vimmoos@Thor
commited on
Commit
·
c6a28ec
1
Parent(s):
62f50f1
base gym env
Browse files- app.py +276 -0
- old_code/experiment_3/q_networks/buffers/CartPole-v0/1/DQN/memory_buffer.p +0 -0
- poetry.lock +0 -0
- pyproject.toml +1 -0
- udrl/__main__.py +3 -3
- udrl/agent.py +3 -3
- udrl/inference.py +2 -2
- udrl/plot.py +2 -2
- udrl/viz.py +2 -2
app.py
CHANGED
@@ -0,0 +1,276 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
import streamlit as st
|
2 |
+
import gymnasium as gym
|
3 |
+
import numpy as np
|
4 |
+
from PIL import Image
|
5 |
+
import time
|
6 |
+
|
7 |
+
# Initialize session state variables if they don't exist
|
8 |
+
if "env" not in st.session_state:
|
9 |
+
st.session_state.env = gym.make("LunarLander-v2", render_mode="rgb_array")
|
10 |
+
st.session_state.env.reset()
|
11 |
+
st.session_state.frame = st.session_state.env.render()
|
12 |
+
if "paused" not in st.session_state:
|
13 |
+
st.session_state.paused = False
|
14 |
+
|
15 |
+
|
16 |
+
# Function to reset the environment
|
17 |
+
def reset_environment():
|
18 |
+
st.session_state.env.reset()
|
19 |
+
|
20 |
+
|
21 |
+
# Function to toggle pause state
|
22 |
+
def toggle_pause():
|
23 |
+
st.session_state.paused = not st.session_state.paused
|
24 |
+
|
25 |
+
|
26 |
+
# Create the Streamlit app
|
27 |
+
st.title("Gymnasium Environment Viewer")
|
28 |
+
|
29 |
+
# Add control buttons in a horizontal layout
|
30 |
+
col1, col2 = st.columns(2)
|
31 |
+
with col1:
|
32 |
+
st.button("Reset Environment", on_click=reset_environment)
|
33 |
+
with col2:
|
34 |
+
if st.session_state.paused:
|
35 |
+
st.button("Resume", on_click=toggle_pause)
|
36 |
+
else:
|
37 |
+
st.button("Pause", on_click=toggle_pause)
|
38 |
+
|
39 |
+
# Create a placeholder for the image
|
40 |
+
image_placeholder = st.empty()
|
41 |
+
|
42 |
+
# Create a container for environment info
|
43 |
+
sidebar_container = st.sidebar.container()
|
44 |
+
|
45 |
+
# Main simulation loop using rerun
|
46 |
+
if not st.session_state.paused:
|
47 |
+
# Take a random action
|
48 |
+
action = st.session_state.env.action_space.sample()
|
49 |
+
observation, reward, terminated, truncated, info = (
|
50 |
+
st.session_state.env.step(action)
|
51 |
+
)
|
52 |
+
|
53 |
+
# Render the environment
|
54 |
+
st.session_state.frame = st.session_state.env.render()
|
55 |
+
|
56 |
+
# Reset if the episode is done
|
57 |
+
if terminated or truncated:
|
58 |
+
st.session_state.env.reset()
|
59 |
+
# Display the frame
|
60 |
+
if st.session_state.paused:
|
61 |
+
image_placeholder.image(
|
62 |
+
st.session_state.frame,
|
63 |
+
caption="Environment Visualization (Paused)",
|
64 |
+
use_column_width=True,
|
65 |
+
)
|
66 |
+
else:
|
67 |
+
image_placeholder.image(
|
68 |
+
st.session_state.frame,
|
69 |
+
caption="Environment Visualization",
|
70 |
+
use_column_width=True,
|
71 |
+
)
|
72 |
+
|
73 |
+
# Display some information about the environment
|
74 |
+
with sidebar_container:
|
75 |
+
st.header("Environment Info")
|
76 |
+
st.write(f"Action Space: {st.session_state.env.action_space}")
|
77 |
+
st.write(f"Observation Space: {st.session_state.env.observation_space}")
|
78 |
+
|
79 |
+
# Add auto-refresh logic
|
80 |
+
if not st.session_state.paused:
|
81 |
+
time.sleep(0.1) # Add a small delay to control refresh rate
|
82 |
+
st.rerun()
|
83 |
+
|
84 |
+
