app.py

import os
import gradio as gr
from random import randint
from operator import itemgetter
import bisect
from all_models import tags_plus_models,models,models_plus_tags
from datetime import datetime
from externalmod import gr_Interface_load
import asyncio
import os
from threading import RLock
lock = RLock()
HF_TOKEN = os.environ.get("HF_TOKEN") if os.environ.get("HF_TOKEN") else None # If private or gated models aren't used, ENV setting is unnecessary.


now2 = 0
inference_timeout = 300
MAX_SEED = 2**32-1


nb_rep=2
nb_mod_dif=20
nb_models=nb_mod_dif*nb_rep

cache_image={}
cache_image_actu={}


def load_fn(models):
    global models_load
    global num_models
    global default_models
    models_load = {}
    num_models = len(models)
    i=0
    if num_models!=0:
        default_models = models[:num_models]
    else:
        default_models = {}
    for model in models:
        i+=1
        if i%50==0:
            print("\n\n\n-------"+str(i)+'/'+str(len(models))+"-------\n\n\n")
        if model not in models_load.keys():
            try:
                m = gr_Interface_load(f'models/{model}', hf_token=HF_TOKEN)
            except Exception as error:
                m = gr.Interface(lambda txt: None, ['text'], ['image'])
                print(error)
            models_load.update({model: m})


load_fn(models)
    
def test_pass(test):
    if test==os.getenv('p'):
        print("ok")
        return gr.Dropdown(label="Lists Tags", show_label=True, choices=list(models_test) , interactive = True)
    else:
        print("nop")
        return gr.Dropdown(label="Lists Tags", show_label=True, choices=list([]) , interactive = True)

def test_pass_aff(test):
    if test==os.getenv('p'):
        return gr.Accordion( open=True, visible=True) ,gr.Row(visible=False)
    else:
        return gr.Accordion( open=True, visible=False) , gr.Row()


# https://huggingface.co/docs/api-inference/detailed_parameters
# https://huggingface.co/docs/huggingface_hub/package_reference/inference_client
async def infer(model_str, prompt, nprompt="", height=None, width=None, steps=None, cfg=None, seed=-1, timeout=inference_timeout):
    from pathlib import Path
    kwargs = {}
    if height is not None and height >= 256: kwargs["height"] = height
    if width is not None and width >= 256: kwargs["width"] = width
    if steps is not None and steps >= 1: kwargs["num_inference_steps"] = steps
    if cfg is not None and cfg > 0: cfg = kwargs["guidance_scale"] = cfg
    noise = ""
    if seed >= 0: kwargs["seed"] = seed
    else:
        rand = randint(1, 500)
        for i in range(rand):
            noise += " "
    task = asyncio.create_task(asyncio.to_thread(models_load[model_str].fn,
                               prompt=f'{prompt} {noise}', negative_prompt=nprompt, **kwargs, token=HF_TOKEN))
    await asyncio.sleep(0)
    try:
        result = await asyncio.wait_for(task, timeout=timeout)
    except (Exception, asyncio.TimeoutError) as e:
        print(e)
        print(f"Task timed out: {model_str}")
        if not task.done(): task.cancel()
        result = None
    if task.done() and result is not None:
        with lock:
            png_path = "image.png"
            result.save(png_path)
            image = str(Path(png_path).resolve())
        return image
    return None

def gen_fn(model_str, prompt, nprompt="", height=None, width=None, steps=None, cfg=None, seed=-1):
    if model_str == 'NA':
        return None
    try:
        loop = asyncio.new_event_loop()
        result = loop.run_until_complete(infer(model_str, prompt, nprompt,
                                         height, width, steps, cfg, seed, inference_timeout))
    except (Exception, asyncio.CancelledError) as e:
        print(e)
        print(f"Task aborted: {model_str}")
        result = None
    finally:
        loop.close()
    return result

   
def add_gallery(image, model_str, gallery):
    if gallery is None: gallery = []
    #with lock:
    if image is not None: gallery.append((image, model_str))
    return gallery

def reset_gallery(gallery):
    return add_gallery(None,"",[])

