Upload folder using huggingface_hub
Browse files- .github/workflows/update_space.yml +28 -0
- .gitignore +174 -0
- README.md +4 -8
- requirements.txt +3 -0
- rl_gradio.py +564 -0
.github/workflows/update_space.yml
ADDED
@@ -0,0 +1,28 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
name: Run Python script
|
2 |
+
|
3 |
+
on:
|
4 |
+
push:
|
5 |
+
branches:
|
6 |
+
- main
|
7 |
+
|
8 |
+
jobs:
|
9 |
+
build:
|
10 |
+
runs-on: ubuntu-latest
|
11 |
+
|
12 |
+
steps:
|
13 |
+
- name: Checkout
|
14 |
+
uses: actions/checkout@v2
|
15 |
+
|
16 |
+
- name: Set up Python
|
17 |
+
uses: actions/setup-python@v2
|
18 |
+
with:
|
19 |
+
python-version: '3.12.8'
|
20 |
+
|
21 |
+
- name: Install Gradio
|
22 |
+
run: python -m pip install gradio
|
23 |
+
|
24 |
+
- name: Log in to Hugging Face
|
25 |
+
run: python -c 'import huggingface_hub; huggingface_hub.login(token="${{ secrets.hf_token }}")'
|
26 |
+
|
27 |
+
- name: Deploy to Spaces
|
28 |
+
run: gradio deploy
|
.gitignore
ADDED
@@ -0,0 +1,174 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
# Byte-compiled / optimized / DLL files
|
2 |
+
__pycache__/
|
3 |
+
*.py[cod]
