opensportslib 0.3.0.dev15__tar.gz → 0.3.0.dev17__tar.gz

{opensportslib-0.3.0.dev15/opensportslib.egg-info → opensportslib-0.3.0.dev17}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: opensportslib
-Version: 0.3.0.dev15
+Version: 0.3.0.dev17
 Summary: OpenSportsLib is the professional library, designed for advanced video understanding in sports. It provides state-of-the-art tools for action recognition, spotting, retrieval, and captioning, making it ideal for researchers, analysts, and developers working with sports video data.
 Author: Jeet Vora
 Requires-Python: >=3.12

opensportslib-0.3.0.dev17/opensportslib/setup/setup.py

@@ -0,0 +1,291 @@
+import platform
+import subprocess
+import sys
+
+
+CUDA_WHEEL_VERSIONS = {
+    "cu126": (12, 6),
+    "cu128": (12, 8),
+    "cu130": (13, 0),
+}
+MIN_SUPPORTED_COMPUTE_CAPABILITY = (5, 0)
+LEGACY_GPU_MAX_COMPUTE_CAPABILITY = (7, 4)
+LEGACY_GPU_CUDA_WHEEL = "cu126"
+LEGACY_GPU_CUDA_WHEEL_MAX_COMPUTE_CAPABILITY = (9, 0)
+CUDA13_REQUIRED_MIN_COMPUTE_CAPABILITY = (10, 0)
+LEGACY_GPU_TORCH_PACKAGES = (
+    "torch==2.10.0",
+    "torchvision==0.25.0",
+    "torchaudio==2.10.0",
+)
+
+XVARS_DEPENDENCY_PINS = {
+    "transformers": "4.38.2",
+    "peft": "0.9.0",
+    "tokenizers": "0.15.2",
+    "accelerate": "0.27.2",
+    "trl": "0.10.1",
+}
+
+QWEN_DEPENDENCY_PINS = {
+    "transformers": "5.13.0",
+    "peft": "0.19.0",
+    "tokenizers": "0.22.1",
+    "accelerate": "1.14.0",
+    "trl": "1.7.1",
+}
+
+def get_cuda_version():
+    try:
+        output = subprocess.check_output(["nvidia-smi"]).decode()
+
+        for line in output.split("\n"):
+            if "CUDA Version" in line:
+                ver  = line.split("CUDA Version:")[1].strip().split()[0]
+                print(f"CUDA Version found : {ver}")
+                cuda_tag = f"cu{ver.replace('.', '')}"
+                return ver, cuda_tag
+    except Exception:
+        return None, None
+    return None, None
+
+
+def get_gpu_compute_capabilities():
+    try:
+        output = subprocess.check_output(
+            ["nvidia-smi", "--query-gpu=compute_cap", "--format=csv,noheader"],
+            text=True,
+        )
+    except Exception:
+        return []
+
+    capabilities = []
+    for line in output.splitlines():
+        value = line.strip()
+        if not value:
+            continue
+        try:
+            major, minor = value.split(".", 1)
+            capabilities.append((int(major), int(minor)))
+        except ValueError:
+            print(f"Ignoring unrecognized GPU compute capability: {value}")
+    return capabilities
+
+
+def select_cuda_wheel(cuda_version, compute_capabilities):
+    if not cuda_version:
+        return "cpu"
+
+    try:
+        driver_version = tuple(int(part) for part in cuda_version.split(".", 1))
+    except ValueError as exc:
+        raise RuntimeError(f"Unable to parse CUDA version reported by nvidia-smi: {cuda_version}") from exc
+
+    if any(capability < MIN_SUPPORTED_COMPUTE_CAPABILITY for capability in compute_capabilities):
+        raise RuntimeError(
+            "OpenSportsLib PyTorch wheels require GPUs with compute capability "
+            f"{MIN_SUPPORTED_COMPUTE_CAPABILITY[0]}.{MIN_SUPPORTED_COMPUTE_CAPABILITY[1]} or newer. "
+            f"Detected: {compute_capabilities}."
+        )
+
+    compatible_tags = [
+        tag for tag, wheel_version in CUDA_WHEEL_VERSIONS.items() if wheel_version <= driver_version
+    ]
+    has_legacy_gpu = compute_capabilities and any(
+        capability <= LEGACY_GPU_MAX_COMPUTE_CAPABILITY for capability in compute_capabilities
+    )
+    if has_legacy_gpu:
+        if platform.machine().lower() in {"aarch64", "arm64"}:
+            raise RuntimeError(
+                "Official Linux ARM64 CUDA 12.6 PyTorch wheels support Ampere and newer GPUs only. "
+                "Use an x86_64 host or build PyTorch from source for the visible legacy GPU architecture."
+            )
+        if any(
+            capability > LEGACY_GPU_CUDA_WHEEL_MAX_COMPUTE_CAPABILITY
+            for capability in compute_capabilities
+        ):
+            raise RuntimeError(
+                "Visible GPUs require incompatible prebuilt PyTorch wheels. Set CUDA_VISIBLE_DEVICES to "
+                "either the legacy GPUs or the newer GPUs, then run setup again."
+            )
+        compatible_tags = [tag for tag in compatible_tags if tag == LEGACY_GPU_CUDA_WHEEL]
+
+    if any(
+        capability >= CUDA13_REQUIRED_MIN_COMPUTE_CAPABILITY
+        for capability in compute_capabilities
+    ):
+        compatible_tags = [tag for tag in compatible_tags if tag == "cu130"]
+
+    if not compatible_tags:
+        if any(
+            capability >= CUDA13_REQUIRED_MIN_COMPUTE_CAPABILITY
+            for capability in compute_capabilities
+        ):
+            raise RuntimeError(
+                "The visible GPU architecture requires the CUDA 13.0 PyTorch wheel. "
+                "Update the NVIDIA driver so nvidia-smi reports CUDA 13.0 or newer, "
+                "then run setup again."
+            )
+        raise RuntimeError(
+            f"CUDA {cuda_version} is too old for the supported PyTorch wheels: "
