fix a small bug in `data clone` command
support downloading subdirectory from dataset and job output
* data status command fix (restrict to dataset)
* progress bar for data compression
* fix login for Safari
* new login experience
* better error message for invalid data name in floyd config
* floyd run --follow command now exits automatically on job exit
* deprecate support for GPU+ and CPU+ instance type
* handle file permission issue on data upload
* support clone without init
* fix mode support in floyd config
Fix data mount for command job working directory
Fix --gpu and --cpu flag.
Support floyd config. For more info, see: https://docs.floydhub.com/floyd_config
- Use username/password as the default login mechanism. Token based login is still retained.
- Include latest metrics info to the experiment status and info commands.
add support for --max-runtime to run command
* fixed crash due to latest version of pip (9.0.2)
* display error to user about upload that's over allowed size
* added --follow option to floyd run command by mrahtz
* Add data listfiles and data getfile command by mrahtz
* sort env list
* fix `unexpected keyword argument 'strict'` crash
* (UX) display job URL at the end of run and restart command
* (bug) fix error in `floyd info` command
Thanks msoedov for the following patches!
* better error message for network timeout error
* new -f option for logs command
- login command now takes password as a hidden field for improved security
- Fix for 135
- Smart resolution of project and dataset names. The names can now be shortened and the CLI will intelligently resolve it. Some example usage:
| Previous Version | New Version |
| ----------------- | --------------- |
| $ floyd logs alice/projects/cnn/2 | $ floyd logs |
| $ floyd output alice/projects/rnn/5 | $ floyd output rnn/5 |
| $ floyd data output alice/datasets/cats/5 | $ floyd data output cats/5 |
| $ floyd run --data alice/datasets/cats/5:/cats script.py | $ floyd run --data cats/5:/cats script.py |
More details about this feature is available on [FloydHub docs](https://docs.floydhub.com/guides/shortnames/).
- Add option to login with username and password:
`floyd login -u <username> -p <password`
This can be used as an alternative to using the token from the website.
- Remove whitespace in tokens pasted during login.
* fixed default input mount path for mounting datasets
* make output command work with running job's output
* suggest user to split out data for large code sync
* fix exception in floyd clone command
* improve error message
* added to_dict method to jobs
* replace all IDs with human friendly names
* make human friendly name consistent with what's on the web UI
* show error message when a data delete request is sent for job output
* fixed issue with fetching job and data outputs
* updated default ignore list
* fixed bug with force version upgrade
- Update the logs request to use Resource api. (Hotfix)
- Log cli output to stdout
- Replace more IDs with NAME from output
- Tensorflow 1.1 environment is now available. It becomes the default environment.
- Keras environment now contains Keras 2.0.4 and Tensorflow 1.1.0.
- For backwards compatibility, `tensorflow-0.12` and `tensorflow-1.0` environments are also available.
- Support for attaching multiple datasets to the same floyd run. Now you can specify `--data <ID1> --data <ID2>`
- Specify mount points for data. You can specify the path where your data should be mounted. Eg: `--data <ID>:pathname` and the data will be available at `/pathname`.
- Support for deleting multiple datasets and experiments in a single command. 14 and 21
- Bug Fix: Too many files error was getting thrown even if directories were added to .floydignore
- Additional tests
Support for deleting multiple experiment ids. 14
Use proper encoding for server response 11