# fig, ax = plt.subplots()
|
85 |
+
# ax.imshow(env.render())
|
86 |
+
# st.pyplot(fig)
|
87 |
+
# st.image(env.render())
|
88 |
+
|
89 |
+
|
90 |
+
# import gymnasium as gym
|
91 |
+
# import streamlit as st
|
92 |
+
# import numpy as np
|
93 |
+
# from udrl.policies import SklearnPolicy
|
94 |
+
# from udrl.agent import UpsideDownAgent, AgentHyper
|
95 |
+
# from pathlib import Path
|
96 |
+
|
97 |
+
# # import json
|
98 |
+
|
99 |
+
|
100 |
+
# def normalize_value(value, is_bounded, low=None, high=None):
|
101 |
+
# return (value - low) / (high - low)
|
102 |
+
|
103 |
+
|
104 |
+
# def visualize_environment(
|
105 |
+
# state,
|
106 |
+
# env,
|
107 |
+
# # paused,
|
108 |
+
# feature_importances,
|
109 |
+
# epoch,
|
110 |
+
# max_epoch=200,
|
111 |
+
# ):
|
112 |
+
|
113 |
+
# st.image(env.render())
|
114 |
+
# st.image(e)
|
115 |
+
# # Render the Gym environment
|
116 |
+
# # env_render = env.render()
|
117 |
+
|
118 |
+
# # # Display the rendered image using Streamlit
|
119 |
+
# # st.image(env_render, caption=f"Epoch {epoch}", use_column_width=True)
|
120 |
+
|
121 |
+
# # Display feature importances using Streamlit metrics
|
122 |
+
# # cols = st.columns(len(feature_importances))
|
123 |
+
# # for i, col in enumerate(cols):
|
124 |
+
# # col.metric(
|
125 |
+
# # label=f"Importance {i}", value=f"{feature_importances[i]:.2f}"
|
126 |
+
# # )
|
127 |
+
|
128 |
+
# # Create buttons using Streamlit
|
129 |
+
# # reset_button = st.button("Reset")
|
130 |
+
# # pause_play_button = st.button("Pause" if not paused else "Play")
|
131 |
+
# # next_button = st.button("Next")
|
132 |
+
# # save_button = st.button("Save")
|
133 |
+
|
134 |
+
# # return reset_button, pause_play_button, next_button, save_button
|
135 |
+
|
136 |
+
|
137 |
+
# def run_visualization(
|
138 |
+
# env_name,
|
139 |
+
# agent,
|
140 |
+
# init_desired_return,
|
141 |
+
# init_desired_horizon,
|
142 |
+
# max_epoch,
|
143 |
+
# base_path,
|
144 |
+
# ):
|
145 |
+
# # base_path = (
|
146 |
+
# # Path(base_path) / env_name / agent.policy.estimator.__str__()[:-2]
|
147 |
+
# # )
|
148 |
+
# # base_path.mkdir(parents=True, exist_ok=True)
|
149 |
+
# desired_return = init_desired_return
|
150 |
+
# desired_horizon = init_desired_horizon
|
151 |
+
|
152 |
+
# # Initialize the Gym environment
|
153 |
+
# env = gym.make(env_name, render_mode="rgb_array")
|
154 |
+
# state, _ = env.reset()
|
155 |
+
|
156 |
+
# epoch = 0
|
157 |
+
# # save_index = 0
|
158 |
+
|
159 |
+
# # paused = False
|
160 |
+
# # step = False
|
161 |
+
|
162 |
+
# # # Use Streamlit session state to manage paused state
|
163 |
+
# # if "paused" not in st.session_state:
|
164 |
+
# # st.session_state.paused = False
|
165 |
+
|
166 |
+
# while True:
|
167 |
+
# # Render and display the environment
|
168 |
+
# env_render = env.render()
|
169 |
+
# # if not st.session_state.pausedor step:
|
170 |
+
# command = np.array(
|
171 |
+
# [
|
172 |
+
# desired_return * agent.conf.return_scale,
|
173 |
+
# desired_horizon * agent.conf.horizon_scale,
|
174 |
+
# ]
|
175 |
+
# )
|
176 |
+
# command = np.expand_dims(command, axis=0)
|
177 |
+
# state = np.expand_dims(state, axis=0)
|
178 |
+
|
179 |
+
# action = agent.policy(state, command, True)
|
180 |
+
|
181 |
+
# ext_state = np.concatenate((state, command), axis=1)
|
182 |
+
|
183 |
+
