Merge branch 'main_copy25.09' into 'main'

sync with internal repo1 (commit 62f139f5a)

See merge request cuda-hpc-libraries/cuquantum-sdk/cuquantum-public!39

Author: daniellowell

python/README.md

@@ -13,12 +13,12 @@ For instructions on installing *cuQuantum Python*, refer to our
 
 The build-time dependencies of the cuQuantum Python package include:
 
-* CUDA Toolkit 11.x or 12.x
+* CUDA Toolkit 12.x or 13.x
 * Python >=3.11, <3.14
-* Cython >=0.29.22
+* Cython >=3.0.4,!=3.1.0,!=3.1.1
 * pip 21.3.1+
 * [packaging](https://packaging.pypa.io/en/latest/)
-* setuptools 61.0.0+
+* setuptools 77.0.3+
 * wheel 0.34.0+
 
 > **Note:** Starting with cuQuantum Python v25.06, cuQuantum C libraries including cuDensityMat, cuStateVec and cuTensorNet are no longer build-time dependencies. However, they are still required at runtime.
@@ -29,37 +29,36 @@ To compile and install cuQuantum Python from source, please follow the steps bel
 
 1. Clone the [NVIDIA/cuQuantum](https://github.com/NVIDIA/cuQuantum) repository: `git clone https://github.com/NVIDIA/cuQuantum`
 2. Set `CUDA_PATH` to point to your CUDA installation
-3. [optional] Set `CUQUANTUM_ROOT` to point to your cuQuantum installation
-4. [optional] Set `CUTENSOR_ROOT` to point to your cuTENSOR installation
-5. [optional] Make sure cuQuantum and cuTENSOR are visible in your `LD_LIBRARY_PATH`
-6. Switch to the directory containing the Python implementation: `cd cuQuantum/python`
-7. Build and install:
-   - Run `pip install .` if you skip Step 3-5 above
+3. [optional] Make sure cuQuantum and cuTENSOR are visible in your `LD_LIBRARY_PATH`
+4. Switch to the directory containing the Python implementation: `cd cuQuantum/python`
+5. Build and install:
+   - Run `pip install .` if you skip Step 3 above
    - Run `pip install -v --no-deps --no-build-isolation .` otherwise (advanced)
 
 Notes:
-- For Step 7, if you are building from source for testing/developing purposes you'd likely want to insert a `-e` flag before the last period (so `pip ... .` becomes `pip ... -e .`):
+- For Step 5, if you are building from source for testing/developing purposes you'd likely want to insert a `-e` flag before the last period (so `pip ... .` becomes `pip ... -e .`):
   * `-e`: use the "editable" (in-place) mode
   * `-v`: enable more verbose output
   * `--no-deps`: avoid installing the *run-time* dependencies
   * `--no-build-isolation`: reuse the current Python environment instead of creating a new one for building the package (this avoids installing any *build-time* dependencies)
-- As an alternative to setting `CUQUANTUM_ROOT`, `CUSTATEVEC_ROOT`, `CUTENSORNET_ROOT` and `CUDENSITYMAT_ROOT` can be set to point to the cuStateVec, cuTensorNet and cuDensityMat libraries, respectively. The latter three environment variables take precedence if defined.
-- Please ensure that you use consistent binaries and packages for either CUDA 11 or 12. Mixing-and-matching will result in undefined behavior.
+- Please ensure that you use consistent binaries and packages for either CUDA 12 or 13. Mixing-and-matching will result in undefined behavior.
 
 ## Running
 
 ### Requirements
 
 Runtime dependencies of the cuQuantum Python package include:
 
