You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
{{ message }}
This repository has been archived by the owner on Oct 19, 2024. It is now read-only.
Please describe the bug
Hi, according to the alpa installation doc, we need to pip3 install cupy-cuda11x to install cupy. However, when CUDA version is 11.1, according to cupy package info here and here, the correct command should be pip3 install cupy-cuda111, as using cupy-cuda11x will result in errors in the next step (python3 -c "from cupy.cuda import nccl").
Please describe the expected behavior python3 -c "from cupy.cuda import nccl" does not print any error.
System information and environment
OS Platform and Distribution (e.g., Linux Ubuntu 16.04, docker): Linux Ubuntu 22.04
Python version: 3.8.10
CUDA version: 11.1
NCCL version: 2.7.8
To Reproduce
Steps to reproduce the behavior:
Follow the installation guide using CUDA 11.1 and cudnn 8.0.5.
The text was updated successfully, but these errors were encountered:
Hi, I pip install cupy-cuda11x, and pip3 install alpa and jaxlib. But when I check the installation by python3 -m alpa.test_install, An error occurred:
ERROR: test_2_pipeline_parallel (__main__.InstallationTest)
-------------------------------------------------------------------
......
self._sender_tasks[sender_worker].append(
jax._src.traceback_util.UnfilteredStackTrace: KeyError: Actor(MeshHostWorker, d022bdfe8bc5c769e2ce7fa6302000000)
The stack trace below excludes JAX-internal frames.
The preceding is the original exception that occurred, unmodified.
Hi, I pip install cupy-cuda11x, and pip3 install alpa and jaxlib. But when I check the installation by python3 -m alpa.test_install, An error occurred:
ERROR: test_2_pipeline_parallel (__main__.InstallationTest)
-------------------------------------------------------------------
......
self._sender_tasks[sender_worker].append(
jax._src.traceback_util.UnfilteredStackTrace: KeyError: Actor(MeshHostWorker, d022bdfe8bc5c769e2ce7fa6302000000)
The stack trace below excludes JAX-internal frames.
The preceding is the original exception that occurred, unmodified.
Have you ever encountered this problem? Thanks!
I was not able to pass installation test 2 all the time due to different errors with various versions. You can file a new issue about the specific error you met, but I feel the project is not actively maintained.
Sign up for freeto subscribe to this conversation on GitHub.
Already have an account?
Sign in.
Please describe the bug
Hi, according to the alpa installation doc, we need to
pip3 install cupy-cuda11x
to install cupy. However, when CUDA version is 11.1, according to cupy package info here and here, the correct command should bepip3 install cupy-cuda111
, as usingcupy-cuda11x
will result in errors in the next step (python3 -c "from cupy.cuda import nccl"
).Please describe the expected behavior
python3 -c "from cupy.cuda import nccl"
does not print any error.System information and environment
To Reproduce
Steps to reproduce the behavior:
Follow the installation guide using CUDA 11.1 and cudnn 8.0.5.
The text was updated successfully, but these errors were encountered: