@@ -7,13 +7,14 @@ dependencies are as follows:
77
88| | CUDA 11 | CUDA 12 |
99| ------------------ | ------------ | ----------- |
10- | CUDA Toolkit [ ^ 1 ] | 11.2 - 11.8 | 12.0 - 12.6 |
10+ | CUDA Toolkit [ ^ 1 ] | 11.2 - 11.8 | 12.x |
1111| Driver | 450.80.02+ (Linux), 452.39+ (Windows) | 525.60.13+ (Linux), 527.41+ (Windows) |
1212
1313[ ^ 1 ] : Including ` cuda-python ` .
1414
1515` cuda.core ` supports Python 3.9 - 3.13, on Linux (x86-64, arm64) and Windows (x86-64).
1616
17+
1718## Installing from PyPI
1819
1920` cuda.core ` works with ` cuda.bindings ` (part of ` cuda-python ` ) 11 or 12. For example with CUDA 12:
@@ -22,8 +23,9 @@ $ pip install cuda-core[cu12]
2223```
2324and likewise use ` [cu11] ` for CUDA 11.
2425
25- Note that using ` cuda.core ` with NVRTC or nvJitLink installed from PyPI via ` pip install ` requires
26- ` cuda.bindings ` 12.8.0+ or 11.8.6+.
26+ Note that using ` cuda.core ` with NVRTC installed from PyPI via ` pip install ` requires
27+ ` cuda.bindings ` 12.8.0+ or 11.8.6+. Likewise, with nvJitLink it requires 12.8.0+.
28+
2729
2830## Installing from Conda (conda-forge)
2931
@@ -33,12 +35,8 @@ $ conda install -c conda-forge cuda-core cuda-version=12
3335```
3436and likewise use ` cuda-version=11 ` for CUDA 11.
3537
36- Note that to use ` cuda.core ` with nvJitLink installed from conda-forge currently requires it to
37- be separately installed:
38- ``` console
39- $ conda install -c conda-forge libnvjitlink
40- ```
41- (can be combined with the command above). This extra step will be removed in a future release.
38+ Note that to use ` cuda.core ` with nvJitLink installed from conda-forge requires ` cuda.bindings ` 12.8.0+.
39+
4240
4341## Installing from Source
4442
@@ -47,4 +45,4 @@ $ git clone https://github.com/NVIDIA/cuda-python
4745$ cd cuda-python/cuda_core
4846$ pip install .
4947```
50- For now ` cuda-python ` ( ` cuda- bindings` later) 11.x or 12.x is a required dependency.
48+ ` cuda-bindings ` 11.x or 12.x is a required dependency.
0 commit comments