- removed command group aliases
- streaming logs for
experiments createAndStart
by default
- new params for
job create
andhyperparameters create
- fixed missing
--experimentType
when creating new single node experiment - deployments can be of
TFSERVING
type only - fixed autocomplete instructions
- autocomplete for bash (GNU/Linux and iOS)
- aliases in singular form for command groups
- showing logs for experiments and jobs
- restored and deprecated
run
command - workspace uploading for experiments
- workspace upload progress bar
- partial support for hyperparameter tuning
- prompt for apiKey if not provided
- deprecate
login
command --ignoreFiles
option forexperiments create
--ignoreFiles
and--workspace none
options forjobs create
- restore apiKey command
- colorful error and help messages
- removed
paperspace-python run
command
- partial support for managing jobs, models and projects
- filtering deployments list by project ID and model ID
- reading logs of experiments and jobs
- handling error response when listing experiments
- experiments and deployments management
- --apiKey for all commands for one-time api key overwrite
- help messages for all commands and parameters
- filtering experiments list with
--projectHandle
- paginating experiments list when does not fit the terminal width
- added
--modelType
and--modelPath
toexperiments create
- some minor bug fixes
- support for experiments API - creating, starting, listing, etc.
- Fix missing import
- Add statsd agent support with StatsdClient
- Fix crashing on converting to bytes in python < 3
-
Run gradient jobs from custom Dockerfile built containers (see: https://docs.paperspace.com/gradient/jobs/create-a-job#new-run-jobs-from-dockerfiles)
-
Push GPU-enabled container images to a registry of your choice
- Add custom headers to requests
- Allow user to select preemptible machine types
- Fix package.egg-info issue affecting Windows users downloading paperspace from PyPI
- Allow user to select custom port mappings between container and host
- Fix circular import bug causing problems for python2 users on v0.0.12
- New cluster and machineType functionality supported: send jobs to a gradient-node based on the cluster name or clusterId
- For more information on running gradient-node see the Gradient Private Cloud docs
- Send jobs to a specific node by specifying node attributes
- Changed default handling of machineType: cluster preferences come first
- Jobs create automatically records the git commit hash for local workspaces using git
- New jobs machineTypes method for discovering available job machine types.
- Minor doc fixes
paperspace-python run
command for running python scripts unmodified on paperspace- run a python script, module, or command string remotely
- run a executable or shell command remotely
- use pip, pipenv, or a shell script to manage python dependencies
- Improved docs
- Minor doc fixes
- The default machineType for paperspace-python is now P5000, since GPU+ machines are not yet available for jobs.
- The default container image for paperspace-python is 'docker.io/paperspace/tensorflow-python' which adds python3 support to Google's 'gcr.io/tensorflow/tensorflow:1.5.0-gpu' container.
- This is the first official release of the Paperspace Python package.
- It provides support for Gradient jobs and running python modules as jobs.
- There are python packages available on PyPI: https://pypi.python.org/pypi/paperspace
- And on Anaconda.org for Conda: https://anaconda.org/paperspace/paperspace