Update app.py

app.py

@@ -7,7 +7,6 @@ Original file is located at
     https://colab.research.google.com/drive/1RxVCwOkq3Q5qlEkQxhFGeUxICBujjEjR
 """
 
-
 from transformers import AutoTokenizer, AutoModelForSeq2SeqLM
 
 tokenizer = AutoTokenizer.from_pretrained("Helsinki-NLP/opus-mt-zh-en")
@@ -24,10 +23,17 @@ from diffusers import StableDiffusionControlNetPipeline, ControlNetModel, UniPCM
 import torch
 from controlnet_aux import HEDdetector
 from diffusers.utils import load_image
+
 import concurrent.futures
 from threading import Thread
 from compel import Compel
 
+from transformers import pipeline
+
+
+model_ckpt = "papluca/xlm-roberta-base-language-detection"
+pipe = pipeline("text-classification", model=model_ckpt)
+
 
 
 device="cuda" if torch.cuda.is_available() else "cpu"
@@ -37,6 +43,9 @@ hidden_negative = "big boobs, huge boobs, sexy, dirty, d cup, e cup, g cup, slut
 hidden_cn_booster_text = "漂亮的脸,小胸,贫乳,a罩杯"
 hidden_cn_negative = "大胸, ，, ！, 。, ；,巨乳,性感,脏,d罩杯,e罩杯,g罩杯,骚,骚气,badhandv4, ng_deepnegative_v1_75t"
 
+# text_tokenizer = CLIPTokenizer.from_pretrained("IDEA-CCNL/Taiyi-Stable-Diffusion-1B-Chinese-EN-v0.1")
+# text_encoder = CLIPTextModel.from_pretrained("IDEA-CCNL/Taiyi-Stable-Diffusion-1B-Chinese-EN-v0.1").eval()
+# text_encoder = ChineseCLIPModel.from_pretrained("OFA-Sys/chinese-clip-vit-base-patch16").eval()
 
 def translate(prompt):
     trans_text = prompt
@@ -85,6 +94,7 @@ pipe_scribble.safety_checker = None
 pipe_scribble.requires_safety_checker = False
 pipe_scribble.to(device)
 
+text = "handsome doctor, black eyes, sparkling eyes, smiling, handsome, glasses, blue tie with yellow dots, doctor's white coat, white collar, blue gradient background"
 
 def scribble_to_image(text, input_img, chinese_check):
     """
@@ -92,13 +102,19 @@ def scribble_to_image(text, input_img, chinese_check):
     include Adetailer, detail tweaker lora, prompt backend include: beautiful eyes, beautiful face, beautiful hand, (maybe infer from user's prompt for gesture and facial
     expression to improve hand)
     """
+
+
+
+# if auto detect detects chinese => auto turn on chinese prompting checkbox
     # change param "bag" below to text, image param below to input_img
     input_img = Image.fromarray(input_img)
     input_img = hed(input_img, scribble=True)
     input_img = load_image(input_img)
     # global prompt
     compel_proc = Compel(tokenizer=pipe_scribble.tokenizer, text_encoder=pipe_scribble.text_encoder)
-    if chinese_check:
+    lang_check_label = pipe(text, top_k=1, truncation=True)[0]['label']
+    lang_check_score = pipe(text, top_k=1, truncation=True)[0]['score']
+    if lang_check_label == 'zh' and lang_check_score >= 0.85:
         text = translate(text)
         print("prompt text:", text)
     prompt = text + hidden_booster_text
@@ -111,6 +127,7 @@ def scribble_to_image(text, input_img, chinese_check):
 
     return res_image0, res_image1, res_image2, res_image3
 
+
 from PIL import Image
 
 from transformers import pipeline
@@ -140,6 +157,7 @@ pipe_img2img.safety_checker = None
 pipe_img2img.requires_safety_checker = False
 pipe_img2img.to(device)
 
+
 def real_img2img_to_anime(text, input_img, chinese_check):
     """
     pass the sd model and do scribble to image
@@ -219,6 +237,7 @@ with gr.Blocks(theme=theme, css="footer {visibility: hidden}", title="ShellAI Ap
                 image_box = gr.Image(label="Input Image", height=350)
                 gen_btn = gr.Button(value="Generate")
         with gr.Row(equal_height=True):
+
             image1 = gr.Image()
             image2 = gr.Image()
             image3 = gr.Image()
@@ -228,8 +247,9 @@ with gr.Blocks(theme=theme, css="footer {visibility: hidden}", title="ShellAI Ap
             with concurrent.futures.ThreadPoolExecutor(max_workers=12000) as executor:
                 future = executor.submit(scribble_to_image, prompt_box, image_box, chinese_check)
                 image1, image2, image3, image4 = future.result()
+
             return image1, image2, image3, image4
 
         gen_btn.click(mult_thread, [prompt_box, image_box, chinese_check], [image1, image2, image3, image4])
 
-iface.launch(debug=True, share=True, auth=["user", "1qh-32ba"])
+iface.launch(debug=True, share=True, auth=["shenrym", "shjdqw%23-sw2&"])