Kernels:
Trusted publisher
Uploaded using `kernel-builder`.
Browse files
README.md
ADDED
|
@@ -0,0 +1,41 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
---
|
| 2 |
+
library_name: kernels
|
| 3 |
+
license: mit
|
| 4 |
+
---
|
| 5 |
+
|
| 6 |
+
This is the repository card of kernels-community/deep-ep that has been pushed on the Hub. It was built to be used with the [`kernels` library](https://github.com/huggingface/kernels). This card was automatically generated.
|
| 7 |
+
|
| 8 |
+
## How to use
|
| 9 |
+
|
| 10 |
+
Usage example not available.
|
| 11 |
+
|
| 12 |
+
## Available functions
|
| 13 |
+
|
| 14 |
+
Function list not available.
|
| 15 |
+
|
| 16 |
+
## Requirements
|
| 17 |
+
|
| 18 |
+
This is [DeepEP](https://github.com/deepseek-ai/DeepEP) 2.5.0 (`def8651`), with
|
| 19 |
+
[DeepJIT](https://github.com/deepseek-ai/DeepJIT) `cba802d`. Only DeepEP's host
|
| 20 |
+
extension is prebuilt; its communication kernels are compiled at runtime. On
|
| 21 |
+
top of the Hub kernel's own requirements, using it needs:
|
| 22 |
+
|
| 23 |
+
- NVIDIA Hopper or newer GPUs, connected with NVLink (RDMA for multi-node).
|
| 24 |
+
- NCCL >= 2.32.3, loaded by PyTorch. The kernel does not ship NCCL: it uses the
|
| 25 |
+
NCCL library that PyTorch loads, and compiles its kernels against that
|
| 26 |
+
library's headers. PyTorch 2.14 and 2.15 ship NCCL 2.30.7; upgrade it with
|
| 27 |
+
`pip install "nvidia-nccl-cu13>=2.32.3"` (or use a PyTorch build that ships
|
| 28 |
+
2.32.3 or newer).
|
| 29 |
+
- A CUDA toolkit >= 13.1 (`nvcc`) and a C++20 host compiler at runtime, found
|
| 30 |
+
through `CUDA_HOME`, `CUDA_PATH`, `nvcc` on `PATH` or `/usr/local/cuda`.
|
| 31 |
+
|
| 32 |
+
These are checked when the first buffer is created, with an error explaining
|
| 33 |
+
what is missing. Compiled kernels are cached in `$HOME/.dj` (`EP_JIT_CACHE_DIR`).
|
| 34 |
+
|
| 35 |
+
## Benchmarks
|
| 36 |
+
|
| 37 |
+
No benchmark available yet.
|
| 38 |
+
|
| 39 |
+
## Source code
|
| 40 |
+
|
| 41 |
+
Source code of this kernel originally comes from https://github.com/deepseek-ai/DeepEP and it was repurposed for compatibility with `kernels`.
|