envd
envd copied to clipboard
tensorchord
bug: arm64 binaries are not available on PyPI
Currently, we only distribute amd64 on PyPI for both Mac and Linux but not arm64.
- name: Configure linux build environment
if: runner.os == 'Linux'
run: |
mkdir -p bin
mv dist/envd_linux_amd64_v1/envd bin/envd
chmod +x bin/envd
- name: Configure macOS build environment
if: runner.os == 'macOS'
run: |
mkdir -p bin
mv dist/envd_darwin_amd64_v1/envd bin/envd
chmod +x bin/envd
We should support arm64, especially for Apple M1 users.
/assign
I can help improve the CI, but I cannot test if it works since I do not have a machine for it.
@popfido
Thanks! could you please run the commands to build and test if the build works on ARM64 macOS:
make
./bin/envd bootstrap
./bin/envd up -p ./pkg/app/testdata/build-test
We can add ci with arm by qemu https://github.com/docker/setup-qemu-action. Although I head the performance is very bad
@gaocegege
Here is the ./bin/envd up -p ./pkg/app/testdata/build-test results in cmd.
[+] ⌚ parse build.envd and download/cache dependencies 0.0s ✅ (finished)
[+] 🐋 build envd environment 90.1s (11/11) ✅ (finished)
=> docker-image://docker.io/tensorchord/python:3.8-ubuntu20.04 61.4s
=> => resolve docker.io/tensorchord/python:3.8-ubuntu20.04 5.7s
=> => sha256:c03f0eb9ed3e9ecc9777137ca64d13d1e7dba88eadcc151074a918d94821415c 3.20MB / 3.20MB 1.4s
=> => sha256:dd0b398750f4b004a65c1bb9bf2c93b7a753756afaea45ac0dd29ab305661141 58.96MB / 58.96MB 31.1s
=> => sha256:3e66387987ad1b955b18ed108c3e53074f997ec0b7d93a02b11308276e0ddcf6 91.22MB / 91.22MB 51.1s
=> => sha256:d7bfe07ed8476565a440c2113cc64d7c0409dba8ef761fb3ec019d7e6b5952df 28.57MB / 28.57MB 11.2s
=> => extracting sha256:d7bfe07ed8476565a440c2113cc64d7c0409dba8ef761fb3ec019d7e6b5952df 0.9s
=> => extracting sha256:3e66387987ad1b955b18ed108c3e53074f997ec0b7d93a02b11308276e0ddcf6 2.7s
=> => extracting sha256:dd0b398750f4b004a65c1bb9bf2c93b7a753756afaea45ac0dd29ab305661141 1.8s
=> => extracting sha256:c03f0eb9ed3e9ecc9777137ca64d13d1e7dba88eadcc151074a918d94821415c 0.1s
=> [internal] create user group envd 0.2s
=> [internal] create user envd 0.2s
=> [internal] add user envd to sudoers 0.3s
=> [internal] configure user permissions 0.1s
=> [internal] configure user permissions 7.0s
=> install ssh keys 0.0s
=> install PyPI packages 0.0s
=> install ssh keys 0.0s
=> merging all components into one 0.0s
=> exporting to oci image format 20.8s
=> => exporting layers 5.5s
=> => exporting manifest sha256:06b240e4fe4d92d1110347c4fa2356f5ca0ed436cf132d640220ba547581dde3 0.0s
=> => exporting config sha256:5538a6145c5e321d4d2dcd068970e85ab85c117b4119794182b80043556ab665 0.0s
=> => sending tarball 15.4s
To run a command as administrator (user "root"), use "sudo <command>".
See "man sudo_root" for details.
Thanks! Then it should work!
@all-contributors please add @popfido for platform
![allcontributors[bot] avatar](https://avatars.githubusercontent.com/in/23186?v=4 "Published by a pub.dev verified publisher"){kind=small}
I am not sure how to add M1 build in our CI and CD process.
Here are some references:
- https://github.com/actions/runner/issues/805
- https://github.com/atoti/jdk4py/pull/54/
m1 macbook pro failed to run mnist cpu version example
[+] ⌚ parse build.envd and download/cache dependencies 0.0s ✅ (finished)
=> 💽 (cached) download oh-my-zsh 0.0s
=> 💽 (cached) download ms-python.python 0.0s
[+] 🐋 build envd environment 7.6s (18/20)
=> docker-image://docker.io/tensorchord/python:3.8-ubuntu20.04 2.8s
=> => resolve docker.io/tensorchord/python:3.8-ubuntu20.04 2.8s
=> local://cache-dir 1.8s
=> => transferring cache-dir: 430.41kB 1.6s
=> 💽 (cached) [internal] create user group envd 0.0s
=> 💽 (cached) [internal] create user envd 0.0s
=> 💽 (cached) [internal] add user envd to sudoers 0.0s
=> 💽 (cached) [internal] configure user permissions 0.0s
=> 💽 (cached) [internal] configure user permissions 0.0s
=> 💽 (cached) [internal] settings pip cache mount permissions 0.0s
=> 💽 (cached) install vscode plugin ms-python.python 0.0s
=> 💽 (cached) install system packages 0.0s
=> 💽 (cached) install ssh keys 0.0s
=> 💽 (cached) install ssh keys 0.0s
=> 🔥 (error) pip install tensorflow numpy jupyter 4.7s
=> 💽 (cached) copy /oh-my-zsh /home/envd/.oh-my-zsh 0.0s
=> 💽 (cached) mkfile /home/envd/install.sh 0.0s
=> 💽 (cached) install oh-my-zsh 0.0s
=> 💽 (cached) mkfile /home/envd/.zshrc 0.0s
=> 💽 (cached) install PyPI packages 0.0s
------
> pip install tensorflow numpy jupyter:
#0 4.479 ERROR: Exception:
#0 4.479 Traceback (most recent call last):
#0 4.479 File "/usr/lib/python3/dist-packages/pip/_internal/cli/base_command.py", line 186, in _main
#0 4.479 status = self.run(options, args)
