Upload 9 files

app.py

@@ -3,10 +3,10 @@ import gradio as gr
 import json
 import torch
 from diffusers import DiffusionPipeline, AutoencoderTiny, AutoencoderKL, AutoPipelineForImage2Image
-from live_preview_helpers import flux_pipe_call_that_returns_an_iterable_of_images
+from live_preview_helpers import calculate_shift, retrieve_timesteps, flux_pipe_call_that_returns_an_iterable_of_images
 from diffusers.utils import load_image
 from diffusers import FluxControlNetPipeline, FluxControlNetModel, FluxMultiControlNetModel, FluxControlNetImg2ImgPipeline
-from huggingface_hub import HfFileSystem, ModelCard
+from huggingface_hub import hf_hub_download, HfFileSystem, ModelCard, snapshot_download
 import os
 import copy
 import random
@@ -589,9 +589,16 @@ def check_custom_model(link):
         # Assume it's a Hugging Face model path
         return get_huggingface_safetensors(link)
 
+def update_history(new_image, history):
+    """Updates the history gallery with the new image."""
+    if history is None:
+        history = []
+    history.insert(0, new_image)
+    return history
+
 css = '''
-#gen_btn{height: 100%}
 #gen_column{align-self: stretch}
+#gen_btn{height: 100%}
 #title{text-align: center}
 #title h1{font-size: 3em; display:inline-flex; align-items:center}
 #title img{width: 100px; margin-right: 0.25em}
@@ -605,11 +612,12 @@ css = '''
 #progress .generating{display:none}
 .progress-container {width: 100%;height: 30px;background-color: #f0f0f0;border-radius: 15px;overflow: hidden;margin-bottom: 20px}
 .progress-bar {height: 100%;background-color: #4f46e5;width: calc(var(--current) / var(--total) * 100%);transition: width 0.5s ease-in-out}
-.button_total{height: 100%}
+#component-8, .button_total{height: 100%; align-self: stretch;}
 #loaded_loras [data-testid="block-info"]{font-size:80%}
 #custom_lora_structure{background: var(--block-background-fill)}
 #custom_lora_btn{margin-top: auto;margin-bottom: 11px}
 #random_btn{font-size: 300%}
+#component-11{align-self: stretch;}
 .info {text-align:center; !important}
 '''
 with gr.Blocks(theme='NoCrypt/miku@>=1.2.2', fill_width=True, css=css, delete_cache=(60, 3600)) as app:
@@ -682,6 +690,8 @@ with gr.Blocks(theme='NoCrypt/miku@>=1.2.2', fill_width=True, css=css, delete_ca
             with gr.Column():
                 progress_bar = gr.Markdown(elem_id="progress",visible=False)
                 result = gr.Image(label="Generated Image", format="png", show_share_button=False)
+                with gr.Accordion("History", open=False):
+                    history_gallery = gr.Gallery(label="History", columns=6, object_fit="contain", interactive=False)
         with gr.Group():
             model_name = gr.Dropdown(label="Base Model", info="You can enter a huggingface model repo_id to want to use.", choices=models, value=models[0], allow_custom_value=True)
             model_info = gr.Markdown(elem_classes="info")
@@ -810,6 +820,10 @@ with gr.Blocks(theme='NoCrypt/miku@>=1.2.2', fill_width=True, css=css, delete_ca
         outputs=[result, seed, progress_bar],
         queue=True,
         show_api=True,
+    ).then(  # Update the history gallery
+        fn=lambda x, history: update_history(x, history),
+        inputs=[result, history_gallery],
+        outputs=history_gallery,
     )
 
     input_image.upload(preprocess_i2i_image, [input_image, input_image_preprocess, height, width], [input_image], queue=False, show_api=False)