|
4 |
+
*$py.class
|
5 |
+
|
6 |
+
# C extensions
|
7 |
+
*.so
|
8 |
+
|
9 |
+
# Distribution / packaging
|
10 |
+
.Python
|
11 |
+
build/
|
12 |
+
develop-eggs/
|
13 |
+
dist/
|
14 |
+
downloads/
|
15 |
+
eggs/
|
16 |
+
.eggs/
|
17 |
+
lib/
|
18 |
+
lib64/
|
19 |
+
parts/
|
20 |
+
sdist/
|
21 |
+
var/
|
22 |
+
wheels/
|
23 |
+
share/python-wheels/
|
24 |
+
*.egg-info/
|
25 |
+
.installed.cfg
|
26 |
+
*.egg
|
27 |
+
MANIFEST
|
28 |
+
|
29 |
+
# PyInstaller
|
30 |
+
# Usually these files are written by a python script from a template
|
31 |
+
# before PyInstaller builds the exe, so as to inject date/other infos into it.
|
32 |
+
*.manifest
|
33 |
+
*.spec
|
34 |
+
|
35 |
+
# Installer logs
|
36 |
+
pip-log.txt
|
37 |
+
pip-delete-this-directory.txt
|
38 |
+
|
39 |
+
# Unit test / coverage reports
|
40 |
+
htmlcov/
|
41 |
+
.tox/
|
42 |
+
.nox/
|
43 |
+
.coverage
|
44 |
+
.coverage.*
|
45 |
+
.cache
|
46 |
+
nosetests.xml
|
47 |
+
coverage.xml
|
48 |
+
*.cover
|
49 |
+
*.py,cover
|
50 |
+
.hypothesis/
|
51 |
+
.pytest_cache/
|
52 |
+
cover/
|
53 |
+
|
54 |
+
# Translations
|
55 |
+
*.mo
|
56 |
+
*.pot
|
57 |
+
|
58 |
+
# Django stuff:
|
59 |
+
*.log
|
60 |
+
local_settings.py
|
61 |
+
db.sqlite3
|
62 |
+
db.sqlite3-journal
|
63 |
+
|
64 |
+
# Flask stuff:
|
65 |
+
instance/
|
66 |
+
.webassets-cache
|
67 |
+
|
68 |
+
# Scrapy stuff:
|
69 |
+
.scrapy
|
70 |
+
|
71 |
+
# Sphinx documentation
|
72 |
+
docs/_build/
|
73 |
+
|
74 |
+
# PyBuilder
|
75 |
+
.pybuilder/
|
76 |
+
target/
|
77 |
+
|
78 |
+
# Jupyter Notebook
|
79 |
+
.ipynb_checkpoints
|
80 |
+
|
81 |
+
# IPython
|
82 |
+
profile_default/
|
83 |
+
ipython_config.py
|
84 |
+
|
85 |
+
# pyenv
|
86 |
+
# For a library or package, you might want to ignore these files since the code is
|
87 |
+
# intended to run in multiple environments; otherwise, check them in:
|
88 |
+
# .python-version
|
89 |
+
|
90 |
+
# pipenv
|
91 |
+
# According to pypa/pipenv#598, it is recommended to include Pipfile.lock in version control.
|
92 |
+
# However, in case of collaboration, if having platform-specific dependencies or dependencies
|
93 |
+
# having no cross-platform support, pipenv may install dependencies that don't work, or not
|
94 |
+
# install all needed dependencies.
|
95 |
+
#Pipfile.lock
|
96 |
+
|
97 |
+
# UV
|
98 |
+
# Similar to Pipfile.lock, it is generally recommended to include uv.lock in version control.
|
99 |
+
# This is especially recommended for binary packages to ensure reproducibility, and is more
|
100 |
+
# commonly ignored for libraries.
|
101 |
+
#uv.lock
|
102 |
+
|
103 |
+
# poetry
|
104 |
+
# Similar to Pipfile.lock, it is generally recommended to include poetry.lock in version control.
|
105 |
+
# This is especially recommended for binary packages to ensure reproducibility, and is more
|
106 |
+
# commonly ignored for libraries.
|
107 |
+
# https://python-poetry.org/docs/basic-usage/#commit-your-poetrylock-file-to-version-control
|
108 |
+
#poetry.lock
|
109 |
+
|
110 |
+
# pdm
|
111 |
+
# Similar to Pipfile.lock, it is generally recommended to include pdm.lock in version control.
|
112 |
+
#pdm.lock
|
113 |
+
# pdm stores project-wide configurations in .pdm.toml, but it is recommended to not include it
|
114 |
+
# in version control.
|
115 |
+
# https://pdm.fming.dev/latest/usage/project/#working-with-version-control
|
116 |
+
.pdm.toml
|
117 |
+
.pdm-python
|
118 |
+
.pdm-build/
|
119 |
+
|
120 |
+
# PEP 582; used by e.g. github.com/David-OConnor/pyflow and github.com/pdm-project/pdm
|
121 |
+
__pypackages__/
|
122 |
+
|
123 |
+
# Celery stuff
|
124 |
+
celerybeat-schedule
|
125 |
+
celerybeat.pid
|
126 |
+
|
127 |
+
# SageMath parsed files
|
128 |
+
*.sage.py
|
129 |
+
|
130 |
+
# Environments
|
131 |
+
.env
|
132 |
+
.venv
|
133 |
+
env/
|
134 |
+
venv/
|
135 |
+
ENV/
|
136 |
+
env.bak/
|
137 |
+
venv.bak/
|
138 |
+
|
139 |
+
# Spyder project settings
|
140 |
+
.spyderproject
|
141 |
+
.spyproject
|
142 |
+
|
143 |
+
# Rope project settings
|
144 |
+
.ropeproject
|
145 |
+
|
146 |
+
# mkdocs documentation
|
147 |
+
/site
|
148 |
+
|
149 |
+
# mypy
|
150 |
+
.mypy_cache/
|
151 |
+
.dmypy.json
|
152 |
+
dmypy.json
|
153 |
+
|
154 |
+
# Pyre type checker
|
155 |
+
.pyre/
|
156 |
+
|
157 |
+
# pytype static type analyzer
|
158 |
+
.pytype/
|
159 |
+
|
160 |
+
# Cython debug symbols
|
161 |
+
cython_debug/
|
162 |
+
|
163 |
+
# PyCharm
|
164 |
+
# JetBrains specific template is maintained in a separate JetBrains.gitignore that can
|
165 |
+
# be found at https://github.com/github/gitignore/blob/main/Global/JetBrains.gitignore
|
166 |
+
# and can be added to the global gitignore or merged into this file. For a more nuclear
|
167 |
+
# option (not recommended) you can uncomment the following to ignore the entire idea folder.
|
168 |
+
#.idea/
|
169 |
+
|
170 |
+
# PyPI configuration file
|
171 |
+
.pypirc
|
172 |
+
|
173 |
+
.gradio
|
174 |
+
.hugging_face
|
README.md
CHANGED
@@ -1,12 +1,8 @@
|
|
1 |
---
|
2 |
-
title: Q-
|
3 |
-
|
4 |
-
colorFrom: purple
|
5 |
-
colorTo: blue
|
6 |
sdk: gradio
|
7 |
sdk_version: 5.19.0
|
8 |
-
app_file: app.py
|
9 |
-
pinned: false
|
10 |
---
|
11 |
-
|
12 |
-
|
|
|
1 |
---
|
2 |
+
title: Q-Learning_GridWorld_Simulator
|
3 |
+
app_file: rl_gradio.py
|
|
|
|
|
4 |
sdk: gradio
|
5 |
sdk_version: 5.19.0
|
|
|
|
|
6 |
---
|
7 |
+
# Reinforcement_Learning_Project
|
8 |
+
Simple Project to enforce learning by Q-Learning
|
requirements.txt
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
gradio==5.19.0
|
2 |
+
matplotlib==3.10.1
|
3 |
+
numpy==2.2.3
|
rl_gradio.py
ADDED
@@ -0,0 +1,564 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
import numpy as np
|
2 |
+
import matplotlib.pyplot as plt
|
3 |
+
import gradio as gr
|
4 |
+
import time
|
5 |
+
from matplotlib.colors import ListedColormap
|
6 |
+
import matplotlib.patches as patches
|
7 |
+
|
8 |
+
class GridWorld:
|
9 |
+
"""A simple grid world environment with obstacles."""
