Upload 2 files

app.py

@@ -1,19 +1,21 @@
 import gradio as gr
-from random import randint
+
 from all_models import models
-from externalmod import gr_Interface_load
+from externalmod import gr_Interface_load, save_image, randomize_seed
 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.
+
 
 def load_fn(models):
     global models_load
     models_load = {}
-    
     for model in models:
         if model not in models_load.keys():
             try:
-                m = gr_Interface_load(f'models/{model}')
+                m = gr_Interface_load(f'models/{model}', hf_token=HF_TOKEN)
             except Exception as error:
                 print(error)
                 m = gr.Interface(lambda: None, ['text'], ['image'])
@@ -24,8 +26,10 @@ load_fn(models)
 
 
 num_models = 6
+timeout = 600
 default_models = models[:num_models]
-timeout = 300
+MAX_SEED = 3999999999
+
 
 def extend_choices(choices):
     return choices[:num_models] + (num_models - len(choices[:num_models])) * ['NA']
@@ -33,45 +37,58 @@ def extend_choices(choices):
 
 def update_imgbox(choices):
     choices_plus = extend_choices(choices[:num_models])
-    return [gr.Image(None, label = m, visible = (m != 'NA')) for m in choices_plus]
-
-
-def update_imgbox_gallery(choices):
-    choices_plus = extend_choices(choices[:num_models])
-    return [gr.Gallery(None, label = m, visible = (m != 'NA')) for m in choices_plus]
-
-
-async def infer(model_str, prompt, timeout):
-    from PIL import Image
-    noise = ""
-    rand = randint(1, 500)
-    for i in range(rand):
-        noise += " "
-    task = asyncio.create_task(asyncio.to_thread(models_load[model_str], f'{prompt} {noise}'))
+    return [gr.Image(None, label=m, visible=(m!='NA')) for m in choices_plus]
+
+
+def random_choices():
+    import random
+    random.seed()
+    return random.choices(models, k=561)
+
+
+# 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=0, width=0, cfg=0, seed=-1, timeout=inference_timeout):
+    kwargs = {}
+    if height > 0: kwargs["height"] = height
+    if width > 0: kwargs["width"] = width
+    if cfg > 0: cfg = kwargs["guidance_scale"] = cfg
+    if seed == -1: kwargs["seed"] = randomize_seed()
+    else: kwargs["seed"] = seed
+    task = asyncio.create_task(asyncio.to_thread(models_load[model_str].fn,
+                               prompt=prompt, 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:
+    except 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:
+        raise Exception(f"Task timed out: {model_str}") from e
+    except Exception as e:
+        print(e)
+        if not task.done(): task.cancel()
+        result = None
+        raise Exception() from e
+    if task.done() and result is not None and not isinstance(result, tuple):
         with lock:
-            image = Image.open(result).convert('RGBA')
+            png_path = "image.png"
+            image = save_image(result, png_path, model_str, prompt, nprompt, height, width, steps, cfg, seed)
         return image
     return None
 
-def gen_fn(model_str, prompt):
-    if model_str == 'NA':
-        return None
+
+def gen_fn(model_str, prompt, nprompt="", height=0, width=0, steps=0, cfg=0, seed=-1):
     try:
         loop = asyncio.new_event_loop()
-        result = loop.run_until_complete(infer(model_str, prompt, timeout))
+        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
+        raise gr.Error(f"Task aborted: {model_str}, Error: {e}")
     finally:
         loop.close()
     return result
@@ -84,53 +101,48 @@ def add_gallery(image, model_str, gallery):
     return gallery
 
 
-def gen_fn_gallery(model_str, prompt, gallery):
-    if gallery is None: gallery = []
-    if model_str == 'NA':
-        yield gallery
-    try:
-        loop = asyncio.new_event_loop()
-        result = loop.run_until_complete(infer(model_str, prompt, timeout))
-        with lock:
-            if result: gallery.insert(0, result)
-    except (Exception, asyncio.CancelledError) as e:
-        print(e)
-        print(f"Task aborted: {model_str}")
-    finally:
-        loop.close()
-    yield gallery
-
-
 CSS="""
-#container { max-width: 1200px; margin: 0 auto; !important; }
-.output { width=112px; height=112px; !important; }
-.gallery { width=100%; min_height=768px; !important; }
+.gradio-container { max-width: 1200px; margin: 0 auto; !important; }
+.output { width=112px; height=112px; max_width=112px; max_height=112px; !important; }
+.gallery { min_width=512px; min_height=512px; max_height=1024px; !important; }
 .guide { text-align: center; !important; }
 """
 
