aakaash912 commited on
Commit
1cfd7ee
·
verified ·
1 Parent(s): e88b614

Update app.py

Browse files
Files changed (1) hide show
  1. app.py +48 -46
app.py CHANGED
@@ -2,73 +2,76 @@ import google.generativeai as genai
2
  import gradio as gr
3
  from diffusers import StableDiffusionPipeline
4
  import torch
 
5
 
6
- # Initialize the models
7
- def initialize_models():
8
- # Configure Gemini
9
- GOOGLE_API_KEY = "AIzaSyCK-tdqs-O7sMUR938ZZy0kF-2vblNJKDo"
10
-
11
- genai.configure(api_key=GOOGLE_API_KEY)
12
- gemini_model = genai.GenerativeModel("gemini-1.5-flash-8b")
13
-
14
- # Initialize Stable Diffusion for CPU
15
- model_id = "CompVis/stable-diffusion-v1-4"
16
- pipeline = StableDiffusionPipeline.from_pretrained(
17
- model_id,
18
- torch_dtype=torch.float32
19
- )
20
- print("Model Initialization Done")
21
-
22
- return gemini_model, pipeline
23
 
24
- # Define the processing functions
25
- def translate_tamil_to_english(gemini_model, tamil_text):
 
 
 
 
 
 
 
 
 
 
 
 
 
26
  try:
27
- response = gemini_model.generate_content(
28
- f"Translate this Tamil text to English: {tamil_text}"
29
- )
30
- print("Translation Done")
31
  return response.text
32
  except Exception as e:
33
  return f"Translation error: {str(e)}"
34
 
35
- def generate_image(pipeline, prompt):
36
  try:
37
- image = pipeline(prompt).images[0]
38
- print("Image Generation Done")
 
 
 
 
 
 
 
 
 
 
 
39
  return image
40
  except Exception as e:
 
41
  return None
42
 
43
- def process_input(tamil_text, gemini_model, pipeline):
44
  if not tamil_text.strip():
45
  return "Please enter some text", None
46
 
47
  try:
48
  # Translate Tamil to English
49
- english_translation = translate_tamil_to_english(gemini_model, tamil_text)
50
  print(english_translation)
51
 
52
- # Generate image based on translation
53
  if english_translation.startswith("Translation error"):
54
  return english_translation, None
55
 
56
- image = generate_image(pipeline, english_translation)
 
57
  return english_translation, image
58
-
59
  except Exception as e:
60
- return f"Error: {str(e)}", None
61
-
62
- # Initialize models
63
- try:
64
- gemini_model, pipeline = initialize_models()
65
- except Exception as e:
66
- print(f"Error initializing models: {str(e)}")
67
- gemini_model, pipeline = None, None
68
 
69
- # Create Gradio interface
70
  with gr.Blocks() as demo:
71
- gr.Markdown("### Tamil to English Translator and Image Generator")
72
 
73
  with gr.Row():
74
  tamil_input = gr.Textbox(
@@ -82,19 +85,18 @@ with gr.Blocks() as demo:
82
  label="English Translation",
83
  interactive=False
84
  )
85
-
86
  with gr.Row():
87
  image_output = gr.Image(label="Generated Image")
88
 
 
89
  btn_process = gr.Button("Generate")
90
-
91
- # Update process function to include model parameters
92
  btn_process.click(
93
- fn=lambda x: process_input(x, gemini_model, pipeline),
94
  inputs=[tamil_input],
95
  outputs=[english_output, image_output]
96
  )
97
 
98
  # Launch the interface
99
  if __name__ == "__main__":
100
- demo.launch(__debug__=True)
 
2
  import gradio as gr
3
  from diffusers import StableDiffusionPipeline
4
  import torch
5
+ from PIL import Image
6
 
7
+ # Initialize the Stable Diffusion model with optimizations
8
+ model_id = "CompVis/stable-diffusion-v1-4"
9
+ genai.configure(api_key="AIzaSyCK-tdqs-O7sMUR938ZZy0kF-2vblNJKDo")
10
+ model = genai.GenerativeModel("gemini-1.5-flash-8b")
 
 
 
 
 
 
 
 
 
 
 
 
 
11
 
12
+ # Initialize optimized pipeline
13
+ pipeline = StableDiffusionPipeline.from_pretrained(
14
+ model_id,
15
+ torch_dtype=torch.float32,
16
+ revision="fp16",
17
+ safety_checker=None,
18
+ requires_safety_checking=False
19
+ )
20
+
21
+ # Apply memory optimizations
22
+ pipeline.enable_attention_slicing(slice_size="max")
23
+ pipeline.enable_model_cpu_offload()
24
+
25
+ # Define the Gemini interaction functions
26
+ def translate_tamil_to_english_gemini(tamil_text):
27
  try:
28
+ response = model.generate_content(f"Translate this Tamil text to English: {tamil_text}")
 
 
 
29
  return response.text
30
  except Exception as e:
31
  return f"Translation error: {str(e)}"
32
 
33
+ def generate_image_gemini(prompt):
34
  try:
35
+ # Optimized image generation parameters
36
+ image = pipeline(
37
+ prompt,
38
+ num_inference_steps=20, # Reduced steps for faster generation
39
+ guidance_scale=7.0, # Slightly reduced from default
40
+ height=512, # Optimized dimensions
41
+ width=512
42
+ ).images[0]
43
+
44
+ # Optimize output image size
45
+ if isinstance(image, Image.Image):
46
+ image = image.resize((384, 384), Image.Resampling.LANCZOS)
47
+
48
  return image
49
  except Exception as e:
50
+ print(f"Image generation error: {str(e)}")
51
  return None
52
 
53
+ def process_input_gemini(tamil_text):
54
  if not tamil_text.strip():
55
  return "Please enter some text", None
56
 
57
  try:
58
  # Translate Tamil to English
59
+ english_translation = translate_tamil_to_english_gemini(tamil_text)
60
  print(english_translation)
61
 
 
62
  if english_translation.startswith("Translation error"):
63
  return english_translation, None
64
 
65
+ # Generate an image based on the translation
66
+ image = generate_image_gemini(english_translation)
67
  return english_translation, image
 
68
  except Exception as e:
69
+ print(f"Processing error: {str(e)}")
70
+ return str(e), None
 
 
 
 
 
 
71
 
72
+ # Create Gradio UI
73
  with gr.Blocks() as demo:
74
+ gr.Markdown("### Tamil to English Translator and Image Generator (Optimized)")
75
 
76
  with gr.Row():
77
  tamil_input = gr.Textbox(
 
85
  label="English Translation",
86
  interactive=False
87
  )
88
+
89
  with gr.Row():
90
  image_output = gr.Image(label="Generated Image")
91
 
92
+ # Button to process
93
  btn_process = gr.Button("Generate")
 
 
94
  btn_process.click(
95
+ fn=process_input_gemini,
96
  inputs=[tamil_input],
97
  outputs=[english_output, image_output]
98
  )
99
 
100
  # Launch the interface
101
  if __name__ == "__main__":
102
+ demo.launch(share=True, debug=True)