|
10 |
+
|
11 |
+
def __init__(self, height=4, width=4):
|
12 |
+
# Grid dimensions
|
13 |
+
self.height = height
|
14 |
+
self.width = width
|
15 |
+
|
16 |
+
# Define states
|
17 |
+
self.n_states = self.height * self.width
|
18 |
+
|
19 |
+
# Actions: 0: up, 1: right, 2: down, 3: left
|
20 |
+
self.n_actions = 4
|
21 |
+
self.action_names = ['Up', 'Right', 'Down', 'Left']
|
22 |
+
|
23 |
+
# Define rewards
|
24 |
+
self.rewards = np.zeros((self.height, self.width))
|
25 |
+
# Goal state
|
26 |
+
self.rewards[self.height-1, self.width-1] = 1.0
|
27 |
+
# Obstacles (negative reward)
|
28 |
+
self.obstacles = []
|
29 |
+
if height >= 4 and width >= 4:
|
30 |
+
self.rewards[1, 1] = -1.0
|
31 |
+
self.rewards[1, 2] = -1.0
|
32 |
+
self.rewards[2, 1] = -1.0
|
33 |
+
self.obstacles = [(1, 1), (1, 2), (2, 1)]
|
34 |
+
|
35 |
+
# Start state
|
36 |
+
self.start_state = (0, 0)
|
37 |
+
|
38 |
+
# Goal state
|
39 |
+
self.goal_state = (self.height-1, self.width-1)
|
40 |
+
|
41 |
+
# Reset the environment
|
42 |
+
self.reset()
|
43 |
+
|
44 |
+
def reset(self):
|
45 |
+
"""Reset the agent to the start state."""
|
46 |
+
self.agent_position = self.start_state
|
47 |
+
return self._get_state()
|
48 |
+
|
49 |
+
def _get_state(self):
|
50 |
+
"""Convert the agent's (row, col) position to a state number."""
|
51 |
+
row, col = self.agent_position
|
52 |
+
return row * self.width + col
|
53 |
+
|
54 |
+
def _get_pos_from_state(self, state):
|
55 |
+
"""Convert a state number to (row, col) position."""
|
56 |
+
row = state // self.width
|
57 |
+
col = state % self.width
|
58 |
+
return (row, col)
|
59 |
+
|
60 |
+
def step(self, action):
|
61 |
+
"""Take an action and return next_state, reward, done."""
|
62 |
+
row, col = self.agent_position
|
63 |
+
|
64 |
+
# Apply the action
|
65 |
+
if action == 0: # up
|
66 |
+
row = max(0, row - 1)
|
67 |
+
elif action == 1: # right
|
68 |
+
col = min(self.width - 1, col + 1)
|
69 |
+
elif action == 2: # down
|
70 |
+
row = min(self.height - 1, row + 1)
|
71 |
+
elif action == 3: # left
|
72 |
+
col = max(0, col - 1)
|
73 |
+
|
74 |
+
# Update agent position
|
75 |
+
self.agent_position = (row, col)
|
76 |
+
|
77 |
+
# Get reward
|
78 |
+
reward = self.rewards[row, col]
|
79 |
+
|
80 |
+
# Check if episode is done
|
81 |
+
done = (row, col) == self.goal_state
|
82 |
+
|
83 |
+
return self._get_state(), reward, done
|
84 |
+
|
85 |
+
class QLearningAgent:
|
86 |
+
"""A simple Q-learning agent."""
|
87 |
+
|
88 |
+
def __init__(self, n_states, n_actions, learning_rate=0.1, discount_factor=0.9, exploration_rate=1.0, exploration_decay=0.995):
|
89 |
+
"""Initialize the Q-learning agent."""
|
90 |
+
self.n_states = n_states
|
91 |
+
self.n_actions = n_actions
|
92 |
+
self.learning_rate = learning_rate
|
93 |
+
self.discount_factor = discount_factor
|
94 |
+
self.exploration_rate = exploration_rate
|
95 |
+
self.exploration_decay = exploration_decay
|
96 |
+
|
97 |
+
# Initialize Q-table
|
98 |
+
self.q_table = np.zeros((n_states, n_actions))
|
99 |
+
|
100 |
+
# Track visited states for visualization
|
101 |
+
self.visit_counts = np.zeros(n_states)
|
102 |
+
|
103 |
+
# Training metrics
|
104 |
+
self.rewards_history = []
|
105 |
+
self.exploration_rates = []
|
106 |
+
|
107 |
+
def select_action(self, state):
|
108 |
+
"""Select an action using epsilon-greedy policy."""
