Skip to content

Latest commit

 

History

History
33 lines (29 loc) · 1.14 KB

README.md

File metadata and controls

33 lines (29 loc) · 1.14 KB
Raw

Terminal Plotting Using Conspiracy

The main semantics are to use a Log object to store values, the use plot_logs to display them. Example:

from conspiracy import Log, plot_logs
my_log = Log(capacity=1024)
for i in range(5000):
  my_log.log(i)

chart = plot_logs(
  {'my_log':my_log},
  colors={'my_log':'RED'},
  border='line',
  legend=True,
  min_max_y=True,
)
print(chart)

The capacity argument of Log indicates how many values are stored in the log. When you overrun the capacity, the log starts compressing (and averaging) the data stored inside it. capacity can also be 'adaptive' in which case the log will continue to grow to store all logged data. See conspiracy/example.py for more examples.

If you want, you can also simultaneously send data to tensorboard using:

from torch.utils.tensorboard import SummaryWriter
from conspiracy import Log
my_writer = SummaryWriter()
my_log = Log(capacity=1024)
my_log.add_tensorboard_callback(my_writer, 'my_scalar_name')

At this point any calls to my_log.log(value) will also call my_writer.add_scalar('my_scalar_name', value) under the hood.