e:
         print(f"Failed to download {file_name}: {e}")
         raise
-    
+
     return output_path