|
109 |
+
if np.random.random() < self.exploration_rate:
|
110 |
+
# Explore: select a random action
|
111 |
+
return np.random.randint(self.n_actions)
|
112 |
+
else:
|
113 |
+
# Exploit: select the action with the highest Q-value
|
114 |
+
return np.argmax(self.q_table[state])
|
115 |
+
|
116 |
+
def update(self, state, action, reward, next_state, done):
|
117 |
+
"""Update the Q-table using the Q-learning update rule."""
|
118 |
+
# Calculate the Q-target
|
119 |
+
if done:
|
120 |
+
q_target = reward
|
121 |
+
else:
|
122 |
+
q_target = reward + self.discount_factor * np.max(self.q_table[next_state])
|
123 |
+
|
124 |
+
# Update the Q-value
|
125 |
+
self.q_table[state, action] += self.learning_rate * (q_target - self.q_table[state, action])
|
126 |
+
|
127 |
+
# Update visit count for visualization
|
128 |
+
self.visit_counts[state] += 1
|
129 |
+
|
130 |
+
def decay_exploration(self):
|
131 |
+
"""Decay the exploration rate."""
|
132 |
+
self.exploration_rate *= self.exploration_decay
|
133 |
+
self.exploration_rates.append(self.exploration_rate)
|
134 |
+
|
135 |
+
def get_policy(self):
|
136 |
+
"""Return the current greedy policy."""
|
137 |
+
return np.argmax(self.q_table, axis=1)
|
138 |
+
|
139 |
+
def reset(self):
|
140 |
+
"""Reset the agent for a new training session."""
|
141 |
+
self.q_table = np.zeros((self.n_states, self.n_actions))
|
142 |
+
self.visit_counts = np.zeros(self.n_states)
|
143 |
+
self.rewards_history = []
|
144 |
+
self.exploration_rates = []
|
145 |
+
|
146 |
+
|
147 |
+
def create_gridworld_figure(env, agent, episode_count=0, total_reward=0):
|
148 |
+
"""Create a figure with environment, visit heatmap, and Q-values."""
|
149 |
+
fig, axes = plt.subplots(1, 3, figsize=(15, 5))
|
150 |
+
fig.suptitle(f"Episode: {episode_count}, Total Reward: {total_reward:.2f}, Exploration Rate: {agent.exploration_rate:.2f}")
|
151 |
+
|
152 |
+
# Define colors for different cell types
|
153 |
+
colors = {
|
154 |
+
'empty': 'white',
|
155 |
+
'obstacle': 'black',
|
156 |
+
'goal': 'green',
|
157 |
+
'start': 'blue',
|
158 |
+
'agent': 'red'
|
159 |
+
}
|
160 |
+
|
161 |
+
# Helper function to draw grid
|
162 |
+
def draw_grid(ax):
|
163 |
+
# Create a grid
|
164 |
+
for i in range(env.height + 1):
|
165 |
+
ax.axhline(i, color='black', lw=1)
|
166 |
+
for j in range(env.width + 1):
|
167 |
+
ax.axvline(j, color='black', lw=1)
|
168 |
+
|
169 |
+
# Set limits and remove ticks
|
170 |
+
ax.set_xlim(0, env.width)
|
171 |
+
ax.set_ylim(0, env.height)
|
172 |
+
ax.invert_yaxis() # Invert y-axis to match grid coordinates
|
173 |
+
ax.set_xticks(np.arange(0.5, env.width, 1))
|
174 |
+
ax.set_yticks(np.arange(0.5, env.height, 1))
|
175 |
+
ax.set_xticklabels(range(env.width))
|
176 |
+
ax.set_yticklabels(range(env.height))
|
177 |
+
|
178 |
+
# Helper function to draw a cell
|
179 |
+
def draw_cell(ax, row, col, cell_type):
|
180 |
+
color = colors.get(cell_type, 'white')
|
181 |
+
rect = patches.Rectangle((col, row), 1, 1, linewidth=1, edgecolor='black', facecolor=color, alpha=0.7)
|
182 |
+
ax.add_patch(rect)
|
183 |
+
|
184 |
+
# Helper function to draw an arrow
|
185 |
+
def draw_arrow(ax, row, col, action):
|
186 |
+
# Coordinates for arrows
|
187 |
+
arrow_starts = {
|
188 |
+
0: (col + 0.5, row + 0.7), # up
|
189 |
+
1: (col + 0.3, row + 0.5), # right
|
190 |
+
2: (col + 0.5, row + 0.3), # down
|
191 |
+
3: (col + 0.7, row + 0.5) # left
|
192 |