+            f"{', '.join(CUDA_WHEEL_VERSIONS)}."
+        )
+
+    selected = max(compatible_tags, key=lambda tag: CUDA_WHEEL_VERSIONS[tag])
+    print(
+        "Selected PyTorch wheel "
+        f"{selected} for CUDA {cuda_version} and GPU compute capabilities {compute_capabilities or 'unknown'}"
+    )
+    return selected
+
+
+def select_torch_packages(compute_capabilities):
+    if compute_capabilities and any(
+        capability <= LEGACY_GPU_MAX_COMPUTE_CAPABILITY for capability in compute_capabilities
+    ):
+        return LEGACY_GPU_TORCH_PACKAGES
+    return ("torch", "torchvision", "torchaudio")
+
+
+CUDA_VERSION, _DETECTED_CUDA_TAG = get_cuda_version()
+GPU_COMPUTE_CAPABILITIES = get_gpu_compute_capabilities()
+CUDA_TAG = select_cuda_wheel(CUDA_VERSION, GPU_COMPUTE_CAPABILITIES)
+
+
+def install_xvars_dependencies(DEPENDENCY_PINS):
+    python = sys.executable
+    packages = list(DEPENDENCY_PINS)
+    pinned_packages = [f"{name}=={version}" for name, version in DEPENDENCY_PINS.items()]
+
+    print(f"\nInstalling {list(DEPENDENCY_PINS.keys())} dependency overrides...\n")
+    print("This overrides the default Hugging Face dependency set with XVars-compatible versions.")
+    subprocess.call([python, "-m", "pip", "uninstall", "-y", *packages])
+    subprocess.check_call([python, "-m", "pip", "install", *pinned_packages])
+    print("Dependencies installed successfully.")
+
+def install_torch():
+    python = sys.executable
+    subprocess.call([python, "-m", "pip", "uninstall", "-y", "torch", "torchvision", "torchaudio"])
+    packages = select_torch_packages(GPU_COMPUTE_CAPABILITIES)
+
+    subprocess.check_call([
+        python, "-m", "pip", "install",
+        *packages,
+        "--index-url",
+        f"https://download.pytorch.org/whl/{CUDA_TAG}",
+    ])
+    print(f"\nSuccess with {CUDA_TAG}: {', '.join(packages)}")
+    return CUDA_TAG
+
+def install_dali():
+
+    python = sys.executable
+
+    print("\nInstalling dali extras...\n")
+
+    # DALI (only if GPU)
+    if CUDA_VERSION:
+        
+        if CUDA_TAG == "cu130":
+            subprocess.check_call([
+                python, "-m", "pip", "install",
+                "nvidia-dali-cuda130"
+            ])
+
+            # CuPy (CUDA-aware but auto-resolves internally)
+            subprocess.check_call([
+                python, "-m", "pip", "install",
+                "cupy-cuda130"
+            ])
+        else:
+            subprocess.check_call([
+                python, "-m", "pip", "install",
+                "nvidia-dali-cuda120"
+            ])
+
+            # CuPy (CUDA-aware but auto-resolves internally)
+            subprocess.check_call([
+                python, "-m", "pip", "install",
+                "cupy-cuda12x"
+            ])
+
+def install_pyg():
+    import torch
+    from packaging import version
+
+    python = sys.executable
+    torch_version = "2.10.0" if version.parse(torch.__version__.split("+")[0]) > version.parse("2.10.0") else torch.__version__.split("+")[0]
+    cuda_tag = CUDA_TAG
+    print("\nInstalling Py-Geometric ecosystem...\n")
+    if cuda_tag == "cpu":
+        url =  f"https://data.pyg.org/whl/torch-{torch_version}+cpu.html"
+    else:
+        url = f"https://data.pyg.org/whl/torch-{torch_version}+{cuda_tag}.html"
+
+    subprocess.check_call([
+        python, "-m", "pip", "install",
+        "torch-geometric", "-f", url
+    ])
+    subprocess.check_call([
+        python, "-m", "pip", "install",
+        "torch-scatter", "-f", url
+    ])
+    subprocess.check_call([
+        python, "-m", "pip", "install",
+        "torch-sparse", "-f", url
+    ])
+    subprocess.check_call([
+        python, "-m", "pip", "install",
+        "torch-cluster", "-f", url
+    ])
+    subprocess.check_call([
+        python, "-m", "pip", "install",
+        "torch-spline-conv", "-f", url
+    ])
+
+def install_extras(dali=False, pyg=False):
+    if dali:
+        install_dali()
+        print("NVIDIA DALI installed successfully.")
+    if pyg:
+        install_pyg()
+        print("PyTorch Geometric installed successfully.")
+
+
+def verify():
+    import torch
+
+    print("\n Verifying installation...\n")
+    print("Torch:", torch.__version__)
+
+    if torch.cuda.is_available():
+        print("CUDA available")
+        print("GPU:", torch.cuda.get_device_name(0))
+    else:
+        print("Running on CPU")
+
+def setup(dali=False, pyg=False, vqa_xvars=False, vqa_qwen=False):
+    install_torch()
+    install_extras(dali=dali, pyg=pyg)
+    if vqa_xvars:
+        install_xvars_dependencies(XVARS_DEPENDENCY_PINS)
+    if vqa_qwen:
+        install_xvars_dependencies(QWEN_DEPENDENCY_PINS)
+    verify()
+
+
+# ----------------------------
+# CLI entry
+# ----------------------------
+if __name__ == "__main__":
+    import argparse
+
+    parser = argparse.ArgumentParser()
+    parser.add_argument("--dali", action="store_true")
+    parser.add_argument("--pyg", action="store_true")
+    parser.add_argument("--vqa_xvars", action="store_true")
+    parser.add_argument("--vqa_qwen", action="store_true")
+
+    args = parser.parse_args()
+
+    setup(dali=args.dali, pyg=args.pyg, vqa_xvars=args.vqa_xvars, vqa_qwen=args.vqa_qwen)

