Spaces:
Running
on
Zero
Running
on
Zero
Update app.py
Browse files
app.py
CHANGED
@@ -2,6 +2,7 @@ import spaces
|
|
2 |
import gradio as gr
|
3 |
import torch
|
4 |
from transformers import PaliGemmaForConditionalGeneration, PaliGemmaProcessor, pipeline
|
|
|
5 |
import re
|
6 |
import random
|
7 |
import os
|
@@ -40,12 +41,35 @@ kolors_pipe.enable_model_cpu_offload()
|
|
40 |
vlm_model = PaliGemmaForConditionalGeneration.from_pretrained("gokaygokay/sd3-long-captioner-v2").to(device).eval()
|
41 |
vlm_processor = PaliGemmaProcessor.from_pretrained("gokaygokay/sd3-long-captioner-v2")
|
42 |
|
|
|
|
|
|
|
|
|
43 |
# Prompt Enhancer
|
44 |
enhancer_medium = pipeline("summarization", model="gokaygokay/Lamini-Prompt-Enchance", device=device)
|
45 |
enhancer_long = pipeline("summarization", model="gokaygokay/Lamini-Prompt-Enchance-Long", device=device)
|
46 |
|
47 |
MAX_SEED = 2**32 - 1
|
48 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
49 |
# VLM Captioner function
|
50 |
def create_captions_rich(image):
|
51 |
prompt = "caption en"
|
@@ -112,9 +136,12 @@ def generate_image(prompt, negative_prompt, seed, randomize_seed, width, height,
|
|
112 |
|
113 |
# Gradio Interface
|
114 |
@spaces.GPU
|
115 |
-
def process_workflow(image, text_prompt, use_vlm, use_enhancer, model_choice, negative_prompt, seed, randomize_seed, width, height, guidance_scale, num_inference_steps):
|
116 |
if use_vlm and image is not None:
|
117 |
-
|
|
|
|
|
|
|
118 |
else:
|
119 |
prompt = text_prompt
|
120 |
|
@@ -161,6 +188,7 @@ with gr.Blocks(css=custom_css, theme=gr.themes.Soft(primary_hue="blue", secondar
|
|
161 |
with gr.Group(elem_classes="input-group"):
|
162 |
input_image = gr.Image(label="Input Image for VLM")
|
163 |
use_vlm = gr.Checkbox(label="Use VLM Captioner", value=False)
|
|
|
164 |
|
165 |
with gr.Group(elem_classes="input-group"):
|
166 |
text_prompt = gr.Textbox(label="Text Prompt")
|
@@ -187,7 +215,7 @@ with gr.Blocks(css=custom_css, theme=gr.themes.Soft(primary_hue="blue", secondar
|
|
187 |
generate_btn.click(
|
188 |
fn=process_workflow,
|
189 |
inputs=[
|
190 |
-
input_image, text_prompt, use_vlm, use_enhancer, model_choice,
|
191 |
negative_prompt, seed, randomize_seed, width, height, guidance_scale, num_inference_steps
|
192 |
],
|
193 |
outputs=[output_image, final_prompt, used_seed]
|
|
|
2 |
import gradio as gr
|
3 |
import torch
|
4 |
from transformers import PaliGemmaForConditionalGeneration, PaliGemmaProcessor, pipeline
|
5 |
+
from transformers import AutoProcessor, AutoModelForCausalLM
|
6 |
import re
|
7 |
import random
|
8 |
import os
|
|
|
41 |
vlm_model = PaliGemmaForConditionalGeneration.from_pretrained("gokaygokay/sd3-long-captioner-v2").to(device).eval()
|
42 |
vlm_processor = PaliGemmaProcessor.from_pretrained("gokaygokay/sd3-long-captioner-v2")
|
43 |
|
44 |
+
# Initialize Florence model
|
45 |
+
florence_model = AutoModelForCausalLM.from_pretrained('microsoft/Florence-2-base', trust_remote_code=True).to(device).eval()
|
46 |
+
florence_processor = AutoProcessor.from_pretrained('microsoft/Florence-2-base', trust_remote_code=True)
|
47 |
+
|
48 |
# Prompt Enhancer
|
49 |
enhancer_medium = pipeline("summarization", model="gokaygokay/Lamini-Prompt-Enchance", device=device)
|
50 |
enhancer_long = pipeline("summarization", model="gokaygokay/Lamini-Prompt-Enchance-Long", device=device)
|
51 |
|
52 |
MAX_SEED = 2**32 - 1
|
53 |
|
54 |
+
# Florence caption function
|
55 |
+
def florence_caption(image):
|
56 |
+
inputs = florence_processor(text="<MORE_DETAILED_CAPTION>", images=image, return_tensors="pt").to(device)
|
57 |
+
generated_ids = florence_model.generate(
|
58 |
+
input_ids=inputs["input_ids"],
|
59 |
+
pixel_values=inputs["pixel_values"],
|
60 |
+
max_new_tokens=1024,
|
61 |
+
early_stopping=False,
|
62 |
+
do_sample=False,
|
63 |
+
num_beams=3,
|
64 |
+
)
|
65 |
+
generated_text = florence_processor.batch_decode(generated_ids, skip_special_tokens=False)[0]
|
66 |
+
parsed_answer = florence_processor.post_process_generation(
|
67 |
+
generated_text,
|
68 |
+
task="<MORE_DETAILED_CAPTION>",
|
69 |
+
image_size=(image.width, image.height)
|
70 |
+
)
|
71 |
+
return parsed_answer["<MORE_DETAILED_CAPTION>"]
|
72 |
+
|
73 |
# VLM Captioner function
|
74 |
def create_captions_rich(image):
|
75 |
prompt = "caption en"
|
|
|
136 |
|
137 |
# Gradio Interface
|
138 |
@spaces.GPU
|
139 |
+
def process_workflow(image, text_prompt, use_vlm, use_enhancer, model_choice, vlm_model_choice, negative_prompt, seed, randomize_seed, width, height, guidance_scale, num_inference_steps):
|
140 |
if use_vlm and image is not None:
|
141 |
+
if vlm_model_choice == "Long Captioner":
|
142 |
+
prompt = create_captions_rich(image)
|
143 |
+
else: # Florence
|
144 |
+
prompt = florence_caption(image)
|
145 |
else:
|
146 |
prompt = text_prompt
|
147 |
|
|
|
188 |
with gr.Group(elem_classes="input-group"):
|
189 |
input_image = gr.Image(label="Input Image for VLM")
|
190 |
use_vlm = gr.Checkbox(label="Use VLM Captioner", value=False)
|
191 |
+
vlm_model_choice = gr.Radio(["Long Captioner", "Florence"], label="VLM Model", value="Long Captioner")
|
192 |
|
193 |
with gr.Group(elem_classes="input-group"):
|
194 |
text_prompt = gr.Textbox(label="Text Prompt")
|
|
|
215 |
generate_btn.click(
|
216 |
fn=process_workflow,
|
217 |
inputs=[
|
218 |
+
input_image, text_prompt, use_vlm, use_enhancer, model_choice, vlm_model_choice,
|
219 |
negative_prompt, seed, randomize_seed, width, height, guidance_scale, num_inference_steps
|
220 |
],
|
221 |
outputs=[output_image, final_prompt, used_seed]
|