-* An NVIDIA GPU with compute capability 7.0+
-* Driver: Linux (450.80.02+ for CUDA 11, 525.60.13+ for CUDA 12)
-* CUDA Toolkit 11.x or 12.x
-* cuStateVec 1.9.0+
-* cuTensorNet 2.8.0+
-* cuDensityMat >=0.2.0, <0.3.0
+* An NVIDIA GPU with compute capability 7.5+
+* Driver: Linux (525.60.13+ for CUDA 12, 580.65.06+ for CUDA 13)
+* CUDA Toolkit 12.x or 13.x
+* cuStateVec 1.10.0+
+* cuTensorNet 2.9.0+
+* cuDensityMat >=0.3.0, <0.4.0
 * Python >=3.11, <3.14
 * NumPy v1.21+
+* nvmath-python ==0.6.0
+* cuda-bindings >=12.9.2, <13.0.0 for CUDA 12 or cuda-bindings >=13.0.1, <14.0.0
 * CuPy v13.0.0+ (see [installation guide](https://docs.cupy.dev/en/stable/install.html))
 * PyTorch v1.10+ (optional, see [installation guide](https://pytorch.org/get-started/locally/))
 * Qiskit v1.4.2+ (optional, see [installation guide](https://qiskit.org/documentation/getting_started.html))
@@ -76,7 +75,7 @@ to your `LD_LIBRARY_PATH` environment variable, and that a compatible CuPy is in
 
 Known issues:
 - If a system has multiple copies of cuTENSOR, one of which is installed in a default system path, the Python runtime could pick it up despite cuQuantum Python is linked to another copy installed elsewhere, potentially causing a version-mismatch error. The proper fix is to remove cuTENSOR from the system paths to ensure the visibility of the proper copy. **DO NOT ATTEMPT** to use `LD_PRELOAD` to overwrite it --- it could cause hard to debug behaviors!
-- Please ensure that you use consistent binaries and packages for either CUDA 11 or 12. Mixing-and-matching will result in undefined behavior.
+- Please ensure that you use consistent binaries and packages for either CUDA 12 or 13. Mixing-and-matching will result in undefined behavior.
 
 ### Samples
 

python/builder/utils.py

@@ -49,18 +49,21 @@ def check_cuda_version():
     except:
         raise
     else:
-        # 11020 -> "11.2"
+        # 12020 -> "12.2"
         return str(ver // 1000) + '.' + str((ver % 100) // 10)
 
 
-# We support CUDA 11/12 starting 23.03
+# We support CUDA 12/13 starting 25.09
 cuda_ver = check_cuda_version()
-if cuda_ver == '11.0':
-    cuda_major_ver = '11'
-elif '11.0' < cuda_ver < '12.0':
-    cuda_major_ver = '11'
-elif '12.0' <= cuda_ver < '13.0':
+print("\n"+"*"*80)
+print("CUDA version:", cuda_ver)
+print("CUDA path:", cuda_path)
+print("*"*80+"\n")
+
+if '12.0' <= cuda_ver < '13.0':
     cuda_major_ver = '12'
+elif '13.0' <= cuda_ver < '14.0':
+    cuda_major_ver = '13'
 else:
     raise RuntimeError(f"Unsupported CUDA version: {cuda_ver}")
 
@@ -78,67 +81,13 @@ def run(self):
 
 class build_ext(_build_ext):
 
-    def _set_library_roots(self):
-        custatevec_root = cutensornet_root = cudensitymat_root = cuquantum_root = None
-        # Note that we need sys.path because of build isolation (since PEP 517)
-        py_paths = sys.path + [site.getusersitepackages()] + site.getsitepackages()
-
-        # search order:
-        # 1. installed "cuquantum" package
-        # 2. env var
-        for path in py_paths:
-            path = os.path.join(path, 'cuquantum')
-            if os.path.isdir(os.path.join(path, 'include')):
-                custatevec_root = cutensornet_root = cudensitymat_root = path
-                break
-        else:
-            # We allow setting CUSTATEVEC_ROOT, CUTENSORNET_ROOT & CUDENSITYMAT_ROOT separately for the ease
-            # of development, but users are encouraged to either install cuquantum from PyPI
-            # or conda, or set CUQUANTUM_ROOT to the existing installation.
-            cuquantum_root = os.environ.get('CUQUANTUM_ROOT')
-            try:
-                custatevec_root = os.environ['CUSTATEVEC_ROOT']
-            except KeyError as e:
-                if cuquantum_root is None:
-                    raise RuntimeError('cuStateVec is not found, please set $CUQUANTUM_ROOT '
-                                       'or $CUSTATEVEC_ROOT') from e
-            try:
-                cutensornet_root = os.environ['CUTENSORNET_ROOT']
-            except KeyError as e:
-                if cuquantum_root is None:
-                    raise RuntimeError('cuTensorNet is not found, please set $CUQUANTUM_ROOT '
-                                       'or $CUTENSORNET_ROOT') from e
-            
-            try:
-                cudensitymat_root = os.environ['CUDENSITYMAT_ROOT']
-            except KeyError as e:
-                if cuquantum_root is None:
-                    raise RuntimeError('cuDensityMat is not found, please set $CUQUANTUM_ROOT '
-                                       'or $CUDENSITYMAT_ROOT') from e
-
-        return custatevec_root, cutensornet_root, cudensitymat_root, cuquantum_root
-
-    def _prep_includes_libs_rpaths(self):
+    def _prep_includes_libs_rpaths(self, ext_name):
         """
-        Set global vars cusv_incl_dir, cutn_incl_dir, cudm_incl_dir, and extra_linker_flags.
+        Set global vars extra_linker_flags.
 
