Version 1.0.3

Fixing issue with shapely version >2.0

Version 1.0.3 RC1

deepracer-utils-1.0.3rc1

Fix shapely dependency

Version 1.0.2

deepracer-utils-1.0.2

Flexible loading

Version 1.0.1

Merge pull request #41 from aws-deepracer-community/eval-folder-only

Additional LogFolder Type

Version 1.0.0

Major feature update - includes a set of new capabilities, including the ability to read logs directly from Console and S3 bucket.

v1.0.0 RC1

Merge pull request #37 from aws-deepracer-community/dev

Feature Development

Add AWS CLI support

After you install this version from pypi, run

python -m deepracer install-cli

and then

aws deepracer help

(assuming you have aws-cli installed for that to work)

Introduce a unified log loader which supports trace logs in csv format

If you have AWS DeepRacer console logs downloaded extracted in folder model-name, simply do

from deepracer.logs import DeepRacerLog
log = DeepRacerLog("path/to/model-name")
log.load()

df = log.dataframe()

try:
    print(log.agent_and_network())
    print("-------------")
    print(log.hyperparameters())
    print("-------------")
    print(log.action_space())
except Exception:
    print("Robomaker logs not available")

Not that everything but log.load() works only if the robomaker log is also available (so only for console logs for now)

There is a known issue where converting a dataframe into simulation_aggregates fails. This is due to a slight naming mismatch. Until it's fixed you can work around it with

df["throttle"]=df["speed"]
df["timestamp"]=df["tstamp"]
df["steer"]=df["steering_angle"]