+
}
|
193 |
+
|
194 |
+
arrow_ends = {
|
195 |
+
0: (col + 0.5, row + 0.3), # up
|
196 |
+
1: (col + 0.7, row + 0.5), # right
|
197 |
+
2: (col + 0.5, row + 0.7), # down
|
198 |
+
3: (col + 0.3, row + 0.5) # left
|
199 |
+
}
|
200 |
+
|
201 |
+
ax.annotate('', xy=arrow_ends[action], xytext=arrow_starts[action],
|
202 |
+
arrowprops=dict(arrowstyle='->', lw=2, color='blue'))
|
203 |
+
|
204 |
+
# Draw Environment
|
205 |
+
ax = axes[0]
|
206 |
+
ax.set_title('GridWorld Environment')
|
207 |
+
draw_grid(ax)
|
208 |
+
|
209 |
+
# Draw cells
|
210 |
+
for i in range(env.height):
|
211 |
+
for j in range(env.width):
|
212 |
+
if (i, j) in env.obstacles:
|
213 |
+
draw_cell(ax, i, j, 'obstacle')
|
214 |
+
elif (i, j) == env.goal_state:
|
215 |
+
draw_cell(ax, i, j, 'goal')
|
216 |
+
elif (i, j) == env.start_state:
|
217 |
+
draw_cell(ax, i, j, 'start')
|
218 |
+
|
219 |
+
# Draw agent
|
220 |
+
row, col = env.agent_position
|
221 |
+
draw_cell(ax, row, col, 'agent')
|
222 |
+
|
223 |
+
# Draw policy arrows
|
224 |
+
policy = agent.get_policy()
|
225 |
+
for state in range(env.n_states):
|
226 |
+
row, col = env._get_pos_from_state(state)
|
227 |
+
if (row, col) not in env.obstacles and (row, col) != env.goal_state:
|
228 |
+
draw_arrow(ax, row, col, policy[state])
|
229 |
+
|
230 |
+
# Ensure proper aspect ratio
|
231 |
+
ax.set_aspect('equal')
|
232 |
+
|
233 |
+
# Draw Visit Heatmap
|
234 |
+
ax = axes[1]
|
235 |
+
ax.set_title('State Visitation Heatmap')
|
236 |
+
draw_grid(ax)
|
237 |
+
|
238 |
+
# Create heatmap data
|
239 |
+
heatmap_data = np.zeros((env.height, env.width))
|
240 |
+
for state in range(env.n_states):
|
241 |
+
row, col = env._get_pos_from_state(state)
|
242 |
+
heatmap_data[row, col] = agent.visit_counts[state]
|
243 |
+
|
244 |
+
# Normalize values for coloring
|
245 |
+
max_visits = max(1, np.max(heatmap_data))
|
246 |
+
|
247 |
+
# Draw heatmap
|
248 |
+
for i in range(env.height):
|
249 |
+
for j in range(env.width):
|
250 |
+
if (i, j) in env.obstacles:
|
251 |
+
draw_cell(ax, i, j, 'obstacle')
|
252 |
+
elif (i, j) == env.goal_state:
|
253 |
+
draw_cell(ax, i, j, 'goal')
|
254 |
+
else:
|
255 |
+
intensity = heatmap_data[i, j] / max_visits
|
256 |
+
color = plt.cm.viridis(intensity)
|
257 |
+
rect = patches.Rectangle((j, i), 1, 1, linewidth=1, edgecolor='black', facecolor=color, alpha=0.7)
|
258 |
+
ax.add_patch(rect)
|
259 |
+
# Add visit count text
|
260 |
+
if heatmap_data[i, j] > 0:
|
261 |
+
ax.text(j + 0.5, i + 0.5, int(heatmap_data[i, j]), ha='center', va='center', color='white' if intensity > 0.5 else 'black')
|
262 |
+
|
263 |
+
# Ensure proper aspect ratio
|
264 |
+
ax.set_aspect('equal')
|
265 |
+
|
266 |
+
# Draw Q-values
|
267 |
+
ax = axes[2]
|
268 |
+
ax.set_title('Q-Values')
|
269 |
+
draw_grid(ax)
|
270 |
+
|
271 |
+
# Draw Q-values for each cell
|
272 |
+
for state in range(env.n_states):
|
273 |
+
row, col = env._get_pos_from_state(state)
|
274 |
+
|
275 |
+
if (row, col) in env.obstacles:
|
276 |
+
draw_cell(ax, row, col, 'obstacle')
|
277 |
+
continue
|
278 |
+
|
279 |
+
if (row, col) == env.goal_state:
|
280 |
+
draw_cell(ax, row, col, 'goal')
|
281 |
+
continue
|
282 |
+
|
283 |
+
# Calculate q-values for each action
|
284 |
+
q_values = agent.q_table[state]
|
285 |
+
|
286 |
+
# Draw arrows proportional to Q-values
|
287 |
+
for action in range(env.n_actions):
|
288 |
+
q_value = q_values[action]
|
289 |
+
|
290 |