         With the new bindings, we no longer need to link to cuQuantum DSOs.
         """
-        custatevec_root, cutensornet_root, cudensitymat_root, cuquantum_root = self._set_library_roots()
-
-        global cusv_incl_dir, cutn_incl_dir, cudm_incl_dir, cuqnt_incl_dir
-        cusv_incl_dir = cutn_incl_dir = cudm_incl_dir = cuqnt_incl_dir = None
-        base_incl_dir = (os.path.join(cuda_path, 'include'),)
-        if cuquantum_root is not None:
-            cuqnt_incl_dir = base_incl_dir + (os.path.join(cuquantum_root, 'include'),)
-        if custatevec_root is not None:
-            cusv_incl_dir = base_incl_dir + (os.path.join(custatevec_root, 'include'),)
-        if cutensornet_root is not None:
-            cutn_incl_dir = base_incl_dir + (os.path.join(cutensornet_root, 'include'),)
-        if cudensitymat_root is not None:
-            cudm_incl_dir = base_incl_dir + (os.path.join(cudensitymat_root, 'include'),)
-
-        global extra_linker_flags
+
         if not building_wheel:
             # Note: with PEP-517 the editable mode would not build a wheel for installation
             # (and we purposely do not support PEP-660).
@@ -147,43 +96,29 @@ def _prep_includes_libs_rpaths(self):
             # Note: soname = library major version
             # We don't need to link to cuBLAS/cuSOLVER/cuTensor at build time
             # The rpaths must be adjusted given the following full-wheel installation:
-            # - cuquantum-python: site-packages/cuquantum/{custatevec, cutensornet, cudensitymat}/_internal/  [=$ORIGIN]
+            # - cuquantum-python: site-packages/cuquantum/bindings/_internal/  [=$ORIGIN]
             # - cusv, cutn & cudm:      site-packages/cuquantum/lib/
             # - cutensor:         site-packages/cutensor/lib/
             # - cublas:           site-packages/nvidia/cublas/lib/
             # - cusolver:         site-packages/nvidia/cusolver/lib/
             # (Note that starting v22.11 we use the new wheel format, so all lib wheels have suffix -cuXX,
             #  and cuBLAS/cuSOLVER additionally have prefix nvidia-.)
-            ldflag = "-Wl,--disable-new-dtags,"
-            ldflag += "-rpath,$ORIGIN/../../lib,"
-            ldflag += "-rpath,$ORIGIN/../../../cutensor/lib,"
-            ldflag += "-rpath,$ORIGIN/../../../nvidia/cublas/lib,"
-            ldflag += "-rpath,$ORIGIN/../../../nvidia/cusolver/lib"
+            ldflag = "-Wl,--disable-new-dtags"
+            ldflag += ",-rpath,$ORIGIN/../../lib"
+            ldflag += ",-rpath,$ORIGIN/../../../nvidia/cublas/lib"
+            if "cutensornet" in ext_name or "cudensitymat" in ext_name:
+                ldflag += ",-rpath,$ORIGIN/../../../cutensor/lib"
+                ldflag += ",-rpath,$ORIGIN/../../../nvidia/cusolver/lib"
+                ldflag += ",-rpath,$ORIGIN/../../../nvidia/curand/lib"
             extra_linker_flags = [ldflag]
 