+
 with gr.Blocks(theme='Nymbo/Nymbo_Theme', fill_width=True, css=CSS) as demo:
     gr.HTML(
     """
         <div>
-        <p> <center>For simultaneous generations without hidden queue check out <a href="https://huggingface.co/spaces/Yntec/ToyWorld">Toy World</a>! For more options like single model x6 check out <a href="https://huggingface.co/spaces/John6666/Diffusion80XX4sg">Diffusion80XX4sg</a> by John6666!</center>
+        <p> <center>For more options like single model x6 check out <a href="https://huggingface.co/spaces/John6666/Diffusion80XX4sg">Diffusion80XX4sg</a> by John6666!</center>
         </p></div>
     """
 )  
     with gr.Tab('Huggingface Diffusion'):
         with gr.Column(scale=2):
-            txt_input = gr.Textbox(label='Your prompt:', lines=4)
+            with gr.Group():
+                txt_input = gr.Textbox(label='Your prompt:', lines=4)
+                neg_input = gr.Textbox(label='Negative prompt:', lines=1)
+                with gr.Accordion("Advanced", open=False, visible=True):
+                    with gr.Row():
+                        width = gr.Slider(label="Width", info="If 0, the default value is used.", maximum=1216, step=32, value=0)
+                        height = gr.Slider(label="Height", info="If 0, the default value is used.", maximum=1216, step=32, value=0)
+                    with gr.Row():
+                        cfg = gr.Slider(label="Guidance scale", info="If 0, the default value is used.", maximum=30.0, step=0.1, value=0)
+                        seed = gr.Slider(label="Seed", info="Randomize Seed if -1.", minimum=-1, maximum=MAX_SEED, step=1, value=-1)
+                        seed_rand = gr.Button("Randomize Seed ??", size="sm", variant="secondary")
+                        seed_rand.click(randomize_seed, None, [seed], queue=False)
             with gr.Row():
-                gen_button = gr.Button(f'Generate up to {int(num_models)} images from 1 to {int(num_models)*3} minutes total', scale=2)
-                stop_button = gr.Button('Stop', variant='secondary', interactive=False, scale=1)
-                gen_button.click(lambda: gr.update(interactive = True), None, stop_button)
+                gen_button = gr.Button(f'Generate up to {int(num_models)} images from 10 seconds to 3 minutes total', variant='primary', scale=3)
+                random_button = gr.Button(f'Random {int(num_models)} ??', variant='secondary', scale=1)
+                #gen_button.click(lambda: gr.update(interactive = True), None, stop_button)
             gr.Markdown("Scroll down to see more images and select models.", elem_classes="guide")
 
         with gr.Column(scale=1):
             with gr.Group():
                 with gr.Row():
-                    output = [gr.Image(label=m, show_download_button=True, elem_classes="output", interactive=False, min_width=80, show_share_button=False, visible=True) for m in default_models]
-                    #output = [gr.Image(label=m, show_download_button=True, elem_classes="output", interactive=False, show_share_button=True) for m in default_models]
-                    #output = [gr.Gallery(label=m, show_download_button=True, elem_classes="output", interactive=False, show_share_button=True, container=True, format="png", object_fit="cover") for m in default_models]
+                    output = [gr.Image(label=m, show_download_button=True, elem_classes="output",
+                              interactive=False, min_width=80, show_share_button=False, format="png",
+                              visible=True) for m in default_models]
                     current_models = [gr.Textbox(m, visible=False) for m in default_models]
 