+
# Only draw arrows for positive Q-values
|
291 |
+
if q_value > 0:
|
292 |
+
# Normalize arrow size
|
293 |
+
max_q = max(0.1, np.max(q_values))
|
294 |
+
arrow_size = 0.3 * (q_value / max_q)
|
295 |
+
|
296 |
+
# Position calculations
|
297 |
+
center_x = col + 0.5
|
298 |
+
center_y = row + 0.5
|
299 |
+
|
300 |
+
# Direction vectors
|
301 |
+
directions = [
|
302 |
+
(0, -arrow_size), # up
|
303 |
+
(arrow_size, 0), # right
|
304 |
+
(0, arrow_size), # down
|
305 |
+
(-arrow_size, 0) # left
|
306 |
+
]
|
307 |
+
|
308 |
+
dx, dy = directions[action]
|
309 |
+
|
310 |
+
# Draw arrow
|
311 |
+
ax.arrow(center_x, center_y, dx, dy, head_width=0.1, head_length=0.1,
|
312 |
+
fc='blue', ec='blue', alpha=0.7)
|
313 |
+
|
314 |
+
# Add Q-value text
|
315 |
+
text_positions = [
|
316 |
+
(center_x, center_y - 0.25), # up
|
317 |
+
(center_x + 0.25, center_y), # right
|
318 |
+
(center_x, center_y + 0.25), # down
|
319 |
+
(center_x - 0.25, center_y) # left
|
320 |
+
]
|
321 |
+
|
322 |
+
tx, ty = text_positions[action]
|
323 |
+
ax.text(tx, ty, f"{q_value:.2f}", ha='center', va='center', fontsize=8,
|
324 |
+
bbox=dict(facecolor='white', alpha=0.7, boxstyle='round,pad=0.1'))
|
325 |
+
|
326 |
+
# Ensure proper aspect ratio
|
327 |
+
ax.set_aspect('equal')
|
328 |
+
|
329 |
+
plt.tight_layout()
|
330 |
+
return fig
|
331 |
+
|
332 |
+
def create_metrics_figure(agent):
|
333 |
+
"""Create a figure with training metrics."""
|
334 |
+
fig, axes = plt.subplots(1, 2, figsize=(12, 4))
|
335 |
+
|
336 |
+
# Plot rewards
|
337 |
+
if agent.rewards_history:
|
338 |
+
axes[0].plot(agent.rewards_history)
|
339 |
+
axes[0].set_title('Rewards per Episode')
|
340 |
+
axes[0].set_xlabel('Episode')
|
341 |
+
axes[0].set_ylabel('Total Reward')
|
342 |
+
axes[0].grid(True)
|
343 |
+
else:
|
344 |
+
axes[0].set_title('No reward data yet')
|
345 |
+
|
346 |
+
# Plot exploration rate
|
347 |
+
if agent.exploration_rates:
|
348 |
+
axes[1].plot(agent.exploration_rates)
|
349 |
+
axes[1].set_title('Exploration Rate Decay')
|
350 |
+
axes[1].set_xlabel('Episode')
|
351 |
+
axes[1].set_ylabel('Exploration Rate (ε)')
|
352 |
+
axes[1].grid(True)
|
353 |
+
else:
|
354 |
+
axes[1].set_title('No exploration rate data yet')
|
355 |
+
|
356 |
+
plt.tight_layout()
|
357 |
+
return fig
|
358 |
+
|
359 |
+
def train_single_episode(env, agent):
|
360 |
+
"""Train for a single episode and return the total reward."""
|
361 |
+
state = env.reset()
|
362 |
+
total_reward = 0
|
363 |
+
done = False
|
364 |
+
steps = 0
|
365 |
+
max_steps = env.width * env.height * 3 # Prevent infinite loops
|
366 |
+
|
367 |
+
while not done and steps < max_steps:
|
368 |
+
# Select action
|
369 |
+
action = agent.select_action(state)
|
370 |
+
|
371 |
+
# Take the action
|
372 |
+
next_state, reward, done = env.step(action)
|
373 |
+
|
374 |
+
# Update the Q-table
|
375 |
+
agent.update(state, action, reward, next_state, done)
|
376 |
+
|
377 |
+
# Update state and total reward
|
378 |
+
state = next_state
|
379 |
+
total_reward += reward
|
380 |
+
steps += 1
|
381 |
+
|
382 |
+
# Decay exploration rate
|
383 |
+
agent.decay_exploration()
|
384 |
+
|
385 |
+
# Store the total reward
|
386 |
+
agent.rewards_history.append(total_reward)
|
387 |
+
|
388 |
+
return total_reward
|
389 |
+
|
390 |
+
def train_agent(env, agent, episodes, progress=gr.Progress()):
|
391 |
+
"""Train the agent for a specified number of episodes."""