#0 4.479 File "/usr/lib/python3/dist-packages/pip/_internal/commands/install.py", line 305, in run
#0 4.479 session = self.get_default_session(options)
#0 4.479 File "/usr/lib/python3/dist-packages/pip/_internal/cli/req_command.py", line 74, in get_default_session
#0 4.479 self._session = self.enter_context(self._build_session(options))
#0 4.479 File "/usr/lib/python3/dist-packages/pip/_internal/cli/req_command.py", line 84, in _build_session
#0 4.479 session = PipSession(
#0 4.479 File "/usr/lib/python3/dist-packages/pip/_internal/network/session.py", line 241, in __init__
#0 4.479 self.headers["User-Agent"] = user_agent()
#0 4.479 File "/usr/lib/python3/dist-packages/pip/_internal/network/session.py", line 132, in user_agent
#0 4.479 zip(["name", "version", "id"], distro.linux_distribution()),
#0 4.479 File "/usr/share/python-wheels/distro-1.4.0-py2.py3-none-any.whl/distro.py", line 122, in linux_distribution
#0 4.479 return _distro.linux_distribution(full_distribution_name)
#0 4.479 File "/usr/share/python-wheels/distro-1.4.0-py2.py3-none-any.whl/distro.py", line 677, in linux_distribution
#0 4.479 self.version(),
#0 4.479 File "/usr/share/python-wheels/distro-1.4.0-py2.py3-none-any.whl/distro.py", line 737, in version
#0 4.479 self.lsb_release_attr('release'),
#0 4.479 File "/usr/share/python-wheels/distro-1.4.0-py2.py3-none-any.whl/distro.py", line 899, in lsb_release_attr
#0 4.479 return self._lsb_release_info.get(attribute, '')
#0 4.479 File "/usr/share/python-wheels/distro-1.4.0-py2.py3-none-any.whl/distro.py", line 552, in __get__
#0 4.479 ret = obj.__dict__[self._fname] = self._f(obj)
#0 4.479 File "/usr/share/python-wheels/distro-1.4.0-py2.py3-none-any.whl/distro.py", line 1012, in _lsb_release_info
#0 4.479 stdout = subprocess.check_output(cmd, stderr=devnull)
#0 4.479 File "/usr/lib/python3.8/subprocess.py", line 415, in check_output
#0 4.479 return run(*popenargs, stdout=PIPE, timeout=timeout, check=True,
#0 4.479 File "/usr/lib/python3.8/subprocess.py", line 516, in run
#0 4.479 raise CalledProcessError(retcode, process.args,
#0 4.479 subprocess.CalledProcessError: Command '('lsb_release', '-a')' returned non-zero exit status 1.
------
ERRO[2022-06-15T19:37:44+08:00] failed to solve LLB: failed to solve: process "/dev/.buildkit_qemu_emulator pip install --no-warn-script-location tensorflow numpy jupyter" did not complete successfully: exit code: 2 tag="mnist:dev"
ERRO[2022-06-15T19:37:44+08:00] failed to load docker image: Post "http://%2Fvar%2Frun%2Fdocker.sock/v1.41/images/load?quiet=1": context canceled tag="mnist:dev"
error: process "/dev/.buildkit_qemu_emulator pip install --no-warn-script-location tensorflow numpy jupyter" did not complete successfully: exit code: 2
Gotcha, let me test if it works on Linux arm64 first. Then let's see why it fails on macOS arm64.
Gotcha, let me test if it works on Linux arm64 first. Then let's see why it fails on macOS arm64.
after update docker for mac and brew install gcc. work fine
It is weird. I cannot figure out why it requires GCC.
BTW, what's your previous docker version?
It is weird. I cannot figure out why it requires GCC.
BTW, what's your previous docker version?
docker version: v4.9.0, maybe not GCC, I try again without GCC
It is weird. I cannot figure out why it requires GCC.
BTW, what's your previous docker version?
BTW, how to clear cache
@kenwoodjw Do you mean buildkit cache or envd's cache (for oh-my-zsh and vscode extensions)
@kenwoodjw Do you mean buildkit cache or envd's cache (for oh-my-zsh and vscode extensions)
pip cache is buildkit cache ?
Yep. You can docker stop envd_buildkitd && docker rm envd_buildkitd && envd bootstrap to create a new clean buildkitd.
It is weird. I cannot figure out why it requires GCC. BTW, what's your previous docker version?
docker version: v4.9.0, maybe not GCC, I try again without GCC
Ya, without GCC also work fine!
Then it is related to the docker version. It is reasonable although I do not know why hhh
We can add ci with arm by qemu https://github.com/docker/setup-qemu-action. Although I head the performance is very bad
It should be fine. We've been using it in argo-workflows.
How about this https://veertu.com/?
Originally posted by @aseaday in https://github.com/tensorchord/envd/issues/414#issuecomment-1171382038
https://github.com/github/roadmap/issues/528
https://github.blog/changelog/2022-08-09-github-actions-self-hosted-runners-now-support-apple-m1-hardware/
GitHub supports M1 now (in beta)
Let's give it a try.
https://dev.to/thewraven/universal-macos-binaries-with-go-1-16-3mm3 I am creating a universal bin for the build.
https://github.com/gaocegege/envd/actions/runs/3562709880/jobs/5984732217 It works!
Let's summarize the change in my personal fork and submit a PR here.