marksverdhei commited on
Commit
010edb7
·
1 Parent(s): 74e9a4d
Files changed (1) hide show
  1. views.py +80 -0
views.py ADDED
@@ -0,0 +1,80 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ import streamlit as st
2
+ import vec2text
3
+ import torch
4
+ from umap import UMAP
5
+ import plotly.express as px
6
+ import numpy as np
7
+ from streamlit_plotly_events import plotly_events
8
+ from resources import reduce_embeddings
9
+ import utils
10
+ import pandas as pd
11
+
12
+
13
+ def diffs(embeddings: np.ndarray, corrector):
14
+ st.text(f"Embedding shape: {embeddings.shape}")
15
+ st.html('<a href="https://www.flaticon.com/free-icons/array" title="array icons">Array icons created by Voysla - Flaticon</a>')
16
+
17
+ def plot(df: pd.DataFrame, embeddings: np.ndarray, vectors_2d, reducer, corrector):
18
+
19
+
20
+ # Add a scatter plot using Plotly
21
+ fig = px.scatter(
22
+ x=vectors_2d[:, 0],
23
+ y=vectors_2d[:, 1],
24
+ opacity=0.6,
25
+ hover_data={"Title": df["title"]},
26
+ labels={'x': 'UMAP Dimension 1', 'y': 'UMAP Dimension 2'},
27
+ title="UMAP Scatter Plot of Reddit Titles",
28
+ color_discrete_sequence=["#ff504c"] # Set default blue color for points
29
+ )
30
+
31
+ # Customize the layout to adapt to browser settings (light/dark mode)
32
+ fig.update_layout(
33
+ template=None, # Let Plotly adapt automatically based on user settings
34
+ plot_bgcolor="rgba(0, 0, 0, 0)",
35
+ paper_bgcolor="rgba(0, 0, 0, 0)"
36
+ )
37
+
38
+ x, y = 0.0, 0.0
39
+ vec = np.array([x, y]).astype("float32")
40
+
41
+ # Add a card container to the right of the content with Streamlit columns
42
+ col1, col2 = st.columns([0.6, 0.4]) # Adjusting ratio to allocate space for the card container
43
+
44
+ with col1:
45
+ # Main content stays here (scatterplot, form, etc.)
46
+ selected_points = plotly_events(fig, click_event=True, hover_event=False, #override_height=600, override_width="100%"
47
+ )
48
+ with st.form(key="form1_main"):
49
+ if selected_points:
50
+ clicked_point = selected_points[0]
51
+ x_coord = x = clicked_point['x']
52
+ y_coord = y = clicked_point['y']
53
+
54
+ x = st.number_input("X Coordinate", value=x, format="%.10f")
55
+ y = st.number_input("Y Coordinate", value=y, format="%.10f")
56
+ vec = np.array([x, y]).astype("float32")
57
+
58
+
59
+ submit_button = st.form_submit_button("Submit")
60
+
61
+ if selected_points or submit_button:
62
+ inferred_embedding = reducer.inverse_transform(np.array([[x, y]]) if not isinstance(reducer, UMAP) else np.array([[x, y]]))
63
+ inferred_embedding = inferred_embedding.astype("float32")
64
+
65
+ output = vec2text.invert_embeddings(
66
+ embeddings=torch.tensor(inferred_embedding).cuda(),
67
+ corrector=corrector,
68
+ num_steps=20,
69
+ )
70
+
71
+ st.text(str(output))
72
+ st.text(str(inferred_embedding))
73
+ else:
74
+ st.text("Click on a point in the scatterplot to see its coordinates.")
75
+
76
+ with col2:
77
+ closest_sentence_index = utils.find_exact_match(vectors_2d, vec, decimals=3)
78
+ st.markdown(
79
+ f"### Selected text:\n```console\n{df.title.iloc[closest_sentence_index] if closest_sentence_index > -1 else '[no selected text]'}\n```"
80
+ )