def plot_episode_red(df, E): #, center_line, inner_border, outer_border):
fig = plt.figure(1, figsize=(trkSizeX/5, trkSizeY/5))
ax = fig.add_subplot(211)
ax.set_aspect(aspect=1)
pu.print_border(ax, track, color="WhiteSmoke") # center_line, inner_border, outer_border)
episode_data = df[df['episode'] == E]
for row in episode_data.iterrows():
x1,y1,action,reward = row[1]['x'], row[1]['y'], row[1]['action'], row[1]['reward']
car_x2, car_y2 = x1 - 0.02, y1
plt.plot([x1, car_x2], [y1, car_y2], 'r.')
def plot_episode_color(df, E): #, center_line, inner_border, outer_border):
fig = plt.figure(1, figsize=(trkSizeX/5, trkSizeY/5))
ax = fig.add_subplot(211)
ax.set_aspect(aspect=1)
pu.print_border(ax, track, color="WhiteSmoke") # center_line, inner_border, outer_border)
episode_data = df[df['episode'] == E]
for row in episode_data.iterrows():
x1,y1,action,reward = row[1]['x'], row[1]['y'], row[1]['action'], row[1]['reward']
actidx = int(action)
action_color = asl[int(action)].color
action_s = (asl[int(action)].throttle / maxThrottle)**2 * 50 # tune the size of the dots
#print(action_color)
#plt.scatter([x1, car_x2], [y1, car_y2], color=action_color, s=action_s, alpha=0.65)
plt.scatter(x1, y1, color=action_color, s=action_s, alpha=0.75)
# Remeber that evaluation npy files are a community effort to visualise the tracks in the trainings, they aren't 100% accurate.
#
# Tracks Available:
# +
# !dir tracks
tu = TrackIO()
# -
# Take the name from results above and paste below to load the key elements of the track and view the outline of it.
# +
track: Track = tu.load_track("reinvent_base")
pu.plot_trackpoints(track)
# -
# ## Get the logs
#
# Depending on which way you are training your model, you will need a different way to load the data.
#
# **AWS DeepRacer Console**
# The logs are being stored in CloudWatch, in group `/aws/robomaker/SimulationJobs`. You will be using boto3 to download them based on the training ID (stream name prefix). If you wish to bulk export the logs from Amazon Cloudwatch to Amazon S3 :: https://docs.aws.amazon.com/AmazonCloudWatch/latest/logs/S3ExportTasks.html
#
# **DeepRacer for Dummies/ARCC local training**
# Those two setups come with a container that runs Jupyter Notebook (as you noticed if you're using one of them and reading this text). Logs are stored in `/logs/` and you just need to point at the latest file to see the current training. The logs are split for long running training if they exceed 500 MB. The log loading method has been extended to support that.
#
# **Chris Rhodes' repo**
# Chris repo doesn't come with logs storage out of the box. I would normally run `docker logs dr > /path/to/logfile` and then load the file.
#
complete_ones.nsmallest(15, 'time') # View ten most recent lap attempts simulation_agg.nlargest(10, 'timestamp') # ## Plot all the evaluation laps # # The method below plots your evaluation attempts. Just note that that is a time consuming operation and therefore I suggest using `min_distance_to_plot` to just plot some of them. # # If you would like to, in a below section of this article you can load a single log file to evaluate this. # # In the example below training track data was used for plotting the borders. Since then the community has put a lot of effort into preparing files that resemble the racing ones. # # If you want to plot a single lap, scroll down for an example which lets you do a couple more tricks. pu.plot_evaluations(bulk, track) # ## Single lap # Below you will find some ideas of looking at a single evaluation lap. You may be interested in a specific part of it. This isn't very robust but can work as a starting point. Please submit your ideas for analysis. # # This place is a great chance to learn more about [Pandas](https://pandas.pydata.org/pandas-docs/stable/) and about how to process data series. # Load a single lap lap_df = bulk[(bulk['episode'] == 0) & (bulk['stream'] == 'sim-sample')] # We're adding a lot of columns here to the episode. To speed things up, it's only done per a single episode, so others will currently be missing this information. # # Now try using them as a `graphed_value` parameter. # + lap_df.loc[:, 'distance'] = ((lap_df['x'].shift(1) - lap_df['x'])**2 +
tu = TrackIO()
for f in tu.get_tracks():
print(f)
# -
# Take the name from results above and paste below to load the key elements of the track and view the outline of it.
# + jupyter={"source_hidden": true}
track: Track = tu.load_track("reinvent_base")
l_track = track.center_line
l_outer_border = track.outer_border
l_inner_border = track.inner_border
pu.plot_trackpoints(track)
# -
# ## Get the logs
#
# Depending on which way you are training your model, you will need a different way to load the data.
#
# **AWS DeepRacer Console**
# The logs are being stored in CloudWatch, in group `/aws/robomaker/SimulationJobs`. You will be using boto3 to download them based on the training ID (stream name prefix). If you wish to bulk export the logs from Amazon Cloudwatch to Amazon S3 :: https://docs.aws.amazon.com/AmazonCloudWatch/latest/logs/S3ExportTasks.html
#
# **DeepRacer for Dummies/ARCC local training**
# Those two setups come with a container that runs Jupyter Notebook (as you noticed if you're using one of them and reading this text). Logs are stored in `/logs/` and you just need to point at the latest file to see the current training. The logs are split for long running training if they exceed 500 MB. The log loading method has been extended to support that.
#
# **Chris Rhodes' repo**
# Chris repo doesn't come with logs storage out of the box. I would normally run `docker logs dr > /path/to/logfile` and then load the file.
#