Update app.py

app.py

@@ -1,4 +1,3 @@
-import urllib
 import requests
 from bs4 import BeautifulSoup
 from abc import ABC, abstractmethod
@@ -17,10 +16,8 @@ import requests
 import sys
 import warnings
 import time
-import asyncio
 import math
 from pathlib import Path
-from functools import partial
 from dataclasses import dataclass
 from typing import Any
 import pillow_heif
@@ -44,11 +41,6 @@ from refiners.foundationals.latent_diffusion.stable_diffusion_1.multi_upscaler i
 )
 from datetime import datetime
 
-_HEIGHT_ = None
-_WIDTH_ = None
-
-working = False
-
 model = T5ForConditionalGeneration.from_pretrained("t5-large")
 tokenizer = T5Tokenizer.from_pretrained("t5-large")
 
@@ -445,7 +437,7 @@ MAX_SEED = np.iinfo(np.int32).max
 # precision data
 
 seq=512
-image_steps=50
+image_steps=40
 img_accu=6.5
 
 # ui data
@@ -481,29 +473,6 @@ footer {
 }
 """])
 
-js="""
-function custom(){
-    document.querySelector("div#prompt input").addEventListener("keydown",function(e){
-        e.target.setAttribute("last_value",e.target.value);
-    });
-    document.querySelector("div#prompt input").addEventListener("input",function(e){
-        if( e.target.value.toString().match(/[^ a-zA-Z,]|( |,){2,}/gsm) ){
-            e.target.value = e.target.getAttribute("last_value");
-            e.target.removeAttribute("last_value");
-        }
-    });
-
-    document.querySelector("div#prompt2 input").addEventListener("keydown",function(e){
-        e.target.setAttribute("last_value",e.target.value);
-    });
-    document.querySelector("div#prompt2 input").addEventListener("input",function(e){
-        if( e.target.value.toString().match(/[^ a-zA-Z,]|( |,){2,}/gsm) ){
-            e.target.value = e.target.getAttribute("last_value");
-            e.target.removeAttribute("last_value");
-        }
-    });
-}
-"""
 
 # torch pipes
 
@@ -514,7 +483,22 @@ torch.cuda.empty_cache()
 
 # functionality
 
-@spaces.GPU(duration=240)
+@spaces.GPU(duration=180)
+def hard_scaler(img):
+    return upscaler(img)
+
+@spaces.GPU(duration=90)
+def easy_scaler(img):
+    return upscaler(img)
+
+def handle_upscaler(img):
+    
+    w, h = img.size
+    if w*h > 2 * (10 ** 6):
+        return hard_scaler(img)
+    return easy_scaler(img)
+
+@spaces.GPU(duration=180)
 def upscaler(
     input_image: Image.Image,
     prompt: str = "Accurate, Highly Detailed, Realistic, Best Quality, Hyper-Realistic, Super-Realistic, Natural, Reasonable, Logical.",
@@ -527,7 +511,7 @@ def upscaler(
     tile_width: int = 112,
     tile_height: int = 144,
     denoise_strength: float = 0.35,
-    num_inference_steps: int = 30,
+    num_inference_steps: int = 20,
     solver: str = "DDIM",
 ) -> Image.Image:
 
@@ -572,7 +556,7 @@ def _summarize(text):
     gen = model.generate(
         toks,
         length_penalty=0.1,
-        num_beams=32,
+        num_beams=6,
         early_stopping=True,
         max_length=512
     )
@@ -1262,11 +1246,47 @@ def translate(txt,to_lang="en",from_lang="auto"):
     log(f'RET translate with translation as {translation}')
     return translation.lower()
 
-@spaces.GPU(duration=240)
 def handle_generation(h,w,d):
 
     log(f'CALL handle_generate')
 
+    difficulty_points = 0
+
+    toks_len = get_tensor_length(tokenizer.encode( d, return_tensors="pt", truncation=False))
+    if toks_len > 500:
+        difficulty_points + 2
+    else if toks_len > 50:
+        difficulty_points + 1
+
+    pxs = h*w
+    if pxs > 4 * (10 ** 6):
+        difficulty_points + 3
+    else if pxs > 3 * (10 ** 6):
+        difficulty_points + 2
+    else if pxs > 2 * (10 ** 6):
+        difficulty_points + 1
+
+    if difficulty_points < 2:
+        return easy_generation(h,w,d)
+    else if difficulty_points < 4:
+        return balanced_generation(h,w,d)
+    else:
+        return hard_generation(h,w,d)
+
+@spaces.GPU(duration=90)
+def easy_generation(h,w,d):
+    return generation(h,w,d)
+
+@spaces.GPU(duration=150)
+def balanced_generation(h,w,d):
+    return generation(h,w,d)
+
+@spaces.GPU(duration=210)
+def hard_generation(h,w,d):
+    return generation(h,w,d)
+
+def generation(h,w,d):
+    
     if len(d) > 0:
         d = re.sub(r",( ){1,}",". ",d)
         d_lines = re.split(r"([\n]){1,}", d)
@@ -1282,7 +1302,7 @@ def handle_generation(h,w,d):
         d = re.sub(r"([ \t]){1,}", " ", d)
         d = re.sub(r"(\. \.)", ".", d)
 
-    neg = f"Textual, Text, Sign, Label, Title, Unreal, Exceptional, Irregular, Unusual, Blurry, Smoothed, Polished, Worst Quality, Worse Quality, Painted, Movies Quality."
+    neg = f"Textual, Text, Signs, Labels, Titles, Unreal, Exceptional, Irregular, Unusual, Blurry, Smoothed, Polished, Worst Quality, Worse Quality, Painted, Movies Quality."
     q = "\""
     pos = f'Accurate, Detailed, Realistic.{ "" if d == "" else " " + d }'
 
@@ -1324,17 +1344,17 @@ if __name__ == "__main__":
                 height = gr.Slider(
                     label="Height (px)",
                     minimum=512,
-                    maximum=1024,
+                    maximum=2048,
                     step=16,
-                    value=768,
+                    value=1024,
                 )
 
                 width = gr.Slider(
                     label="Width (px)",
                     minimum=512,
-                    maximum=1024,
+                    maximum=2048,
                     step=16,
-                    value=768,
+                    value=1024,
                 )
 
                 run = gr.Button("Generate",elem_classes="btn")