File size: 3,407 Bytes
4c3ee55
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
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
# %%
import gradio.components as gc
import gradio as gr

import numpy as np
import pandas as pd
import torch
from PIL import Image
from transformers import CLIPModel, CLIPProcessor
device = 'cpu'
torch.no_grad().__enter__()
torch.autocast('cuda').__enter__()

# %%

t = pd.read_pickle("clip_texts_1_fp16.pkl")
words = t.reset_index().word
wordsv = torch.tensor(t.values).to(device)

# %%

# %%
model_name = "openai/clip-vit-large-patch14"
mmm = CLIPModel.from_pretrained(model_name)
mmm.eval()
mmm.to(device)

processor = CLIPProcessor.from_pretrained(model_name)

# %%


def slerp(t, v0, v1, DOT_THRESHOLD=0.9995):
    """ helper function to spherically interpolate two arrays v1 v2 """
    inputs_are_torch = False
    if not isinstance(v0, np.ndarray):
        inputs_are_torch = True
        input_device = v0.device
        v0 = v0.cpu().numpy()
        v1 = v1.cpu().numpy()

    dot = np.sum(v0 * v1 / (np.linalg.norm(v0) * np.linalg.norm(v1)))
    if np.abs(dot) > DOT_THRESHOLD:
        v2 = (1 - t) * v0 + t * v1
    else:
        theta_0 = np.arccos(dot)
        sin_theta_0 = np.sin(theta_0)
        theta_t = theta_0 * t
        sin_theta_t = np.sin(theta_t)
        s0 = np.sin(theta_0 - theta_t) / sin_theta_0
        s1 = sin_theta_t / sin_theta_0
        v2 = s0 * v0 + s1 * v1

    if inputs_are_torch:
        v2 = torch.from_numpy(v2).to(input_device)

    return v2


def query(text: str, img: Image.Image, limit: int, score_threshold: float, slerp_degree: float):
    if text != '':
        inp = processor(text=text, return_tensors='pt').to(device)
        rout = mmm.get_text_features(**inp)
        tout = rout.detach().cpu().numpy()[0]
        out = tout

    if img is not None:
        inp = processor(images=[img], return_tensors="pt",).to(device)
        rout = mmm.get_image_features(**inp)
        iout = rout.detach().cpu().numpy()[0]
        out = iout

    if text != '' and img is not None:
        out = slerp(slerp_degree, tout, iout)

    if out is not None:
        # calculate cosine similarity
        scores = np.dot(out, wordsv.T)
        # sort by score
        topk = (
            pd.concat(
                [words, pd.Series(scores, name='score')],
                axis=1
            )
            .sort_values('score', ascending=False)
            .query(f'score > {score_threshold}')
            .head(limit)
        )

        topwords = "\n".join(
            f'{word}: {score:.2f} '
            for _, word, score in topk.itertuples()
        )

        return topwords


searchtext = gc.Textbox(lines=2, placeholder="Search text")
searchimage = gc.Image(shape=(224, 224), label="Search image", type='pil')
inp_limit = gc.Slider(1, 50, 10, step=1, label='Limit')
score_threshold = gc.Slider(0, 30, 0, step=.5, label='Score threshold')
slerp_degree = gc.Slider(
    0, 1, 0.5, step=.01, label='Slerp degree (if both text and image are provided)\nFinds a midpoint between image and text embeddings')


dsurl = 'https://www.kaggle.com/datasets/yk1598/479k-english-words'
gr.Interface(
    query,
    [searchtext, searchimage, inp_limit, score_threshold, slerp_degree],
    [gc.Textbox(label='Top words')],
    title="Initial Token Finder for Textual Inversion",
    description=f"find the closest single token word for a given text and/or image.\nbased on {model_name}.\n\nData: {dsurl}",
    analytics_enabled=False,
    allow_flagging='never',
).launch()