kwabs22 commited on
Commit
f03a15c
1 Parent(s): 69d6dab

initial commit

Browse files
Files changed (3) hide show
  1. Dockerfile +20 -0
  2. app.py +56 -0
  3. requirements.txt +7 -0
Dockerfile ADDED
@@ -0,0 +1,20 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ # Use an official Python runtime as a parent image
2
+ FROM python:3.8-slim
3
+
4
+ # Set the working directory in the container
5
+ WORKDIR /usr/src/app
6
+
7
+ # Copy the current directory contents into the container at /usr/src/app
8
+ COPY . .
9
+
10
+ # Install any needed packages specified in requirements.txt
11
+ RUN pip install --no-cache-dir -r requirements.txt
12
+
13
+ # Make port 7860 available to the world outside this container
14
+ EXPOSE 7860
15
+
16
+ # Define environment variable
17
+ ENV NAME World
18
+
19
+ # Run app.py when the container launches
20
+ CMD ["python", "app.py"]
app.py ADDED
@@ -0,0 +1,56 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ import gradio as gr
2
+ from transformers import AutoModelForCausalLM, CodeGenTokenizerFast as Tokenizer
3
+ from PIL import Image
4
+ import gc
5
+ from pympler import asizeof
6
+
7
+ # Function to clear model from memory
8
+ def clear_model(model):
9
+ del model
10
+ gc.collect()
11
+
12
+
13
+ def process_image_and_question(image, question):
14
+ # Placeholder for your image processing and question answering
15
+ # Replace this with your actual model processing
16
+ # For example:
17
+ # enc_image = model.encode_image(image)
18
+ # answer = model.answer_question(enc_image, question, tokenizer)
19
+ # return answer
20
+
21
+ FinalOutput = ""
22
+ model_id = "vikhyatk/moondream1"
23
+
24
+ if question == "":
25
+ question = "describe this image?"
26
+
27
+ # Check if the model is already loaded
28
+ try:
29
+ model
30
+ except NameError:
31
+ # clear_model(model) # Example of clearing the model
32
+ model = AutoModelForCausalLM.from_pretrained(model_id, trust_remote_code=True)
33
+ tokenizer = Tokenizer.from_pretrained(model_id)
34
+
35
+ # Assuming you have a correct way to process the image
36
+ #image = Image.open('/content/_57e22ed5-217c-4004-a279-eeecc18cbd55.jpg') #/content/Bard_Generated_Image (3).jpg')
37
+ # This part of the code is incorrect for a standard transformers model
38
+ enc_image = model.encode_image(image)
39
+ FinalOutput += model.answer_question(enc_image, "how many people are there? also explain if the image is weird?", tokenizer)
40
+
41
+ model_size = asizeof.asizeof(model)
42
+ tokenizer_size = asizeof.asizeof(tokenizer)
43
+ FinalOutput += f"\nModel size in RAM: {model_size} bytes, Tokenizer size in RAM: {tokenizer_size} bytes"
44
+
45
+ #model load and set-up = 1 min and inference on CPU = 2 min
46
+ return FinalOutput
47
+
48
+ # Define Gradio interface
49
+ iface = gr.Interface(fn=process_image_and_question,
50
+ inputs=[gr.Image(type="pil"), gr.Textbox(lines=2, placeholder="Ask a question about the image...")],
51
+ outputs="text",
52
+ title="Image Question Answering",
53
+ description="Upload an image and ask a question about it. ( 2 - 3 min response time expected )")
54
+
55
+ # Launch the interface
56
+ iface.launch()
requirements.txt ADDED
@@ -0,0 +1,7 @@
 
 
 
 
 
 
 
 
1
+ gradio
2
+ transformers
3
+ torch
4
+ Pillow
5
+ einops
6
+ timm #dependencies of moondream1 (Trust remote code is on)
7
+ pympler #analyze the memory behavior of Python objects