duplicaction test
Browse files
app.py
CHANGED
@@ -1,3 +1,5 @@
|
|
|
|
|
|
1 |
import gradio as gr
|
2 |
import torch
|
3 |
from transformers import AutoProcessor, LlavaForConditionalGeneration
|
@@ -24,10 +26,16 @@ model = LlavaForConditionalGeneration.from_pretrained(
|
|
24 |
)
|
25 |
|
26 |
|
27 |
-
def text_to_image(image, prompt):
|
28 |
prompt = f'USER: <image>\n{prompt}\nASSISTANT:'
|
29 |
|
30 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
31 |
output = model.generate(**inputs, max_new_tokens=500, temperature=0.3)
|
32 |
generated_text = processor.batch_decode(output, skip_special_tokens=True)
|
33 |
text = generated_text.pop()
|
@@ -41,7 +49,8 @@ demo = gr.Interface(
|
|
41 |
fn=text_to_image,
|
42 |
inputs=[
|
43 |
gr.Image(label='Select an image to analyze', type='pil'),
|
44 |
-
gr.Textbox(label='Enter Prompt')
|
|
|
45 |
],
|
46 |
outputs=[gr.Textbox(label='Maurice says:'), gr.JSON(label='Embedded text')]
|
47 |
)
|
|
|
1 |
+
from copy import deepcopy
|
2 |
+
|
3 |
import gradio as gr
|
4 |
import torch
|
5 |
from transformers import AutoProcessor, LlavaForConditionalGeneration
|
|
|
26 |
)
|
27 |
|
28 |
|
29 |
+
def text_to_image(image, prompt, duplications: int):
|
30 |
prompt = f'USER: <image>\n{prompt}\nASSISTANT:'
|
31 |
|
32 |
+
image_batch = [image]
|
33 |
+
prompt_batch = [prompt]
|
34 |
+
for _ in range(duplications):
|
35 |
+
image_batch.append(deepcopy(image))
|
36 |
+
prompt_batch.append(prompt)
|
37 |
+
|
38 |
+
inputs = processor(prompt_batch, images=image_batch, padding=True, return_tensors="pt").to(model.device)
|
39 |
output = model.generate(**inputs, max_new_tokens=500, temperature=0.3)
|
40 |
generated_text = processor.batch_decode(output, skip_special_tokens=True)
|
41 |
text = generated_text.pop()
|
|
|
49 |
fn=text_to_image,
|
50 |
inputs=[
|
51 |
gr.Image(label='Select an image to analyze', type='pil'),
|
52 |
+
gr.Textbox(label='Enter Prompt'),
|
53 |
+
gr.Number(label='How many duplications of the image (to test memory load)', value=0)
|
54 |
],
|
55 |
outputs=[gr.Textbox(label='Maurice says:'), gr.JSON(label='Embedded text')]
|
56 |
)
|