-        print("\n"+"*"*80)
-        print("CUDA version:", cuda_ver)
-        print("CUDA path:", cuda_path)
-        print("cuStateVec path:", custatevec_root if custatevec_root else cuquantum_root)
-        print("cuTensorNet path:", cutensornet_root if cutensornet_root else cuquantum_root)
-        print("cuDensityMat path:", cudensitymat_root if cudensitymat_root else cuquantum_root)
-        print("*"*80+"\n")
+        return extra_linker_flags
 
     def build_extension(self, ext):
-        ext.include_dirs = ()
-        for include_dir in (cusv_incl_dir, cutn_incl_dir, cudm_incl_dir, cuqnt_incl_dir):
-            if include_dir is not None:
-                ext.include_dirs += include_dir
-        if ext.name.endswith("custatevec"):
-            ext.extra_link_args = extra_linker_flags
-        elif ext.name.endswith("cutensornet"):
-            ext.extra_link_args = extra_linker_flags
-        elif ext.name.endswith("cudensitymat"):
-            ext.extra_link_args = extra_linker_flags
-
+        ext.include_dirs = (os.path.join(cuda_path, 'include'),)
+        ext.extra_link_args = self._prep_includes_libs_rpaths(ext.name)
         super().build_extension(ext)
 
     def build_extensions(self):
-        self._prep_includes_libs_rpaths()
         self.parallel = 4  # use 4 threads
         super().build_extensions()

python/cuquantum/__init__.py

@@ -3,61 +3,8 @@
 # SPDX-License-Identifier: BSD-3-Clause
 
 from cuquantum import bindings
-from cuquantum import custatevec
-from cuquantum import cutensornet
 from cuquantum import densitymat
 from cuquantum import tensornet
-from cuquantum.cutensornet import (
-    contract, contract_path, einsum, einsum_path, tensor, tensor_qualifiers_dtype, BaseCUDAMemoryManager, CircuitToEinsum, MemoryPointer, 
-    Network, NetworkOptions, OptimizerInfo, OptimizerOptions, PathFinderOptions, 
-    ReconfigOptions, SlicerOptions, MemoryLimitExceeded)
+from cuquantum.memory import *
 from cuquantum.bindings._utils import ComputeType, cudaDataType, libraryPropertyType
-from cuquantum._internal import enum_utils
 from cuquantum._version import __version__
-
-
-# We patch all enum values so that they have the correct docstrings
-for enum in (
-        custatevec.Pauli,
-        custatevec.MatrixLayout,
-        custatevec.MatrixType,
-        custatevec.MatrixMapType,
-        custatevec.Collapse,
-        custatevec.SamplerOutput,
-        custatevec.DeviceNetworkType,
-        cutensornet.NetworkAttribute,
-        custatevec.CommunicatorType,
-        custatevec.DataTransferType,
-        custatevec.StateVectorType,
-        cutensornet.Status,
-        cutensornet.ContractionOptimizerInfoAttribute,
-        cutensornet.ContractionOptimizerConfigAttribute,
-        cutensornet.ContractionAutotunePreferenceAttribute,
-        cutensornet.NetworkAttribute,
-        cutensornet.WorksizePref,
-        cutensornet.WorkspaceKind,
-        cutensornet.Memspace,
-        cutensornet.SmartOption,
-        cutensornet.GraphAlgo,
-        cutensornet.MemoryModel,
-        cutensornet.OptimizerCost,
-        cutensornet.TensorSVDConfigAttribute,
-        cutensornet.TensorSVDAlgo,
-        cutensornet.TensorSVDNormalization,
-        cutensornet.TensorSVDPartition,
-        cutensornet.TensorSVDInfoAttribute,
-        cutensornet.GateSplitAlgo,
-        cutensornet.BoundaryCondition,
-        cutensornet.StatePurity,
-        cutensornet.StateAttribute,
-        cutensornet.MarginalAttribute,
-        cutensornet.SamplerAttribute,
-        cutensornet.AccessorAttribute,
-        cutensornet.ExpectationAttribute,
-        cutensornet.StateMPOApplication,
-        cutensornet.StateMPSGaugeOption,
-        ):
-    enum_utils.add_enum_class_doc(enum, chomp="_ATTRIBUTE|_PREFERENCE_ATTRIBUTE")
-
-del enum
-del enum_utils