live_preview_helpers.py

@@ -1,166 +1,166 @@
-import torch
-import numpy as np
-from diffusers import FluxPipeline, AutoencoderTiny, FlowMatchEulerDiscreteScheduler
-from typing import Any, Dict, List, Optional, Union
-
-# Helper functions
-def calculate_shift(
-    image_seq_len,
-    base_seq_len: int = 256,
-    max_seq_len: int = 4096,
-    base_shift: float = 0.5,
-    max_shift: float = 1.16,
-):
-    m = (max_shift - base_shift) / (max_seq_len - base_seq_len)
-    b = base_shift - m * base_seq_len
-    mu = image_seq_len * m + b
-    return mu
-
-def retrieve_timesteps(
-    scheduler,
-    num_inference_steps: Optional[int] = None,
-    device: Optional[Union[str, torch.device]] = None,
-    timesteps: Optional[List[int]] = None,
-    sigmas: Optional[List[float]] = None,
-    **kwargs,
-):
-    if timesteps is not None and sigmas is not None:
-        raise ValueError("Only one of `timesteps` or `sigmas` can be passed. Please choose one to set custom values")
-    if timesteps is not None:
-        scheduler.set_timesteps(timesteps=timesteps, device=device, **kwargs)
-        timesteps = scheduler.timesteps
-        num_inference_steps = len(timesteps)
-    elif sigmas is not None:
-        scheduler.set_timesteps(sigmas=sigmas, device=device, **kwargs)
-        timesteps = scheduler.timesteps
-        num_inference_steps = len(timesteps)
-    else:
-        scheduler.set_timesteps(num_inference_steps, device=device, **kwargs)
-        timesteps = scheduler.timesteps
-    return timesteps, num_inference_steps
-
-# FLUX pipeline function
-@torch.inference_mode()
-def flux_pipe_call_that_returns_an_iterable_of_images(
-    self,
-    prompt: Union[str, List[str]] = None,
-    prompt_2: Optional[Union[str, List[str]]] = None,
-    height: Optional[int] = None,
-    width: Optional[int] = None,
-    num_inference_steps: int = 28,
-    timesteps: List[int] = None,
-    guidance_scale: float = 3.5,
-    num_images_per_prompt: Optional[int] = 1,
-    generator: Optional[Union[torch.Generator, List[torch.Generator]]] = None,
-    latents: Optional[torch.FloatTensor] = None,
-    prompt_embeds: Optional[torch.FloatTensor] = None,
-    pooled_prompt_embeds: Optional[torch.FloatTensor] = None,
-    output_type: Optional[str] = "pil",
-    return_dict: bool = True,
-    joint_attention_kwargs: Optional[Dict[str, Any]] = None,
-    max_sequence_length: int = 512,
-    good_vae: Optional[Any] = None,
-):
-    height = height or self.default_sample_size * self.vae_scale_factor
-    width = width or self.default_sample_size * self.vae_scale_factor
-
-    # 1. Check inputs
-    self.check_inputs(
-        prompt,
-        prompt_2,
-        height,
-        width,
-        prompt_embeds=prompt_embeds,
-        pooled_prompt_embeds=pooled_prompt_embeds,
-        max_sequence_length=max_sequence_length,
-    )
-
-    self._guidance_scale = guidance_scale
-    self._joint_attention_kwargs = joint_attention_kwargs
-    self._interrupt = False
-
-    # 2. Define call parameters
-    batch_size = 1 if isinstance(prompt, str) else len(prompt)
-    device = self._execution_device
-
-    # 3. Encode prompt
-    lora_scale = joint_attention_kwargs.get("scale", None) if joint_attention_kwargs is not None else None
-    prompt_embeds, pooled_prompt_embeds, text_ids = self.encode_prompt(
-        prompt=prompt,
-        prompt_2=prompt_2,
-        prompt_embeds=prompt_embeds,
-        pooled_prompt_embeds=pooled_prompt_embeds,
-        device=device,
-        num_images_per_prompt=num_images_per_prompt,
-        max_sequence_length=max_sequence_length,
-        lora_scale=lora_scale,
-    )
-    # 4. Prepare latent variables
-    num_channels_latents = self.transformer.config.in_channels // 4
-    latents, latent_image_ids = self.prepare_latents(
-        batch_size * num_images_per_prompt,
-        num_channels_latents,
-        height,
-        width,
-        prompt_embeds.dtype,
-        device,
-        generator,
-        latents,
-    )
-    # 5. Prepare timesteps
-    sigmas = np.linspace(1.0, 1 / num_inference_steps, num_inference_steps)
-    image_seq_len = latents.shape[1]
-    mu = calculate_shift(
-        image_seq_len,
-        self.scheduler.config.base_image_seq_len,
-        self.scheduler.config.max_image_seq_len,
