- Slack:
- Email: [email protected]
- Web: https://support.pipeline.ai
- Troubleshooting: Guide
- Docker
- Python 2 or 3 (Conda is Preferred)
- (Windows Only) PowerShell
- For GPU-based models, make sure you specify
--model-chip=gpu
or--start-cmd=nvidia-docker
where appropriate - For GPU-based models, sure you have
nvidia-docker
installed!
- If your server starts, try running
pipeline predict-server-logs
orpipeline train-server-logs
to collect the startup logs. - Please send all relevant information to PipelineAI Support (Slack, Email, or Web).
- You likely have a pickling issue.
- Make sure that you are using Python 3. We see issues when training on Python 2 (your environment), then unpickling with Python 3 (PipelineAI Model Runtime Environment)
- Also, we've found the
cloudpickle
library to be the most stable (versus Python's defaultpickle
anddill
.)
WARNING: Your kernel does not support swap limit capabilities or the cgroup is not mounted. (*-start
)
- Just ignore this
- Instead of
localhost
, you may need to use192.168.99.100
or another IP/Host that maps to your local Docker host. - This usually happens when using Docker Quick Terminal on Windows 7.
- Likely an issue installing the PipelineCLI
- The cli requires Python 2 or 3 - preferably Miniconda
- Try
sudo pip install ...
- Try
pip install --user ...
Could not delete '/usr/local/lib/python3.5/dist-packages/markupsafe/_speedups.cpython-35m-x86_64-linux-gnu.so': Permission denied
- Try
sudo pip install ...
- You need Python 2 or 3 (Conda Preferred)
- (Windows Only) PowerShell
pipeline version
### EXPECTED OUTPUT ###
cli_version: 1.5.xx <-- Exact Version You Are Expecting
sudo pip uninstall cli-pipeline
which pipeline
(or otherwise find which pipeline
binary is being used.)
rm <bad-version-of-pipeline>
ll /usr/local/lib/python3.6/dist-packages/cli-pipeline*
rm -rf /usr/local/lib/python3.6/dist-packages/cli-pipeline*
ll ~/.local/lib/python3.6/site-packages/cli_pipeline*
rm -rf ~/.local/lib/python3.6/site-packages/cli_pipeline*
- Re-install
pip install cli-pipeline==...
--model-path
needs to be relative- On Windows, be sure to use the forward slash
\
for your paths.
- Add
--http-proxy=...
and--https-proxy=...
if you seeCondaHTTPError: HTTP 000 CONNECTION FAILED for url
- If you see
CondaHTTPError: HTTP 000 CONNECTION FAILED for url
or[Errno 111] Connection refused'
orConnectionError(MaxRetryError("HTTPSConnectionPool
, you need to update./tensorflow/census/model/pipeline_condarc
to include proxy servers per THIS document. - For
pip
installs, you may also need toexport HTTP_PROXY
andexport HTTPS_PROXY
within./tensorflow/census/model/pipeline_setup.sh
- You may also need to set the lower-case version, as well:
export http_proxy
andexport https_proxy
- And if your proxy server password uses special characters (ie.
@
, you may need to convert the characters to ASCII (ie.%40
). - You may need to remove the
!#/bin/bash
- Lastly, you may need to set shell-wide (
~/.bashrc
) and re-source (. ~/.bashrc
) - or set them as system-wide environment variables (
http://phutchins.com/blog/2017/01/04/fixing-docker-no-space-left-on-device/
- Slack:
- Email: [email protected]
- Web: https://support.pipeline.ai
- Troubleshooting: Guide