         with gr.Column(scale=2):
@@ -139,20 +151,21 @@ with gr.Blocks(theme='Nymbo/Nymbo_Theme', fill_width=True, css=CSS) as demo:
                                 preview=True, object_fit="cover", columns=2, rows=2) 
 
         for m, o in zip(current_models, output):
-            #gen_event = gen_button.click(gen_fn, [m, txt_input], o)
-            #gen_event = gen_button.click(gen_fn_gallery, [m, txt_input, o], o)
-            gen_event = gr.on(triggers=[gen_button.click, txt_input.submit], fn=gen_fn, inputs=[m, txt_input], outputs=[o])
+            gen_event = gr.on(triggers=[gen_button.click, txt_input.submit], fn=gen_fn,
+                              inputs=[m, txt_input, neg_input, height, width, cfg, seed], outputs=[o],
+                              concurrency_limit=None, queue=False) # Be sure to delete ", queue=False" when activating the stop button
             o.change(add_gallery, [o, m, gallery], [gallery])
-            stop_button.click(lambda: gr.update(interactive = False), None, stop_button, cancels = [gen_event])
+            #stop_button.click(lambda: gr.update(interactive = False), None, stop_button, cancels = [gen_event])
 
         with gr.Column(scale=4):
             with gr.Accordion('Model selection'):
                 model_choice = gr.CheckboxGroup(models, label = f'Choose up to {int(num_models)} different models from the {len(models)} available!', value=default_models, interactive=True)
                 model_choice.change(update_imgbox, model_choice, output)
-                #model_choice.change(update_imgbox_gallery, model_choice, output)
                 model_choice.change(extend_choices, model_choice, current_models)
+                random_button.click(random_choices, None, model_choice)
 
     gr.Markdown("Based on the [TestGen](https://huggingface.co/spaces/derwahnsinn/TestGen) Space by derwahnsinn, the [SpacIO](https://huggingface.co/spaces/RdnUser77/SpacIO_v1) Space by RdnUser77 and Omnibus's Maximum Multiplier!")
 
-demo.queue()
-demo.launch()
+demo.queue(default_concurrency_limit=200, max_size=200)
+demo.launch(show_api=False, max_threads=400)
+# https://github.com/gradio-app/gradio/issues/6339

externalmod.py

@@ -9,7 +9,7 @@ import re
 import tempfile
 import warnings
 from pathlib import Path
-from typing import TYPE_CHECKING, Callable
+from typing import TYPE_CHECKING, Callable, Literal
 
 import httpx
 import huggingface_hub
@@ -33,11 +33,15 @@ if TYPE_CHECKING:
     from gradio.interface import Interface
 
 
+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.
+server_timeout = 600
+
+
 @document()
 def load(
     name: str,
     src: str | None = None,
-    hf_token: str | None = None,
+    hf_token: str | Literal[False] | None = None,
     alias: str | None = None,
     **kwargs,
 ) -> Blocks:
@@ -48,7 +52,7 @@ def load(
     Parameters:
         name: the name of the model (e.g. "gpt2" or "facebook/bart-base") or space (e.g. "flax-community/spanish-gpt2"), can include the `src` as prefix (e.g. "models/facebook/bart-base")
         src: the source of the model: `models` or `spaces` (or leave empty if source is provided as a prefix in `name`)
-        hf_token: optional access token for loading private Hugging Face Hub models or spaces. Find your token here: https://huggingface.co/settings/tokens.  Warning: only provide this if you are loading a trusted private Space as it can be read by the Space you are loading.
+        hf_token: optional access token for loading private Hugging Face Hub models or spaces. Will default to the locally saved token if not provided. Pass `token=False` if you don't want to send your token to the server. Find your token here: https://huggingface.co/settings/tokens.  Warning: only provide a token if you are loading a trusted private Space as it can be read by the Space you are loading.
         alias: optional string used as the name of the loaded model instead of the default name (only applies if loading a Space running Gradio 2.x)
     Returns:
         a Gradio Blocks object for the given model
@@ -65,7 +69,7 @@ def load(
 def load_blocks_from_repo(
     name: str,
     src: str | None = None,
-    hf_token: str | None = None,
+    hf_token: str | Literal[False] | None = None,
     alias: str | None = None,
     **kwargs,
 ) -> Blocks:
@@ -89,7 +93,7 @@ def load_blocks_from_repo(
     if src.lower() not in factory_methods:
         raise ValueError(f"parameter: src must be one of {factory_methods.keys()}")
 