{opensportslib-0.3.0.dev15 → opensportslib-0.3.0.dev17/opensportslib.egg-info}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: opensportslib
-Version: 0.3.0.dev15
+Version: 0.3.0.dev17
 Summary: OpenSportsLib is the professional library, designed for advanced video understanding in sports. It provides state-of-the-art tools for action recognition, spotting, retrieval, and captioning, making it ideal for researchers, analysts, and developers working with sports video data.
 Author: Jeet Vora
 Requires-Python: >=3.12

{opensportslib-0.3.0.dev15 → opensportslib-0.3.0.dev17}/pyproject.toml

@@ -4,7 +4,7 @@ build-backend = "setuptools.build_meta"
 
 [project]
 name = "opensportslib"
-version = "0.3.0.dev15"
+version = "0.3.0.dev17"
 description = "OpenSportsLib is the professional library, designed for advanced video understanding in sports. It provides state-of-the-art tools for action recognition, spotting, retrieval, and captioning, making it ideal for researchers, analysts, and developers working with sports video data."
 readme = "README.md"
 requires-python = ">=3.12"

{opensportslib-0.3.0.dev15 → opensportslib-0.3.0.dev17}/tests/test_setup_cli.py

@@ -1,9 +1,40 @@
 from __future__ import annotations
 
+import pytest
+
 from opensportslib import cli
 from opensportslib.setup import setup as setup_lib
 