# state, reward, done, truncated, info = env.step(action)
|
184 |
+
|
185 |
+
# feature_importances = {idx: [] for idx in range(ext_state.shape[1])}
|
186 |
+
|
187 |
+
# for t in agent.policy.estimator.estimators_:
|
188 |
+
# branch = np.array(t.decision_path(ext_state).todense(), dtype=bool)
|
189 |
+
# imp = t.tree_.impurity[branch[0]]
|
190 |
+
|
191 |
+
# for f, i in zip(
|
192 |
+
# t.tree_.feature[branch[0]][:-1], imp[:-1] - imp[1:]
|
193 |
+
# ):
|
194 |
+
# feature_importances.setdefault(f, []).append(i)
|
195 |
+
|
196 |
+
# # Line 8 Algorithm 2
|
197 |
+
# desired_return -= reward
|
198 |
+
# # Line 9 Algorithm 2
|
199 |
+
# desired_horizon = max(desired_horizon - 1, 1)
|
200 |
+
|
201 |
+
# summed_importances = [
|
202 |
+
# sum(feature_importances.get(k, [0.001]))
|
203 |
+
# for k in range(len(feature_importances.keys()))
|
204 |
+
# ]
|
205 |
+
|
206 |
+
# epoch += 1
|
207 |
+
# visualize_environment(
|
208 |
+
# state,
|
209 |
+
# env,
|
210 |
+
# # st.session_state.paused, # Use session state
|
211 |
+
# summed_importances,
|
212 |
+
# epoch,
|
213 |
+
# max_epoch,
|
214 |
+
# )
|
215 |
+
# # reset_button, pause_play_button, next_button, save_button = (
|
216 |
+
|
217 |
+
# # )
|
218 |
+
|
219 |
+
# if done or truncated:
|
220 |
+
# state, _ = env.reset()
|
221 |
+
# desired_horizon = init_desired_horizon
|
222 |
+
# desired_return = init_desired_return
|
223 |
+
# epoch = 0
|
224 |
+
|
225 |
+
# # step = False
|
226 |
+
|
227 |
+
# # Handle button clicks
|
228 |
+
# # if reset_button:
|
229 |
+
# # state, _ = env.reset()
|
230 |
+
# # desired_horizon = init_desired_horizon
|
231 |
+
# # desired_return = init_desired_return
|
232 |
+
# # epoch = 0
|
233 |
+
# # elif pause_play_button:
|
234 |
+
# # st.session_state.paused = (
|
235 |
+
# # not st.session_state.paused
|
236 |
+
# # ) # Toggle paused state
|
237 |
+
# # elif next_button and st.session_state.paused:
|
238 |
+
# # step = True
|
239 |
+
# # elif save_button:
|
240 |
+
# # # Save image and info using Streamlit
|
241 |
+
# # st.image(
|
242 |
+
# # env_render, caption=f"Epoch {epoch}", use_column_width=True
|
243 |
+
# # )
|
244 |
+
# # st.write(
|
245 |
+
# # {
|
246 |
+
# # "state": {i: str(val) for i, val in enumerate(state)},
|
247 |
+
# # "feature": {
|
248 |
+
# # i: str(val) for i, val in enumerate(summed_importances)
|
249 |
+
# # },
|
250 |
+
# # "action": str(action),
|
251 |
+
# # "reward": str(reward),
|
252 |
+
# # "desired_return": str(desired_return + reward),
|
253 |
+
# # "desired_horizon": str(desired_horizon + 1),
|
254 |
+
# # }
|
255 |
+
# # )
|
256 |
+
|
257 |
+
# env.close()
|
258 |
+
|
259 |
+
|
260 |
+
# env = "Acrobot-v1"
|
261 |
+
# desired_return = -79
|
262 |
+
# desired_horizon = 82
|
263 |
+
# max_epoch = 500
|
264 |
+
|
265 |
+
|
266 |
+
# policy = SklearnPolicy.load("policy")
|
267 |
+
# hyper = AgentHyper(
|
268 |
+
# env,
|
269 |
+
# warm_up=0,
|
270 |
+
# )
|
271 |
+
|
272 |
+
# agent = UpsideDownAgent(hyper, policy)
|
273 |
+
|
274 |
+
# run_visualization(
|
275 |
+
# env, agent, desired_return, desired_horizon, max_epoch, "data/viz_examples"
|
276 |
+
# )
|
old_code/experiment_3/q_networks/buffers/CartPole-v0/1/DQN/memory_buffer.p
DELETED
The diff for this file is too large to render.
See raw diff
|
|
poetry.lock
CHANGED
The diff for this file is too large to render.