-    if hf_token is not None:
+    if hf_token is not None and hf_token is not False:
         if Context.hf_token is not None and Context.hf_token != hf_token:
             warnings.warn(
                 """You are loading a model/Space with a different access token than the one you used to load a previous model/Space. This is not recommended, as it may cause unexpected behavior."""
@@ -100,12 +104,16 @@ def load_blocks_from_repo(
     return blocks
 
 
-def from_model(model_name: str, hf_token: str | None, alias: str | None, **kwargs):
+def from_model(
+    model_name: str, hf_token: str | Literal[False] | None, alias: str | None, **kwargs
+):
     model_url = f"https://huggingface.co/{model_name}"
     api_url = f"https://api-inference.huggingface.co/models/{model_name}"
     print(f"Fetching model from: {model_url}")
 
-    headers = {"Authorization": f"Bearer {hf_token}"} if hf_token is not None else {}
+    headers = (
+        {} if hf_token in [False, None] else {"Authorization": f"Bearer {hf_token}"}
+    )
     response = httpx.request("GET", api_url, headers=headers)
     if response.status_code != 200:
         raise ModelNotFoundError(
@@ -115,7 +123,7 @@ def from_model(model_name: str, hf_token: str | None, alias: str | None, **kwarg
 
     headers["X-Wait-For-Model"] = "true"
     client = huggingface_hub.InferenceClient(
-        model=model_name, headers=headers, token=hf_token,
+        model=model_name, headers=headers, token=hf_token, timeout=server_timeout,
     )
 
     # For tasks that are not yet supported by the InferenceClient
@@ -365,10 +373,14 @@ def from_model(model_name: str, hf_token: str | None, alias: str | None, **kwarg
     else:
         raise ValueError(f"Unsupported pipeline type: {p}")
 
-    def query_huggingface_inference_endpoints(*data):
+    def query_huggingface_inference_endpoints(*data, **kwargs):
         if preprocess is not None:
             data = preprocess(*data)
-        data = fn(*data)  # type: ignore
+        try:
+            data = fn(*data, **kwargs)  # type: ignore
+        except huggingface_hub.utils.HfHubHTTPError as e:
+            if "429" in str(e):
+                raise TooManyRequestsError() from e
         if postprocess is not None:
             data = postprocess(data)  # type: ignore
         return data
@@ -380,7 +392,7 @@ def from_model(model_name: str, hf_token: str | None, alias: str | None, **kwarg
         "inputs": inputs,
         "outputs": outputs,
         "title": model_name,
-    #    "examples": examples,
+        #"examples": examples,
     }
 
     kwargs = dict(interface_info, **kwargs)
@@ -391,19 +403,12 @@ def from_model(model_name: str, hf_token: str | None, alias: str | None, **kwarg
 def from_spaces(
     space_name: str, hf_token: str | None, alias: str | None, **kwargs
 ) -> Blocks:
-    client = Client(
-        space_name,
-        hf_token=hf_token,
-        download_files=False,
-        _skip_components=False,
-    )
-
     space_url = f"https://huggingface.co/spaces/{space_name}"
 
     print(f"Fetching Space from: {space_url}")
 
     headers = {}
-    if hf_token is not None:
+    if hf_token not in [False, None]:
         headers["Authorization"] = f"Bearer {hf_token}"
 
     iframe_url = (
@@ -440,8 +445,7 @@ def from_spaces(
                 "Blocks or Interface locally. You may find this Guide helpful: "
                 "https://gradio.app/using_blocks_like_functions/"
             )
-        if client.app_version < version.Version("4.0.0b14"):
-            return from_spaces_blocks(space=space_name, hf_token=hf_token)
+        return from_spaces_blocks(space=space_name, hf_token=hf_token)
 
 
 def from_spaces_blocks(space: str, hf_token: str | None) -> Blocks:
@@ -486,7 +490,7 @@ def from_spaces_interface(
     config = external_utils.streamline_spaces_interface(config)
     api_url = f"{iframe_url}/api/predict/"
     headers = {"Content-Type": "application/json"}
-    if hf_token is not None:
+    if hf_token not in [False, None]:
         headers["Authorization"] = f"Bearer {hf_token}"
 