|
392 |
+
progress_text = ""
|
393 |
+
progress(0, desc="Starting training...")
|
394 |
+
|
395 |
+
for episode in progress.tqdm(range(episodes)):
|
396 |
+
total_reward = train_single_episode(env, agent)
|
397 |
+
|
398 |
+
if (episode + 1) % 10 == 0 or episode == episodes - 1:
|
399 |
+
progress_text += f"Episode {episode + 1}/{episodes}, Reward: {total_reward}, Exploration: {agent.exploration_rate:.3f}\n"
|
400 |
+
|
401 |
+
# Create final visualization
|
402 |
+
env_fig = create_gridworld_figure(env, agent, episode_count=episodes, total_reward=total_reward)
|
403 |
+
metrics_fig = create_metrics_figure(agent)
|
404 |
+
|
405 |
+
return env_fig, metrics_fig, progress_text
|
406 |
+
|
407 |
+
def run_test_episode(env, agent):
|
408 |
+
"""Run a test episode using the learned policy."""
|
409 |
+
state = env.reset()
|
410 |
+
total_reward = 0
|
411 |
+
done = False
|
412 |
+
path = [env._get_pos_from_state(state)]
|
413 |
+
steps = 0
|
414 |
+
max_steps = env.width * env.height * 3 # Prevent infinite loops
|
415 |
+
|
416 |
+
while not done and steps < max_steps:
|
417 |
+
# Select the best action from the learned policy
|
418 |
+
action = np.argmax(agent.q_table[state])
|
419 |
+
|
420 |
+
# Take the action
|
421 |
+
next_state, reward, done = env.step(action)
|
422 |
+
|
423 |
+
# Update state and total reward
|
424 |
+
state = next_state
|
425 |
+
total_reward += reward
|
426 |
+
path.append(env._get_pos_from_state(state))
|
427 |
+
steps += 1
|
428 |
+
|
429 |
+
# Create visualization
|
430 |
+
env_fig = create_gridworld_figure(env, agent, episode_count="Test", total_reward=total_reward)
|
431 |
+
|
432 |
+
# Format path for display
|
433 |
+
path_text = "Path taken:\n"
|
434 |
+
for i, pos in enumerate(path):
|
435 |
+
path_text += f"Step {i}: {pos}\n"
|
436 |
+
|
437 |
+
return env_fig, path_text, f"Test completed with total reward: {total_reward}"
|
438 |
+
|
439 |
+
def create_ui():
|
440 |
+
"""Create the Gradio interface."""
|
441 |
+
# Create environment and agent
|
442 |
+
env = GridWorld(height=4, width=4)
|
443 |
+
agent = QLearningAgent(
|
444 |
+
n_states=env.n_states,
|
445 |
+
n_actions=env.n_actions,
|
446 |
+
learning_rate=0.1,
|
447 |
+
discount_factor=0.9,
|
448 |
+
exploration_rate=1.0,
|
449 |
+
exploration_decay=0.995
|
450 |
+
)
|
451 |
+
|
452 |
+
# Create initial visualizations
|
453 |
+
init_env_fig = create_gridworld_figure(env, agent)
|
454 |
+
init_metrics_fig = create_metrics_figure(agent)
|
455 |
+
|
456 |
+
with gr.Blocks(title="Q-Learning GridWorld Simulator") as demo:
|
457 |
+
gr.Markdown("# Q-Learning GridWorld Simulator")
|
458 |
+
|
459 |
+
with gr.Tab("Environment Setup"):
|
460 |
+
with gr.Row():
|
461 |
+
with gr.Column():
|
462 |
+
grid_height = gr.Slider(minimum=3, maximum=8, value=4, step=1, label="Grid Height")
|
463 |
+
grid_width = gr.Slider(minimum=3, maximum=8, value=4, step=1, label="Grid Width")
|
464 |
+
setup_btn = gr.Button("Setup Environment")
|
465 |
+
|
466 |
+
env_display = gr.Plot(value=init_env_fig, label="Environment")
|
467 |
+
|
468 |
+
with gr.Row():
|
469 |
+
setup_info = gr.Textbox(label="Environment Info", value="4x4 GridWorld with start at (0,0) and goal at (3,3)")
|
470 |
+
|
471 |
+
with gr.Tab("Train Agent"):
|
472 |
+
with gr.Row():
|
473 |
+
with gr.Column():
|
474 |
+
learning_rate = gr.Slider(minimum=0.01, maximum=1.0, value=0.1, step=0.01, label="Learning Rate (α)")
|
475 |
+