def load_gallery(gallery,id):
    gallery = reset_gallery(gallery)
    for c in cache_image[f"{id}"]:
        gallery=add_gallery(c[0],c[1],gallery)
    return gallery
def load_gallery_sorted(gallery,id):
    gallery = reset_gallery(gallery)
    for c in sorted(cache_image[f"{id}"], key=itemgetter(1)):
        gallery=add_gallery(c[0],c[1],gallery)
    return gallery
def load_gallery_actu(gallery,id):
    gallery = reset_gallery(gallery)
    for c in cache_image_actu[f"{id}"]:
        gallery=add_gallery(c[0],c[1],gallery)
    return gallery

def add_cache_image(image, model_str,id,cache_image=cache_image):
    if image is not None:
        cache_image[f"{id}"].append((image,model_str))
    #cache_image=sorted(cache_image, key=itemgetter(1))
    return 
def add_cache_image_actu(image, model_str,id,cache_image_actu=cache_image_actu):
    if image is not None:
        bisect.insort(cache_image_actu[f"{id}"],(image, model_str), key=itemgetter(1))
    #cache_image_actu=sorted(cache_image_actu, key=itemgetter(1))
    return
def reset_cache_image(id,cache_image=cache_image):
    cache_image[f"{id}"].clear()
    return 
def reset_cache_image_actu(id,cache_image_actu=cache_image_actu):
    cache_image_actu[f"{id}"].clear()
    return 
def reset_cache_image_all_sessions(cache_image=cache_image,cache_image_actu=cache_image_actu):
    for key, listT in cache_image.items():
        listT.clear()
    for key, listT in cache_image_actu.items():
        listT.clear()
    return 
    
def set_session(id):
    if id==0:
        randTemp=randint(1,MAX_SEED)
        cache_image[f"{randTemp}"]=[]
        cache_image_actu[f"{randTemp}"]=[]
        return gr.Number(visible=False,value=randTemp)
    else :
        return id
def print_info_sessions():
    lenTot=0
    print("###################################")
    print("number of sessions : "+str(len(cache_image)))
    for key, listT in cache_image.items():
        print("session "+key+" : "+str(len(listT)))
        lenTot+=len(listT)
    print("images total = "+str(lenTot))
    print("###################################")
    return

def disp_models(group_model_choice,nb_rep=nb_rep):
    listTemp=[]
    strTemp='\n'
    i=0
    for m in group_model_choice:
        if m not in listTemp:
            listTemp.append(m)
    for m in listTemp:
        i+=1
        strTemp+="\"" + m + "\",\n"
        if i%(8/nb_rep)==0:
            strTemp+="\n"
    return gr.Textbox(label="models",value=strTemp)

def search_models(str_search,tags_plus_models=tags_plus_models):
    output1="\n"
    output2=""
    for m in tags_plus_models[0][2]:
        if m.find(str_search)!=-1:
            output1+="\"" + m + "\",\n"
    outputPlus="\n From tags : \n\n"
    for tag_plus_models in tags_plus_models:
        if str_search.lower() == tag_plus_models[0].lower() and str_search!="":
            for m in tag_plus_models[2]:
                output2+="\"" + m + "\",\n"
    if output2 != "":
        output=output1+outputPlus+output2
    else :
        output=output1
    return gr.Textbox(label="out",value=output)

def search_info(txt_search_info,models_plus_tags=models_plus_tags):
    outputList=[]
    if txt_search_info.find("\"")!=-1:
        start=txt_search_info.find("\"")+1
        end=txt_search_info.find("\"",start)
        m_name=cutStrg(txt_search_info,start,end)
    else :
        m_name = txt_search_info
    for m in models_plus_tags:
        if m_name == m[0]:
            outputList=m[1]
    if len(outputList)==0:
        outputList.append("Model Not Find")
    return gr.Textbox(label="out",value=outputList)


def ratio_chosen(choice_ratio,width,height):
    if choice_ratio == [None,None]:
        return width , height
    else :
        return gr.Slider(label="Width", info="If 0, the default value is used.", maximum=2024, step=32, value=choice_ratio[0]), gr.Slider(label="Height", info="If 0, the default value is used.", maximum=2024, step=32, value=choice_ratio[1])