     # The function should call the API with preprocessed data
@@ -526,6 +530,83 @@ def gr_Interface_load(
     src: str | None = None,
     hf_token: str | None = None,
     alias: str | None = None,
-    **kwargs,
+    **kwargs, # ignore
 ) -> Blocks:
-    return load_blocks_from_repo(name, src, hf_token, alias)
+    try:
+        return load_blocks_from_repo(name, src, hf_token, alias)
+    except Exception as e:
+        print(e)
+        return gradio.Interface(lambda: None, ['text'], ['image'])
+
+
+def list_uniq(l):
+    return sorted(set(l), key=l.index)
+
+
+def get_status(model_name: str):
+    from huggingface_hub import AsyncInferenceClient
+    client = AsyncInferenceClient(token=HF_TOKEN, timeout=10)
+    return client.get_model_status(model_name)
+
+
+def is_loadable(model_name: str, force_gpu: bool = False):
+    try:
+        status = get_status(model_name)
+    except Exception as e:
+        print(e)
+        print(f"Couldn't load {model_name}.")
+        return False
+    gpu_state = isinstance(status.compute_type, dict) and "gpu" in status.compute_type.keys()
+    if status is None or status.state not in ["Loadable", "Loaded"] or (force_gpu and not gpu_state):
+        print(f"Couldn't load {model_name}. Model state:'{status.state}', GPU:{gpu_state}")
+    return status is not None and status.state in ["Loadable", "Loaded"] and (not force_gpu or gpu_state)
+
+
+def find_model_list(author: str="", tags: list[str]=[], not_tag="", sort: str="last_modified", limit: int=30, force_gpu=False, check_status=False):
+    from huggingface_hub import HfApi
+    api = HfApi(token=HF_TOKEN)
+    default_tags = ["diffusers"]
+    if not sort: sort = "last_modified"
+    limit = limit * 20 if check_status and force_gpu else limit * 5
+    models = []
+    try:
+        model_infos = api.list_models(author=author, #task="text-to-image",
+                                       tags=list_uniq(default_tags + tags), cardData=True, sort=sort, limit=limit)
+    except Exception as e:
+        print(f"Error: Failed to list models.")
+        print(e)
+        return models
+    for model in model_infos:
+        if not model.private and not model.gated or HF_TOKEN is not None:
+           loadable = is_loadable(model.id, force_gpu) if check_status else True
+           if not_tag and not_tag in model.tags or not loadable: continue
+           models.append(model.id)
+           if len(models) == limit: break
+    return models
+
+
+def save_image(image, savefile, modelname, prompt, nprompt, height=0, width=0, steps=0, cfg=0, seed=-1):
+    from PIL import Image, PngImagePlugin
+    import json
+    try:
+        metadata = {"prompt": prompt, "negative_prompt": nprompt, "Model": {"Model": modelname.split("/")[-1]}}
+        if steps > 0: metadata["num_inference_steps"] = steps
+        if cfg > 0: metadata["guidance_scale"] = cfg
+        if seed != -1: metadata["seed"] = seed
+        if width > 0 and height > 0: metadata["resolution"] = f"{width} x {height}"
+        metadata_str = json.dumps(metadata)
+        info = PngImagePlugin.PngInfo()
+        info.add_text("metadata", metadata_str)
+        image.save(savefile, "PNG", pnginfo=info)
+        return str(Path(savefile).resolve())
+    except Exception as e:
+        print(f"Failed to save image file: {e}")
+        raise Exception(f"Failed to save image file:") from e
+
+
+def randomize_seed():
+    from random import seed, randint
+    MAX_SEED = 2**32-1
+    seed()
+    rseed = randint(0, MAX_SEED)
+    return rseed