You can not select more than 25 topics Topics must start with a chinese character,a letter or number, can include dashes ('-') and can be up to 35 characters long.

README.md 646 B

12345678910111213141516171819
  1. # Generate device kernel registration code for CUTLASS kernels
  2. ## Usage
  3. ```bash
  4. python3 generator.py [--operations {gemm, gemv, conv2d, deconv}] [--type {simt, tensorop8816, tensorop8832}]
  5. output
  6. ```
  7. - operations: operation kind, including gemm|gemv|conv2d|deconv
  8. - type: opcode class, simt|tensorop8816|tensorop8832
  9. - output: the output directory for CUTLASS kernels
  10. ## Generate file list for bazel
  11. We generate `list.bzl` because the `genrule` method of bazel requires that the output file list be specified in the analysis phase.
  12. Please call `gen_list.py` when new operations are added.
  13. ```bash
  14. python3 gen_list.py
  15. ```

MegEngine 安装包中集成了使用 GPU 运行代码所需的 CUDA 环境,不用区分 CPU 和 GPU 版。 如果想要运行 GPU 程序,请确保机器本身配有 GPU 硬件设备并安装好驱动。 如果你想体验在云端 GPU 算力平台进行深度学习开发的感觉,欢迎访问 MegStudio 平台