Spaces:
Runtime error
Runtime error
File size: 5,025 Bytes
bbf97b5 | 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 | import matplotlib.pyplot as plt
import numpy as np
import imageio
import torch
from matplotlib.backends.backend_agg import FigureCanvasAgg as FigureCanvas
def draw_grid_frame(path, grid_size=(4, 4), holes=[(1, 1), (1, 3), (3, 0)], goal=(3, 3), episode=None):
fig, ax = plt.subplots(figsize=(4, 4))
ax.set_xticks(np.arange(grid_size[1]+1)-0.5, minor=True)
ax.set_yticks(np.arange(grid_size[0]+1)-0.5, minor=True)
ax.grid(which="minor", color="black", linestyle='-', linewidth=1)
ax.tick_params(which="minor", bottom=False, left=False)
ax.set_xticks([])
ax.set_yticks([])
for i in range(grid_size[0]):
for j in range(grid_size[1]):
label = chr(65 + i * grid_size[1] + j)
color = "white"
if (i, j) in holes:
color = "#ffcccc"
elif (i, j) == goal:
color = "#ccffcc"
ax.add_patch(plt.Rectangle((j - 0.5, i - 0.5), 1, 1, facecolor=color))
ax.text(j, i, label, ha='center', va='center', fontsize=12)
for i, (x, y) in enumerate(path):
ax.add_patch(plt.Circle((y, x), 0.3, color='blue', alpha=0.3 + 0.5 * (i/len(path))))
if path:
x, y = path[-1]
ax.add_patch(plt.Circle((y, x), 0.3, color='red'))
ax.set_xlim(-0.5, grid_size[1]-0.5)
ax.set_ylim(-0.5, grid_size[0]-0.5)
ax.invert_yaxis()
if episode is not None:
ax.set_title(f"Episode {episode}", fontsize=14)
return fig
def save_agent_walk_gif(trajectory, filename="agent_walk.gif", episode=None, loop=True):
frames = []
for i in range(1, len(trajectory) + 1):
fig = draw_grid_frame(trajectory[:i], episode=episode)
canvas = FigureCanvas(fig)
canvas.draw()
image = np.frombuffer(canvas.buffer_rgba(), dtype='uint8')
image = image.reshape(fig.canvas.get_width_height()[::-1] + (4,))
frames.append(image)
plt.close(fig)
imageio.mimsave(filename, frames, duration=0.5, loop=0 if loop else 1)
def plot_policy(Q, grid_size=4):
fig, ax = plt.subplots(figsize=(6, 6))
arrows = {'up': '↑', 'down': '↓', 'left': '←', 'right': '→'}
labels = np.array([chr(65 + i * grid_size + j) for i in range(grid_size) for j in range(grid_size)])
labels = labels.reshape((grid_size, grid_size))
holes = [(1, 1), (1, 3), (3, 0)]
goal = (3, 3)
start = (0, 0)
cell_text = []
cell_colors = []
for i in range(grid_size):
row_text = []
row_color = []
for j in range(grid_size):
state = (i, j)
label = labels[i, j]
if state in holes:
row_text.append(f"❌\n{label}")
row_color.append("#ffcccc")
elif state == goal:
row_text.append(f"✅\n{label}")
row_color.append("#ccffcc")
elif state == start:
row_text.append(f"🟦\n{label}")
row_color.append("#cce5ff")
elif state in Q:
best_action = max(Q[state], key=Q[state].get)
row_text.append(f"{arrows[best_action]}\n{label}")
row_color.append("white")
else:
row_text.append(label)
row_color.append("white")
cell_text.append(row_text)
cell_colors.append(row_color)
table = ax.table(cellText=cell_text,
cellColours=cell_colors,
loc='center',
cellLoc='center',
colWidths=[0.2]*grid_size)
table.scale(1, 2)
ax.axis('off')
plt.tight_layout()
return fig
def plot_heatmap(Q):
fig, ax = plt.subplots()
values = np.zeros((4, 4))
for i in range(4):
for j in range(4):
state = (i, j)
if state in Q:
values[i][j] = max(Q[state].values())
c = ax.imshow(values, cmap='coolwarm', interpolation='nearest')
plt.colorbar(c)
ax.set_title("Q-Value Heatmap (Best Actions)")
return fig
def plot_visits(visits):
fig, ax = plt.subplots()
counts = np.zeros((4, 4))
for (i, j), count in visits.items():
counts[i][j] = count
c = ax.imshow(counts, cmap='YlGn', interpolation='nearest')
plt.colorbar(c)
ax.set_title("State Visit Frequency")
return fig
def plot_dqn_qvalues(agent_model, actions):
q_grid = np.zeros((4, 4))
with torch.no_grad():
for i in range(4):
for j in range(4):
input_tensor = torch.FloatTensor([i, j])
q_vals = agent_model(input_tensor)
best_action_val = torch.max(q_vals).item()
q_grid[i, j] = best_action_val
fig, ax = plt.subplots()
im = ax.imshow(q_grid, cmap="coolwarm")
ax.set_title("DQN Q-Value Heatmap (Best Actions)")
plt.colorbar(im)
return fig
|