discount_factor = gr.Slider(minimum=0.1, maximum=1.0, value=0.9, step=0.01, label="Discount Factor (γ)")
|
476 |
+
exploration_rate = gr.Slider(minimum=0.1, maximum=1.0, value=1.0, step=0.01, label="Initial Exploration Rate (ε)")
|
477 |
+
exploration_decay = gr.Slider(minimum=0.9, maximum=0.999, value=0.995, step=0.001, label="Exploration Decay Rate")
|
478 |
+
episodes = gr.Slider(minimum=10, maximum=500, value=100, step=10, label="Number of Episodes")
|
479 |
+
train_btn = gr.Button("Train Agent")
|
480 |
+
|
481 |
+
with gr.Row():
|
482 |
+
train_env_display = gr.Plot(label="Training Environment")
|
483 |
+
train_metrics_display = gr.Plot(label="Training Metrics")
|
484 |
+
|
485 |
+
train_log = gr.Textbox(label="Training Log", lines=10)
|
486 |
+
|
487 |
+
with gr.Tab("Test Agent"):
|
488 |
+
with gr.Row():
|
489 |
+
test_btn = gr.Button("Test Trained Agent")
|
490 |
+
|
491 |
+
with gr.Row():
|
492 |
+
test_env_display = gr.Plot(label="Test Environment")
|
493 |
+
|
494 |
+
with gr.Row():
|
495 |
+
with gr.Column():
|
496 |
+
path_display = gr.Textbox(label="Path Taken", lines=10)
|
497 |
+
test_result = gr.Textbox(label="Test Result")
|
498 |
+
|
499 |
+
# Setup environment callback
|
500 |
+
def setup_environment(height, width):
|
501 |
+
nonlocal env, agent
|
502 |
+
env = GridWorld(height=int(height), width=int(width))
|
503 |
+
agent = QLearningAgent(
|
504 |
+
n_states=env.n_states,
|
505 |
+
n_actions=env.n_actions,
|
506 |
+
learning_rate=0.1,
|
507 |
+
discount_factor=0.9,
|
508 |
+
exploration_rate=1.0,
|
509 |
+
exploration_decay=0.995
|
510 |
+
)
|
511 |
+
env_fig = create_gridworld_figure(env, agent)
|
512 |
+
info_text = f"{height}x{width} GridWorld with start at (0,0) and goal at ({height-1},{width-1})"
|
513 |
+
if env.obstacles:
|
514 |
+
info_text += f"\nObstacles at: {env.obstacles}"
|
515 |
+
return env_fig, info_text
|
516 |
+
|
517 |
+
setup_btn.click(
|
518 |
+
setup_environment,
|
519 |
+
inputs=[grid_height, grid_width],
|
520 |
+
outputs=[env_display, setup_info]
|
521 |
+
)
|
522 |
+
|
523 |
+
# Train agent callback
|
524 |
+
def start_training(lr, df, er, ed, eps):
|
525 |
+
nonlocal env, agent
|
526 |
+
agent = QLearningAgent(
|
527 |
+
n_states=env.n_states,
|
528 |
+
n_actions=env.n_actions,
|
529 |
+
learning_rate=float(lr),
|
530 |
+
discount_factor=float(df),
|
531 |
+
exploration_rate=float(er),
|
532 |
+
exploration_decay=float(ed)
|
533 |
+
)
|
534 |
+
env_fig, metrics_fig, log = train_agent(env, agent, int(eps))
|
535 |
+
return env_fig, metrics_fig, log
|
536 |
+
|
537 |
+
train_btn.click(
|
538 |
+
start_training,
|
539 |
+
inputs=[learning_rate, discount_factor, exploration_rate, exploration_decay, episodes],
|
540 |
+
outputs=[train_env_display, train_metrics_display, train_log]
|
541 |
+
)
|
542 |
+
|
543 |
+
# Test agent callback
|
544 |
+
def test_trained_agent():
|
545 |
+
nonlocal env, agent
|
546 |
+
env_fig, path_text, result = run_test_episode(env, agent)
|
547 |
+
return env_fig, path_text, result
|
548 |
+
|
549 |
+
test_btn.click(
|
550 |
+
test_trained_agent,
|
551 |
+
inputs=[],
|
552 |
+
outputs=[test_env_display, path_display, test_result]
|
553 |
+
)
|
554 |
+
|
555 |
+
return demo
|
556 |
+
|
557 |
+
if __name__ == "__main__":
|
558 |
+
# Install required packages
|
559 |
+
# !pip install gradio matplotlib numpy
|
560 |
+
|
561 |
+
# Create and launch the UI
|
562 |
+
demo = create_ui()
|
563 |
+
demo.launch(share=True)
|
564 |
+
|