It seems reasonable to simplify the api by passing NEPTUNE_API_TOKEN directly into the Session object and create Credentials object inside of that class:

import os
from neptunelib import Session

session = Session(api_token=os.environ['NEPTUNE_API_TOKEN'])

I would like to be able to go:

...
experiments = project.get_experiments()
exp = experiments[0]
images = experiments.get_image_channel('images')
for image in images:
   img = image.get_data()
   break
print(img)

".../python3.6/site-packages/neptunelib/utils.py:62: SettingWithCopyWarning:
A value is trying to be set on a copy of a slice from a DataFrame

See the caveats in the documentation: http://pandas.pydata.org/pandas-docs/stable/indexing.html#indexing-view-versus-copy
channel_df.dropna(inplace=True)"

Adjust to google docstrings.

Missing/need improvement:

• Session
• Credentials
• Project
• Experiment
• .get_hardware_utilization
• .get_numerical_utilization

I have encountered this piece of code:

@staticmethod
def _sort_leaderboard_columns(column_names):

in here.

Also:

@staticmethod
def _simple_dict_to_dataframe(d):

in here
I think it is possibly a typo.

It seems odd to have a staticmethod that we want to show to the world as Projects.sort_leaderboard_columns but at the same time it is private so we don't.
I would go with either private or static or even a simple helper function.

Right now, when .get_numeric_channels_values is called for a list of channels, each channel gets x_ and y_ prefixes. What is more, the x_ frequency is different and misaligned for different channels.

Having a common x_ column for different channels and fill channels with NaN should make it easier to use.

Leaderboard dataframe can be divided into channels/parameters/properties/others

Add simple helpers for:

• getting channel column names
• getting property column names
• getting parameter column names

When initializing Session object in python 3.5 (no problems in 3.6.6) it fails to deserialize the API token.

I would like to be able to go:

project = session.get_projects('neptune-ml/Salt-Detection')
jobs = project.get_jobs(id=['SAL-9999', 'SAL-8293'])
for job in jobs:
   job.trash()

.travis.yml contains the following

...
  user: ...

and other things. Please check if that's correct.

Public projects are not available to the user.

Running:

projects.get_projects('USERNAME')

Shows only the private/public projects of the user but all public projects should be visible.

I ran:

from neptunelib.session import Session
from neptunelib.credentials import Credentials

session = Session(credentials=Credentials.from_env())
project = session.get_projects('neptune-ml')['neptune-ml/Salt-Detection']
project.get_members()

and got the following error

SwaggerMappingError: listProjectMembers does not have parameter organizationName

I don't understand what it means and how to handle it.