See raw diff
|
|
pyproject.toml
CHANGED
@@ -26,6 +26,7 @@ python = "3.10.14"
|
|
26 |
scikit-learn = "^1.5.2"
|
27 |
matplotlib = "^3.9.2"
|
28 |
gymnasium = {extras = ["box2d"], version = "^0.29.1"}
|
|
|
29 |
scikit-image = "^0.24.0"
|
30 |
tqdm = "^4.66.5"
|
31 |
torch = "^2.4.1"
|
|
|
26 |
scikit-learn = "^1.5.2"
|
27 |
matplotlib = "^3.9.2"
|
28 |
gymnasium = {extras = ["box2d"], version = "^0.29.1"}
|
29 |
+
numpy = "1.24.4"
|
30 |
scikit-image = "^0.24.0"
|
31 |
tqdm = "^4.66.5"
|
32 |
torch = "^2.4.1"
|
udrl/__main__.py
CHANGED
@@ -1,6 +1,6 @@
|
|
1 |
-
from .agent import UpsideDownAgent, AgentHyper
|
2 |
-
from .policies import SklearnPolicy, NeuralPolicy
|
3 |
-
from .catch import CatchAdaptor
|
4 |
from dataclasses import dataclass, asdict
|
5 |
import gymnasium as gym
|
6 |
from tqdm import trange
|
|
|
1 |
+
from udrl.agent import UpsideDownAgent, AgentHyper
|
2 |
+
from udrl.policies import SklearnPolicy, NeuralPolicy
|
3 |
+
from udrl.catch import CatchAdaptor
|
4 |
from dataclasses import dataclass, asdict
|
5 |
import gymnasium as gym
|
6 |
from tqdm import trange
|
udrl/agent.py
CHANGED
@@ -2,9 +2,9 @@ from dataclasses import dataclass
|
|
2 |
import gymnasium as gym
|
3 |
import numpy as np
|
4 |
|
5 |
-
from .catch import CatchAdaptor
|
6 |
-
from .policies import ABCPolicy
|
7 |
-
from .buffer import ReplayBuffer
|
8 |
|
9 |
|
10 |
@dataclass
|
|
|
2 |
import gymnasium as gym
|
3 |
import numpy as np
|
4 |
|
5 |
+
from udrl.catch import CatchAdaptor
|
6 |
+
from udrl.policies import ABCPolicy
|
7 |
+
from udrl.buffer import ReplayBuffer
|
8 |
|
9 |
|
10 |
@dataclass
|
udrl/inference.py
CHANGED
@@ -1,7 +1,7 @@
|
|
1 |
import matplotlib.pyplot as plt
|
2 |
import numpy as np
|
3 |
-
from .policies import SklearnPolicy, NeuralPolicy
|
4 |
-
from .agent import UpsideDownAgent, AgentHyper
|
5 |
from pathlib import Path
|
6 |
from collections import Counter
|
7 |
from tqdm import trange
|
|
|
1 |
import matplotlib.pyplot as plt
|
2 |
import numpy as np
|
3 |
+
from udrl.policies import SklearnPolicy, NeuralPolicy
|
4 |
+
from udrl.agent import UpsideDownAgent, AgentHyper
|
5 |
from pathlib import Path
|
6 |
from collections import Counter
|
7 |
from tqdm import trange
|
udrl/plot.py
CHANGED
@@ -1,5 +1,5 @@
|
|
1 |
-
from .policies import SklearnPolicy
|
2 |
-
from .agent import UpsideDownAgent, AgentHyper
|
3 |
from pathlib import Path
|
4 |
import matplotlib.pyplot as plt
|
5 |
import numpy as np
|
|
|
1 |
+
from udrl.policies import SklearnPolicy
|
2 |
+
from udrl.agent import UpsideDownAgent, AgentHyper
|
3 |
from pathlib import Path
|
4 |
import matplotlib.pyplot as plt
|
5 |
import numpy as np
|
udrl/viz.py
CHANGED
@@ -1,8 +1,8 @@
|
|
1 |
import gymnasium as gym
|
2 |
import pygame
|
3 |
import numpy as np
|
4 |
-
from .policies import SklearnPolicy
|
5 |
-
from .agent import UpsideDownAgent, AgentHyper
|
6 |
from pathlib import Path
|
7 |
import json
|
8 |
|
|
|
1 |
import gymnasium as gym
|
2 |
import pygame
|
3 |
import numpy as np
|
4 |
+
from udrl.policies import SklearnPolicy
|
5 |
+
from udrl.agent import UpsideDownAgent, AgentHyper
|
6 |
from pathlib import Path
|
7 |
import json
|
8 |
|