list_ratios=[["None",[None,None]],
             ["4:1 (2048 x 512)",[2048,512]],
             ["12:5 (1536 x 640)",[1536,640]],
             ["~16:9 (1344 x 768)",[1344,768]],
             ["~3:2 (1216 x 832)",[1216,832]],
             ["~4:3 (1152 x 896)",[1152,896]],
             ["1:1 (1024 x 1024)",[1024,1024]],
             ["~3:4 (896 x 1152)",[896,1152]],
             ["~2:3 (832 x 1216)",[832,1216]],
             ["~9:16 (768 x 1344)",[768,1344]],
             ["5:12 (640 x 1536)",[640,1536]],
             ["1:4 (512 x 2048)",[512,2048]]]

def fonc_add_param(lp,txt_input,neg_input,width,height,steps,cfg,seed):
    lp.append([txt_input,neg_input,width,height,steps,cfg,seed])
    return gr.Dataset(samples=lp) , gr.Dropdown(choices=[["a",lp]], value=lp)
def fonc_del_param(lp,txt_input,neg_input,width,height,steps,cfg,seed):
    if [txt_input,neg_input,width,height,steps,cfg,seed] in lp :
        lp.remove([txt_input,neg_input,width,height,steps,cfg,seed])
    return gr.Dataset(samples=lp) , gr.Dropdown(choices=[["a",lp]], value=lp)
            
def make_me():
        with gr.Row():
            with gr.Column(scale=4):
                with gr.Group():
                    txt_input = gr.Textbox(label='Your prompt:', lines=3, interactive = True)
                    with gr.Accordion("Advanced", open=False, visible=True):
                        neg_input = gr.Textbox(label='Negative prompt:', lines=1, interactive = True)
                        with gr.Row():
                            width = gr.Slider(label="Width", info="If 0, the default value is used.", maximum=2024, step=32, value=0, interactive = True)
                            height = gr.Slider(label="Height", info="If 0, the default value is used.", maximum=2024, step=32, value=0, interactive = True)
                        with gr.Row():
                            choice_ratio = gr.Dropdown(label="Ratio Width/Height", 
                                                info="OverWrite Width and Height (W*H<1024*1024)", 
                                                show_label=True, choices=list(list_ratios) , interactive = True, value=list_ratios[0][1])
                            choice_ratio.change(ratio_chosen,[choice_ratio,width,height],[width,height])
                        with gr.Row():
                            steps = gr.Slider(label="Number of inference steps", info="If 0, the default value is used.", maximum=100, step=1, value=0, interactive = True)
                            cfg = gr.Slider(label="Guidance scale", info="If 0, the default value is used.", maximum=30.0, step=0.1, value=0, interactive = True)
                            seed = gr.Slider(label="Seed", info="Randomize Seed if -1.", minimum=-1, maximum=MAX_SEED, step=1, value=-1, interactive = True)
            param_actu=[txt_input,neg_input,width,height,steps,cfg,seed]
            list_param=gr.Dropdown(choices=[["a",[]]], value=[], visible=False)
            
            add_param=gr.Button("add to the list")
            del_param=gr.Button("delete to the list")
            #gen_button = gr.Button('Generate images', scale=3)
            #stop_button = gr.Button('Stop', variant='secondary', interactive=False, scale=1)
            #gen_button.click(lambda: gr.update(interactive=True), None, stop_button)

    
        disp_param = gr.Examples(
            examples=list_param.value,
            inputs=[txt_input,neg_input,width,height,steps,cfg,seed],
            outputs=[txt_input,neg_input,width,height,steps,cfg,seed],
        )
    
        add_param.click(fonc_add_param,[list_param,txt_input,neg_input,width,height,steps,cfg,seed],[disp_param.dataset,list_param])
        del_param.click(fonc_del_param,[list_param,txt_input,neg_input,width,height,steps,cfg,seed],[disp_param.dataset,list_param])
    

js_code = """
    
    console.log('ghgh');
"""

with gr.Blocks(theme="Nymbo/Nymbo_Theme", fill_width=True, css="div.float.svelte-1mwvhlq {    position: absolute;    top: var(--block-label-margin);    left: var(--block-label-margin);    background: none;    border: none;}") as demo: 
    gr.Markdown("<script>" + js_code + "</script>")
    make_me()


# https://www.gradio.app/guides/setting-up-a-demo-for-maximum-performance
#demo.queue(concurrency_count=999) # concurrency_count is deprecated in 4.x
demo.queue(default_concurrency_limit=200, max_size=200)
demo.launch(max_threads=400)