File size: 5,076 Bytes
e2b757a
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
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
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
# %%
import asyncio
import pickle as pk
import time
import warnings

import matplotlib as mpl
import matplotlib.pyplot as plt
import mpl_toolkits.mplot3d.art3d as art3d
import numpy as np
import torch
from matplotlib import cm
from matplotlib.animation import FuncAnimation
from matplotlib.gridspec import GridSpec
from matplotlib.patches import Circle, PathPatch
from mpl_toolkits.mplot3d import Axes3D, axes3d
from sklearn.decomposition import PCA

warnings.filterwarnings("ignore", category=UserWarning)

# file_path = "word_embeddings_mpnet.pth"
# embeddings_dict = torch.load(file_path)

# # %%
# words = list(embeddings_dict.keys())

# sentences = [[word] for word in words]

# vectors = list(embeddings_dict.values())
# vectors_list = []
# for item in vectors:
#     vectors_list.append(item.tolist())

# vector_list = vectors_list[:10]
# # %%
# # pca = PCA(n_components=3)
# # pca = pca.fit(vectors_list)
# # pk.dump(pca, open("pca_mpnet.pkl", "wb"))
# score = np.array([0, 0, 0, 0, 0, 1, 1, 1, 1, 1])


# %%
def display_words(words, vector_list, score, bold):
    # %%
    plt.ioff()
    fig = plt.figure()

    ax = fig.add_subplot(111, projection="3d")
    plt.rcParams["image.cmap"] = "magma"
    colormap = cm.get_cmap("magma")  # You can choose any colormap you like

    # Normalize the float values to the range [0, 1]
    score = np.array(score)
    norm = plt.Normalize(0, 10)  # type: ignore
    colors = colormap(norm(score))
    ax.xaxis.pane.fill = False
    ax.yaxis.pane.fill = False
    ax.w_zaxis.set_pane_color(
        (0.87, 0.91, 0.94, 0.8)
    )  # Set the z-axis face color (gray)
    ax.xaxis.line.set_color((1.0, 1.0, 1.0, 0.0))  # Transparent x-axis line
    ax.yaxis.line.set_color((1.0, 1.0, 1.0, 0.0))  # Transparent y-axis line
    ax.zaxis.line.set_color((1.0, 1.0, 1.0, 0.0))

    # Turn off axis labels
    ax.set_xticks([])
    ax.set_yticks([])
    ax.set_zticks([])
    ax.grid(False)
    # %%
    data_pca = vector_list

    if len(data_pca) > 1:
        # for i in range(len(data_pca) - 1):
        #     data = np.append(
        #         data_pca,
        #         [norm_distance(data_pca[0], data_pca[i + 1], score[i + 1])],
        #         axis=0,
        #     )

        # Create copies of the zero-th element of data_pca
        data_pca0 = np.repeat(data_pca[0][None, :], len(data_pca) - 1, axis=0)

        # Use these arrays to construct the calls to norm_distance_v
        data = norm_distance_v(data_pca0, data_pca[1:], score[1:])

    else:
        data = data_pca.transpose()

    (
        x,
        y,
        z,
    ) = data

    center_x = x[0]
    center_y = y[0]
    center_z = z[0]
    # %%
    ax.autoscale(enable=True, axis="both", tight=True)
    # if bold == -1:
    #     k = len(words) - 1
    # else:
    #     k = repeated
    for i, word in enumerate(words):
        if i == bold:
            fontsize = "large"
            fontweight = "demibold"
        else:
            fontsize = "medium"
            fontweight = "normal"

        ax.text(
            x[i],
            y[i],
            z[i] + 0.05,
            word,
            fontsize=fontsize,
            fontweight=fontweight,
            alpha=1,
        )
    # ax.text(
    #     x[0],
    #     y[0],
    #     z[0] + 0.05,
    #     words[0],
    #     fontsize="medium",
    #     fontweight="normal",
    #     alpha=1,
    # )
    ax.scatter(x, y, z, c="black", marker="o", s=75, cmap="magma", vmin=0, vmax=10)
    scatter = ax.scatter(
        x,
        y,
        z,
        marker="o",
        s=70,
        c=colors,
        cmap="magma",
        vmin=0,
        vmax=10,
    )

    # cax = fig.add_subplot(gs[1, :])  # cb = plt.colorbar(sc, cax=cax)
    # a = fig.colorbar(
    #     mappable=scatter,
    #     ax=ax,
    #     cmap="magma",
    #     norm=mpl.colors.Normalize(vmin=0, vmax=10),
    #     orientation="horizontal",
    # )
    fig.colorbar(
        cm.ScalarMappable(norm=mpl.colors.Normalize(0, 10), cmap="magma"),
        ax=ax,
        orientation="horizontal",
    )
    # cbar.set_label("Score Values")

    def update(frame):
        distance = 0.5 * (score.max() - score.min())
        ax.set_xlim(center_x - distance, center_x + distance)
        ax.set_ylim(center_y - distance, center_y + distance)
        ax.set_zlim(center_z - distance, center_z + distance)
        ax.view_init(elev=20, azim=frame)

    # %%

    # Create the animation
    frames = np.arange(0, 360, 5)
    ani = FuncAnimation(fig, update, frames=frames, interval=120)

    ani.save("3d_rotation.gif", writer="pillow", dpi=140)
    plt.close(fig)


# %%
def norm_distance_v(orig, points, distances):
    # Calculate the vector AB

    AB = points - orig

    # Calculate the normalized vector AB
    Normalized_AB = AB / np.linalg.norm(AB, axis=1, keepdims=True)

    # Specify the desired distance from point A
    d = 10 - (distances.reshape(-1, 1) * 1)

    # Calculate the new points C
    C = orig + (Normalized_AB * d)
    C = np.append([orig[0]], C, axis=0)

    return np.array([C[:, 0], C[:, 1], C[:, 2]])