-        self.scheduler.config.base_shift,
-        self.scheduler.config.max_shift,
-    )
-    timesteps, num_inference_steps = retrieve_timesteps(
-        self.scheduler,
-        num_inference_steps,
-        device,
-        timesteps,
-        sigmas,
-        mu=mu,
-    )
-    self._num_timesteps = len(timesteps)
-
-    # Handle guidance
-    guidance = torch.full([1], guidance_scale, device=device, dtype=torch.float32).expand(latents.shape[0]) if self.transformer.config.guidance_embeds else None
-
-    # 6. Denoising loop
-    for i, t in enumerate(timesteps):
-        if self.interrupt:
-            continue
-
-        timestep = t.expand(latents.shape[0]).to(latents.dtype)
-
-        noise_pred = self.transformer(
-            hidden_states=latents,
-            timestep=timestep / 1000,
-            guidance=guidance,
-            pooled_projections=pooled_prompt_embeds,
-            encoder_hidden_states=prompt_embeds,
-            txt_ids=text_ids,
-            img_ids=latent_image_ids,
-            joint_attention_kwargs=self.joint_attention_kwargs,
-            return_dict=False,
-        )[0]
-        # Yield intermediate result
-        latents_for_image = self._unpack_latents(latents, height, width, self.vae_scale_factor)
-        latents_for_image = (latents_for_image / self.vae.config.scaling_factor) + self.vae.config.shift_factor
-        image = self.vae.decode(latents_for_image, return_dict=False)[0]
-        yield self.image_processor.postprocess(image, output_type=output_type)[0]
-        latents = self.scheduler.step(noise_pred, t, latents, return_dict=False)[0]
-        torch.cuda.empty_cache()
-        
-
-    # Final image using good_vae
-    latents = self._unpack_latents(latents, height, width, self.vae_scale_factor)
-    latents = (latents / good_vae.config.scaling_factor) + good_vae.config.shift_factor
-    image = good_vae.decode(latents, return_dict=False)[0]
-    self.maybe_free_model_hooks()
-    torch.cuda.empty_cache()
-    yield self.image_processor.postprocess(image, output_type=output_type)[0]
+import torch
+import numpy as np
+from diffusers import FluxPipeline, AutoencoderTiny, FlowMatchEulerDiscreteScheduler
+from typing import Any, Dict, List, Optional, Union
+
+# Helper functions
+def calculate_shift(
+    image_seq_len,
+    base_seq_len: int = 256,
+    max_seq_len: int = 4096,
+    base_shift: float = 0.5,
+    max_shift: float = 1.16,
+):
+    m = (max_shift - base_shift) / (max_seq_len - base_seq_len)
+    b = base_shift - m * base_seq_len
+    mu = image_seq_len * m + b
+    return mu
+
+def retrieve_timesteps(
+    scheduler,
+    num_inference_steps: Optional[int] = None,
+    device: Optional[Union[str, torch.device]] = None,
+    timesteps: Optional[List[int]] = None,
+    sigmas: Optional[List[float]] = None,
+    **kwargs,
+):
+    if timesteps is not None and sigmas is not None:
+        raise ValueError("Only one of `timesteps` or `sigmas` can be passed. Please choose one to set custom values")
+    if timesteps is not None:
+        scheduler.set_timesteps(timesteps=timesteps, device=device, **kwargs)
+        timesteps = scheduler.timesteps
+        num_inference_steps = len(timesteps)
+    elif sigmas is not None:
+        scheduler.set_timesteps(sigmas=sigmas, device=device, **kwargs)
+        timesteps = scheduler.timesteps
+        num_inference_steps = len(timesteps)
+    else:
+        scheduler.set_timesteps(num_inference_steps, device=device, **kwargs)
+        timesteps = scheduler.timesteps
+    return timesteps, num_inference_steps
+
+# FLUX pipeline function
+@torch.inference_mode()
+def flux_pipe_call_that_returns_an_iterable_of_images(
+    self,
+    prompt: Union[str, List[str]] = None,
+    prompt_2: Optional[Union[str, List[str]]] = None,
+    height: Optional[int] = None,
+    width: Optional[int] = None,
+    num_inference_steps: int = 28,
+    timesteps: List[int] = None,
+    guidance_scale: float = 3.5,
+    num_images_per_prompt: Optional[int] = 1,
+    generator: Optional[Union[torch.Generator, List[torch.Generator]]] = None,
+    latents: Optional[torch.FloatTensor] = None,
+    prompt_embeds: Optional[torch.FloatTensor] = None,
+    pooled_prompt_embeds: Optional[torch.FloatTensor] = None,
+    output_type: Optional[str] = "pil",
+    return_dict: bool = True,
