Age | Commit message (Collapse) | Author | Files | Lines |
|
Make tensorflow (and a bunch of ther things) use CUDA-compatible
toolchain. Introduces cudaPackages.backendStdenv
|
|
|
|
|
|
Previously, the cutensor samples could not find the libcutensor.so.1
shared library at runtime. This patch adds cutensor as a buildInput so
the shared library is linked in properly.
|
|
|
|
|
|
This reverts commit 90c505790c63e34a72a86efcd248de6c9da58ac6.
|
|
|
|
There are many different versions of the `cudatoolkit` and related
cuda packages, and it can be tricky to ensure they remain compatible.
- `cudaPackages` is now a package set with `cudatoolkit`, `cudnn`, `cutensor`, `nccl`, as well as `cudatoolkit` split into smaller packages ("redist");
- expressions should now use `cudaPackages` as parameter instead of the individual cuda packages;
- `makeScope` is now used, so it is possible to use `.overrideScope'` to set e.g. a different `cudnn` version;
- `release-cuda.nix` is introduced to easily evaluate cuda packages using hydra.
|
|
|
|
|
|
Remove old CUDA toolkits (and corresponding CuDNN versions).
- Not supported by upstream anymore.
- We do not use them in nixpkgs.
- We do not test or actively maintain them.
- Anything but ancient GPUs is supported by newer toolkits.
Fixes #107131.
|
|
Fix mispelling. That word always breaks me...
|
|
|
|
Well, strictly speaking, master +
https://github.com/NVIDIA/CUDALibrarySamples/pull/29
|
|
Since CUDA is unfree, we won't actually use this when testing Nixpkgs
officially. But I want to include this as they are useful for users of
Nixpkgs trying to set up / debug a CUDA environment.
|