Spaces:
Runtime error
Runtime error
shubhambhawsar
commited on
app.py
Browse files
app.py
ADDED
@@ -0,0 +1,111 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
import pickle
|
2 |
+
import json
|
3 |
+
import torch
|
4 |
+
import torch
|
5 |
+
import torch.nn as nn
|
6 |
+
import gradio as gr
|
7 |
+
from torch.utils.data import Dataset
|
8 |
+
from transformers import BertTokenizer, BertModel
|
9 |
+
|
10 |
+
tokenizer = BertTokenizer.from_pretrained('bert-base-multilingual-cased')
|
11 |
+
model = BertModel.from_pretrained('bert-base-multilingual-cased')
|
12 |
+
device=torch.device("cuda" if torch.cuda.is_available() else "cpu")
|
13 |
+
|
14 |
+
with open("word_to_index.pkl", "rb") as f:
|
15 |
+
word_to_index = pickle.load(f)
|
16 |
+
|
17 |
+
# Loading index to word mapping
|
18 |
+
with open("index_to_word.pkl", "rb") as f:
|
19 |
+
index_to_word = pickle.load(f)
|
20 |
+
numclass=len(index_to_word)
|
21 |
+
|
22 |
+
|
23 |
+
class DefinitionClassifier(nn.Module):
|
24 |
+
def __init__(self, model, tokenizer, num_classes):
|
25 |
+
super(DefinitionClassifier, self).__init__()
|
26 |
+
|
27 |
+
# Load pre-trained Indic BERT model
|
28 |
+
self.bert = model
|
29 |
+
self.tokenizer = tokenizer
|
30 |
+
self.num_classes = num_classes
|
31 |
+
|
32 |
+
# Define classification layer
|
33 |
+
self.classifier = nn.Linear(self.bert.config.hidden_size, num_classes)
|
34 |
+
|
35 |
+
def forward(self, input_ids, attention_mask):
|
36 |
+
# Forward pass through the BERT model
|
37 |
+
outputs = self.bert(input_ids, attention_mask)
|
38 |
+
|
39 |
+
# Extract the CLS token embeddings
|
40 |
+
cls_embeddings = outputs.pooler_output
|
41 |
+
|
42 |
+
# Pass the CLS embeddings through the classification layer
|
43 |
+
logits = self.classifier(cls_embeddings)
|
44 |
+
|
45 |
+
return logits
|
46 |
+
model_final = DefinitionClassifier(model=model,tokenizer=tokenizer,num_classes=numclass).to(device)
|
47 |
+
state_dict = torch.load("/data5/home/shubhambhaws2/project_drona1/project/modelmbert.pth")
|
48 |
+
|
49 |
+
# Load the state dictionary into the model
|
50 |
+
model_final.load_state_dict(state_dict)
|
51 |
+
|
52 |
+
def formate_story(story):
|
53 |
+
sen= story.split("\n")
|
54 |
+
c=0
|
55 |
+
sen[0]=sen[0]+"\n"
|
56 |
+
fin=[]
|
57 |
+
for i in sen:
|
58 |
+
if(c%2==0):
|
59 |
+
fin.append(i+"\n")
|
60 |
+
else:
|
61 |
+
fin.append(i)
|
62 |
+
c=c+1
|
63 |
+
return "".join(fin)
|
64 |
+
|
65 |
+
print("Model Loaded")
|
66 |
+
|
67 |
+
def generate_word(defination, k):
|
68 |
+
text = "the place where everyone can go"
|
69 |
+
inputs = tokenizer.encode_plus(
|
70 |
+
defination,
|
71 |
+
max_length=128,
|
72 |
+
padding='max_length',
|
73 |
+
truncation=True,
|
74 |
+
return_tensors='pt'
|
75 |
+
)
|
76 |
+
input_ids = inputs['input_ids'].to(device)
|
77 |
+
attention_mask = inputs['attention_mask'].to(device)
|
78 |
+
logits = model_final(input_ids, attention_mask)
|
79 |
+
|
80 |
+
probabilities = torch.softmax(logits, dim=1)
|
81 |
+
topk_probabilities, topk_indices = torch.topk(probabilities, k, dim=1)
|
82 |
+
pred = topk_indices.squeeze().cpu().numpy().tolist()
|
83 |
+
output = []
|
84 |
+
|
85 |
+
for i in pred:
|
86 |
+
output.append(index_to_word[i])
|
87 |
+
|
88 |
+
# Return a string with each word on a new line
|
89 |
+
return "\n".join(output)
|
90 |
+
|
91 |
+
def gradio_reset():
|
92 |
+
return None, None, 10
|
93 |
+
|
94 |
+
# New Title
|
95 |
+
title = """<h1 align='center'>Reverse Dictionary</a></h1>"""
|
96 |
+
|
97 |
+
with gr.Blocks() as demo:
|
98 |
+
gr.Markdown(title)
|
99 |
+
story = gr.Textbox(label="Input Description", lines=2)
|
100 |
+
k = gr.Slider(label="Total Output", minimum=1, maximum=100, step=1, value=10)
|
101 |
+
|
102 |
+
with gr.Row():
|
103 |
+
upload_button = gr.Button(value="Generate Word", interactive=True, variant="primary")
|
104 |
+
clear = gr.Button("Clear")
|
105 |
+
output = gr.Textbox(label="Output Hindi word", lines=20)
|
106 |
+
|
107 |
+
upload_button.click(generate_word, [story, k], [output])
|
108 |
+
clear.click(gradio_reset, [], [story, output,k])
|
109 |
+
|
110 |
+
demo.queue()
|
111 |
+
demo.launch(share=True)
|