[PYTHON] Reinforcement learning 8 Try using Chainer UI

It is assumed that reinforcement learning 7 has been completed. In order to carry out reinforcement learning, you will want to see the learning process. Use "Chainer UI" which is a visualization / experiment management tool of Chainer.

ChainerUI https://github.com/chainer/chainerui https://research.preferred.jp/2017/12/chainerui-release/

Install Chainer UI.

pip install chainerui
chainerui db create
chainerui db upgrade

Go to the parent directory and then create the project.

chainerui project create -d CartPole
chainerui server

Then go to http: // localhost: 5000 / in your web browser. The order around here is important.

scores.txt will not be read. Therefore, add the following code to train.py or execute it with another code.

`convert.py`


import pandas as pd
import json
import codecs
import numpy as np

scores = pd.read_table('result/scores.txt')
columns = scores.columns  #header information

data_list = []
for column in columns:
    data_list.append([column, ''])
data_hash = dict(data_list) #Hash template for each station data

log_json = []
for score in np.array(scores):
    score_hash = data_hash.copy()
    for i in range(columns.size):
        score_hash[columns[i]] = score[i]
    log_json.append(score_hash)

print(log_json)
f = codecs.open('result/log', 'w', 'utf-8')
json.dump(log_json, f, indent=4, ensure_ascii=False)
f.close

It seems that print (log_json) is unnecessary, but when I run it with jupyter notebook, the end of the log file is strange unless I put it in. You can now see the graph in the chainer UI.

ChainerUI has 5 fixed X axes of graph, "epoch", "iteration", "episode", "step", "elapsed_time". scores.txt is slightly different, so I will modify ChainerUI. Demon remodeling in the 9th serialization.

First, regarding the location of the file, there is anaconda3 in the personal folder. Look inside here. userFolder/anaconda3/envs/chainer/lib/python3.7/site-packages/chainerui Here is the program. userFolder/anaconda3/envs/chainer/lib/python3.7/site-packages/chainerui/static/dist/chainerui.js Open this file. In this file, T = ["epoch", "iteration", "episode", "step", "elapsed_time"], , T = ["epoch", "iteration", "episode", "step", "elapsed_time","steps","episodes","elapsed"], Rewrite to. Now you can specify the X axis as steps etc.