 
+def test_select_cuda_wheel_uses_cu126_for_pre_sm75_gpu_with_cuda_13():
+    assert setup_lib.select_cuda_wheel("13.0", [(7, 0)]) == "cu126"
+
+
+def test_select_cuda_wheel_uses_cu126_for_pascal_with_cuda_13():
+    assert setup_lib.select_cuda_wheel("13.0", [(6, 0)]) == "cu126"
+
+
+def test_select_torch_packages_pins_pre_sm75_gpu_compatibility_stack():
+    assert setup_lib.select_torch_packages([(7, 0)]) == (
+        "torch==2.10.0",
+        "torchvision==0.25.0",
+        "torchaudio==2.10.0",
+    )
+
+
+def test_select_cuda_wheel_uses_cu130_for_dgx_spark():
+    assert setup_lib.select_cuda_wheel("13.0", [(12, 1)]) == "cu130"
+
+
+def test_select_cuda_wheel_rejects_new_architecture_without_cuda_13_driver():
+    with pytest.raises(RuntimeError, match="CUDA 13.0"):
+        setup_lib.select_cuda_wheel("12.8", [(10, 0)])
+
+
+def test_select_cuda_wheel_uses_highest_driver_compatible_wheel():
+    assert setup_lib.select_cuda_wheel("12.8", [(8, 0)]) == "cu128"
+
+
 def test_cli_setup_forwards_xvars_flag(monkeypatch):
     captured: dict[str, object] = {}
 

