Upload graph-byclip.py
Browse files- graph-byclip.py +77 -0
graph-byclip.py
ADDED
@@ -0,0 +1,77 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
#!/bin/env python
|
2 |
+
|
3 |
+
""" Work in progress
|
4 |
+
Plan:
|
5 |
+
Modded version of graph-embeddings.py
|
6 |
+
Just to see if using different CLIP module changes values significantly
|
7 |
+
(It does not)
|
8 |
+
This requires
|
9 |
+
pip install git+https://github.com/openai/CLIP.git
|
10 |
+
"""
|
11 |
+
|
12 |
+
|
13 |
+
import sys
|
14 |
+
import json
|
15 |
+
import torch
|
16 |
+
import clip
|
17 |
+
|
18 |
+
import PyQt5
|
19 |
+
import matplotlib
|
20 |
+
matplotlib.use('QT5Agg') # Set the backend to TkAgg
|
21 |
+
|
22 |
+
import matplotlib.pyplot as plt
|
23 |
+
|
24 |
+
CLIPname= "ViT-L/14"
|
25 |
+
|
26 |
+
device=torch.device("cuda")
|
27 |
+
print("loading CLIP model")
|
28 |
+
model, processor = clip.load(CLIPname,device=device)
|
29 |
+
model.cuda().eval()
|
30 |
+
print("done")
|
31 |
+
|
32 |
+
def embed_from_text(text):
|
33 |
+
tokens = clip.tokenize(text).to(device)
|
34 |
+
|
35 |
+
with torch.no_grad():
|
36 |
+
embed = model.encode_text(tokens)
|
37 |
+
return embed
|
38 |
+
|
39 |
+
|
40 |
+
# Expect SINGLE WORD ONLY
|
41 |
+
def standard_embed_calc(text):
|
42 |
+
inputs = processor(text=text, return_tensors="pt")
|
43 |
+
inputs.to(device)
|
44 |
+
with torch.no_grad():
|
45 |
+
text_features = model.get_text_features(**inputs)
|
46 |
+
embedding = text_features[0]
|
47 |
+
return embedding
|
48 |
+
|
49 |
+
|
50 |
+
fig, ax = plt.subplots()
|
51 |
+
|
52 |
+
|
53 |
+
text1 = input("First word or prompt: ")
|
54 |
+
text2 = input("Second prompt(or leave blank): ")
|
55 |
+
|
56 |
+
|
57 |
+
print("generating embeddings for each now")
|
58 |
+
emb1 = embed_from_text(text1)
|
59 |
+
print("shape of emb1:",emb1.shape)
|
60 |
+
|
61 |
+
graph1=emb1[0].tolist()
|
62 |
+
ax.plot(graph1, label=text1[:20])
|
63 |
+
|
64 |
+
if len(text2) >0:
|
65 |
+
emb2 = embed_from_text(text2)
|
66 |
+
graph2=emb2[0].tolist()
|
67 |
+
ax.plot(graph2, label=text2[:20])
|
68 |
+
|
69 |
+
# Add labels, title, and legend
|
70 |
+
#ax.set_xlabel('Index')
|
71 |
+
ax.set_ylabel('Values')
|
72 |
+
ax.set_title('Comparative Graph of Two Embeddings')
|
73 |
+
ax.legend()
|
74 |
+
|
75 |
+
# Display the graph
|
76 |
+
print("Pulling up the graph")
|
77 |
+
plt.show()
|