Param56/Environmental_Ass
0
1# Install required libraries
2# !pip install gradio transformers torch spacy networkx matplotlib diffusers sentencepiece accelerate ftfy
3# !python -m spacy download en_core_web_sm
4
5import gradio as gr
6from transformers import pipeline
7import spacy
8import networkx as nx
9import matplotlib.pyplot as plt
10from transformers import AutoTokenizer, AutoModelForSequenceClassification, AutoModelForMaskedLM
11from diffusers import StableDiffusionPipeline
12import torch
13import warnings
14warnings.filterwarnings("ignore")
15
16# Initialize models
17classifier = None
18image_pipe = None
19ner_pipe = None
20mask_filler = None
21
22def download_models():
23 global classifier, image_pipe, ner_pipe, mask_filler
24
25 # Sentence Classification Model
26 print("Downloading sentence classification model...")
27 classifier = pipeline(
28 "text-classification",
29 model="distilbert-base-uncased-finetuned-sst-2-english"
30 )
31
32 # Image Generation Model
33 print("Downloading image generation model...")
34 image_pipe = StableDiffusionPipeline.from_pretrained(
35 "CompVis/stable-diffusion-v1-4",
36 torch_dtype=torch.float16,
37 use_auth_token=True
38 ).to("cuda" if torch.cuda.is_available() else "cpu")
39
40 # NER Model
41 print("Downloading NER model...")
42 ner_pipe = pipeline(
43 "ner",
44 model="dbmdz/bert-large-cased-finetuned-conll03-english",
45 grouped_entities=True
46 )
47
48 # Mask Filling Model
49 print("Downloading mask filling model...")
50 mask_filler = pipeline(
51 "fill-mask",
52 model="bert-large-uncased"
53 )
54 return "All models downloaded successfully!"
55
56def classify_text(text):
57 environment_categories = {
58 'POSITIVE': 'Environmentally Positive',
59 'NEGATIVE': 'Environmentally Negative',
60 'NEUTRAL': 'Environmentally Neutral',
61 'POLICY': 'Environmental Policy',
62 'SCIENCE': 'Environmental Science'
63 }
64
65 result = classifier(text)[0]
66 label = result['label']
67 score = result['score']
68
69 if label in ['LABEL_0', 'NEGATIVE']:
70 return f"Category: {environment_categories['NEGATIVE']}\nConfidence: {score:.2f}"
71 elif label in ['LABEL_1', 'POSITIVE']:
72 return f"Category: {environment_categories['POSITIVE']}\nConfidence: {score:.2f}"
73 else:
74 return f"Category: {environment_categories['NEUTRAL']}\nConfidence: {score:.2f}"
75
76def generate_environment_image(prompt):
77 generator = torch.Generator("cuda" if torch.cuda.is_available() else "cpu").manual_seed(42)
78 image = image_pipe(
79 f"environmental {prompt}, high quality, detailed, nature, realistic",
80 generator=generator,
81 num_inference_steps=50
82 ).images[0]
83 return image
84
85def create_ner_graph(text):
86 ner_results = ner_pipe(text)
87 G = nx.Graph()
88
89 for entity in ner_results:
90 entity_text = entity['word']
91 entity_type = entity['entity_group']
92 G.add_node(entity_text, type=entity_type)
93
94 for i in range(len(ner_results)-1):
95 node1 = ner_results[i]['word']
96 node2 = ner_results[i+1]['word']
97 G.add_edge(node1, node2)
98
99 plt.figure(figsize=(12, 8))
100 pos = nx.spring_layout(G)
101 colors = []
102 for node in G.nodes():
103 if G.nodes[node]['type'] == 'PER':
104 colors.append('red')
105 elif G.nodes[node]['type'] == 'ORG':
106 colors.append('blue')
107 elif G.nodes[node]['type'] == 'LOC':
108 colors.append('green')
109 else:
110 colors.append('yellow')
111
112 nx.draw(G, pos, with_labels=True, node_color=colors, node_size=2000, font_size=12)
113 plt.title("Environmental NER Graph")
114 plt.axis("off")
115 plt.savefig("ner_graph.png")
116 plt.close()
117 return "ner_graph.png"
118
119def fill_environment_mask(text_with_mask):
120 results = mask_filler(text_with_mask)
121 top_result = results[0]
122 return {
123 "filled_text": text_with_mask.replace("[MASK]", top_result['token_str']),
124 "options": [r['token_str'] for r in results]
125 }
126
127# Gradio Interface
128with gr.Blocks(title="Environmental AI Assistant") as demo:
129 gr.Markdown("# 🌍 Environmental AI Assistant")
130 gr.Markdown("Analyze and generate environmental content using AI models")
131
132 with gr.Tab("Sentence Classification"):
133 gr.Markdown("### Classify text into environmental categories")
134 text_input = gr.Textbox(label="Enter environmental text")
135 classify_btn = gr.Button("Classify")
136 classification_output = gr.Textbox(label="Classification Result")
137 classify_btn.click(classify_text, inputs=text_input, outputs=classification_output)
138
139 with gr.Tab("Image Generation"):
140 gr.Markdown("### Generate environmental images from text")
141 prompt_input = gr.Textbox(label="Describe the environmental scene")
142 generate_btn = gr.Button("Generate Image")
143 image_output = gr.Image(label="Generated Image")
144 generate_btn.click(generate_environment_image, inputs=prompt_input, outputs=image_output)
145
146 with gr.Tab("NER Graph"):
147 gr.Markdown("### Extract entities and visualize relationships")
148 ner_text_input = gr.Textbox(label="Input text with environmental content")
149 ner_btn = gr.Button("Extract Entities")
150 ner_graph_output = gr.Image(label="Entity Relationship Graph")
151 ner_btn.click(create_ner_graph, inputs=ner_text_input, outputs=ner_graph_output)
152
153 with gr.Tab("Fill Mask"):
154 gr.Markdown("### Complete environmental sentences with AI")
155 mask_examples = [
156 "Reducing [MASK] emissions is critical for climate change.",
157 "The [MASK] is home to many endangered species.",
158 "We should recycle more [MASK] to help the environment."
159 ]
160 mask_input = gr.Textbox(label="Enter sentence with [MASK]", value=mask_examples[0])
161 fill_btn = gr.Button("Fill Mask")
162 filled_output = gr.Textbox(label="Completed Sentence")
163 options_output = gr.HighlightedText(label="Alternative Options")
164 fill_btn.click(lambda x: fill_environment_mask(x), inputs=mask_input, outputs=[filled_output, options_output])
165 gr.Examples(examples=mask_examples, inputs=mask_input)
166
167print("Downloading models... (this may take several minutes)")
168download_status = download_models()
169print(download_status)
170demo.launch(share=True)