opensportslib-0.3.0.dev15/opensportslib/setup/setup.py

@@ -1,211 +0,0 @@
-import subprocess
-import sys
-
-
-CUDA_SUPPORT = [
-    "cu126",
-    "cu128",
-    "cu130",
-    "cpu"
-]
-
-XVARS_DEPENDENCY_PINS = {
-    "transformers": "4.38.2",
-    "peft": "0.9.0",
-    "tokenizers": "0.15.2",
-    "accelerate": "0.27.2",
-    "trl": "0.10.1",
-}
-
-QWEN_DEPENDENCY_PINS = {
-    "transformers": "5.13.0",
-    "peft": "0.19.0",
-    "tokenizers": "0.22.1",
-    "accelerate": "1.14.0",
-    "trl": "1.7.1",
-}
-
-def get_cuda_version():
-    try:
-        output = subprocess.check_output(["nvidia-smi"]).decode()
-
-        for line in output.split("\n"):
-            if "CUDA Version" in line:
-                ver  = line.split("CUDA Version:")[1].strip().split()[0]
-                print(f"CUDA Version found : {ver}")
-                cuda_tag = f"cu{ver.replace('.', '')}"
-                return ver, cuda_tag
-    except Exception:
-        return None, None
-
-
-CUDA_VERSION, CUDA_TAG = get_cuda_version()
-
-def get_cpu_tag():
-    if not CUDA_VERSION:
-        return "cpu"
-
-
-def install_xvars_dependencies(DEPENDENCY_PINS):
-    python = sys.executable
-    packages = list(DEPENDENCY_PINS)
-    pinned_packages = [f"{name}=={version}" for name, version in DEPENDENCY_PINS.items()]
-
-    print(f"\nInstalling {list(DEPENDENCY_PINS.keys())} dependency overrides...\n")
-    print("This overrides the default Hugging Face dependency set with XVars-compatible versions.")
-    subprocess.call([python, "-m", "pip", "uninstall", "-y", *packages])
-    subprocess.check_call([python, "-m", "pip", "install", *pinned_packages])
-    print("Dependencies installed successfully.")
-
-def install_torch():
-    python = sys.executable
-    subprocess.call([python, "-m", "pip", "uninstall", "-y", "torch", "torchvision"])
-    
-    if CUDA_TAG == "cu130":
-        cuda = "cu130"
-        subprocess.check_call([
-            python, "-m", "pip", "install",
-            "torch", "torchvision", "torchaudio",
-            "--index-url",
-            f"https://download.pytorch.org/whl/{cuda}"
-        ])
-        print(f"\nSuccess with {cuda}")
-        return cuda
-    for cuda in CUDA_SUPPORT:
-
-        print(f"\n Trying installation: {cuda}\n")
-        try:
-            if get_cpu_tag() == "cpu":
-                subprocess.check_call([
-                    python, "-m", "pip", "install",
-                    "torch", "torchvision", "torchaudio"
-                ])
-            else:
-
-                subprocess.check_call([
-                    python, "-m", "pip", "install",
-                    "torch", "torchvision", "torchaudio",
-                    "--index-url",
-                    f"https://download.pytorch.org/whl/{cuda}"
-                ])
-            print(f"\nSuccess with {cuda}")
-            return cuda
-
-        except Exception as e:
-            print(f"Failed with {cuda}: {e}")
-
-    raise RuntimeError("All CUDA installs failed")
-
-def install_dali():
-
-    python = sys.executable
-
-    print("\nInstalling dali extras...\n")
-
-    # DALI (only if GPU)
-    if CUDA_VERSION:
-        
-        if CUDA_TAG == "cu130":
-            subprocess.check_call([
-                python, "-m", "pip", "install",
-                "nvidia-dali-cuda130"
-            ])
-
-            # CuPy (CUDA-aware but auto-resolves internally)
-            subprocess.check_call([
-                python, "-m", "pip", "install",
-                "cupy-cuda130"
-            ])
-        else:
-            subprocess.check_call([
-                python, "-m", "pip", "install",
-                "nvidia-dali-cuda120"
-            ])
-
-            # CuPy (CUDA-aware but auto-resolves internally)
-            subprocess.check_call([
-                python, "-m", "pip", "install",
-                "cupy-cuda12x"
-            ])
-
-def install_pyg():
-    import torch
-    from packaging import version
-
-    python = sys.executable
-    torch_version = "2.10.0" if version.parse(torch.__version__.split("+")[0]) > version.parse("2.10.0") else torch.__version__.split("+")[0]
-    cuda_tag = next((f"cu{CUDA_VERSION.replace('.', '')}" for _ in [0] if CUDA_VERSION), CUDA_SUPPORT[0])
-    cuda_tag = cuda_tag if cuda_tag in CUDA_SUPPORT else CUDA_SUPPORT[0]
-    print("\nInstalling Py-Geometric ecosystem...\n")
-    if cuda_tag == "cpu":
-        url =  f"https://data.pyg.org/whl/torch-{torch_version}+cpu.html"
-    else:
-        url = f"https://data.pyg.org/whl/torch-{torch_version}+{cuda_tag}.html"
-
-    subprocess.check_call([
-        python, "-m", "pip", "install",
-        "torch-geometric", "-f", url
-    ])
-    subprocess.check_call([
-        python, "-m", "pip", "install",
-        "torch-scatter", "-f", url
-    ])
-    subprocess.check_call([
-        python, "-m", "pip", "install",
-        "torch-sparse", "-f", url
-    ])
-    subprocess.check_call([
-        python, "-m", "pip", "install",
-        "torch-cluster", "-f", url
-    ])
-    subprocess.check_call([
-        python, "-m", "pip", "install",
-        "torch-spline-conv", "-f", url
-    ])
-
-def install_extras(dali=False, pyg=False):
-    if dali:
-        install_dali()
-        print("NVIDIA DALI installed successfully.")
-    if pyg:
-        install_pyg()
-        print("PyTorch Geometric installed successfully.")
-
-
-def verify():
-    import torch
-
-    print("\n Verifying installation...\n")
-    print("Torch:", torch.__version__)
-
-    if torch.cuda.is_available():
-        print("CUDA available")
-        print("GPU:", torch.cuda.get_device_name(0))
-    else:
-        print("Running on CPU")
-
-def setup(dali=False, pyg=False, vqa_xvars=False, vqa_qwen=False):
-    install_torch()
-    install_extras(dali=dali, pyg=pyg)
-    if vqa_xvars:
-        install_xvars_dependencies(XVARS_DEPENDENCY_PINS)
-    if vqa_qwen:
-        install_xvars_dependencies(QWEN_DEPENDENCY_PINS)
-    verify()
-
-
-# ----------------------------
-# CLI entry
-# ----------------------------
-if __name__ == "__main__":
-    import argparse
-
-    parser = argparse.ArgumentParser()
-    parser.add_argument("--dali", action="store_true")
-    parser.add_argument("--pyg", action="store_true")
-    parser.add_argument("--vqa_xvars", action="store_true")
-    parser.add_argument("--vqa_qwen", action="store_true")
-
-    args = parser.parse_args()
-
-    setup(dali=args.dali, pyg=args.pyg, vqa_xvars=args.vqa_xvars, vqa_qwen=args.vqa_qwen)