+    joint_attention_kwargs: Optional[Dict[str, Any]] = None,
+    max_sequence_length: int = 512,
+    good_vae: Optional[Any] = None,
+):
+    height = height or self.default_sample_size * self.vae_scale_factor
+    width = width or self.default_sample_size * self.vae_scale_factor
+
+    # 1. Check inputs
+    self.check_inputs(
+        prompt,
+        prompt_2,
+        height,
+        width,
+        prompt_embeds=prompt_embeds,
+        pooled_prompt_embeds=pooled_prompt_embeds,
+        max_sequence_length=max_sequence_length,
+    )
+
+    self._guidance_scale = guidance_scale
+    self._joint_attention_kwargs = joint_attention_kwargs
+    self._interrupt = False
+
+    # 2. Define call parameters
+    batch_size = 1 if isinstance(prompt, str) else len(prompt)
+    device = self._execution_device
+
+    # 3. Encode prompt
+    lora_scale = joint_attention_kwargs.get("scale", None) if joint_attention_kwargs is not None else None
+    prompt_embeds, pooled_prompt_embeds, text_ids = self.encode_prompt(
+        prompt=prompt,
+        prompt_2=prompt_2,
+        prompt_embeds=prompt_embeds,
+        pooled_prompt_embeds=pooled_prompt_embeds,
+        device=device,
+        num_images_per_prompt=num_images_per_prompt,
+        max_sequence_length=max_sequence_length,
+        lora_scale=lora_scale,
+    )
+    # 4. Prepare latent variables
+    num_channels_latents = self.transformer.config.in_channels // 4
+    latents, latent_image_ids = self.prepare_latents(
+        batch_size * num_images_per_prompt,
+        num_channels_latents,
+        height,
+        width,
+        prompt_embeds.dtype,
+        device,
+        generator,
+        latents,
+    )
+    # 5. Prepare timesteps
+    sigmas = np.linspace(1.0, 1 / num_inference_steps, num_inference_steps)
+    image_seq_len = latents.shape[1]
+    mu = calculate_shift(
+        image_seq_len,
+        self.scheduler.config.base_image_seq_len,
+        self.scheduler.config.max_image_seq_len,
+        self.scheduler.config.base_shift,
+        self.scheduler.config.max_shift,
+    )
+    timesteps, num_inference_steps = retrieve_timesteps(
+        self.scheduler,
+        num_inference_steps,
+        device,
+        timesteps,
+        sigmas,
+        mu=mu,
+    )
+    self._num_timesteps = len(timesteps)
+
+    # Handle guidance
+    guidance = torch.full([1], guidance_scale, device=device, dtype=torch.float32).expand(latents.shape[0]) if self.transformer.config.guidance_embeds else None
+
+    # 6. Denoising loop
+    for i, t in enumerate(timesteps):
+        if self.interrupt:
+            continue
+
+        timestep = t.expand(latents.shape[0]).to(latents.dtype)
+
+        noise_pred = self.transformer(
+            hidden_states=latents,
+            timestep=timestep / 1000,
+            guidance=guidance,
+            pooled_projections=pooled_prompt_embeds,
+            encoder_hidden_states=prompt_embeds,
+            txt_ids=text_ids,
+            img_ids=latent_image_ids,
+            joint_attention_kwargs=self.joint_attention_kwargs,
+            return_dict=False,
+        )[0]
+        # Yield intermediate result
+        latents_for_image = self._unpack_latents(latents, height, width, self.vae_scale_factor)
+        latents_for_image = (latents_for_image / self.vae.config.scaling_factor) + self.vae.config.shift_factor
+        image = self.vae.decode(latents_for_image, return_dict=False)[0]
+        yield self.image_processor.postprocess(image, output_type=output_type)[0]
+        latents = self.scheduler.step(noise_pred, t, latents, return_dict=False)[0]
+        torch.cuda.empty_cache()
+        
+
+    # Final image using good_vae
+    latents = self._unpack_latents(latents, height, width, self.vae_scale_factor)
+    latents = (latents / good_vae.config.scaling_factor) + good_vae.config.shift_factor
+    image = good_vae.decode(latents, return_dict=False)[0]
+    self.maybe_free_model_hooks()
+    torch.cuda.empty_cache()
+    yield self.image_processor.postprocess(image, output_type=output_type)[0]

loras.json

@@ -111,12 +111,6 @@
     "trigger_word": "JojosoStyle",
     "trigger_position": "prepend"
   },
-  {
-    "image": "https://github.com/XLabs-AI/x-flux/blob/main/assets/readme/examples/picture-6-rev1.png?raw=true",
-    "title": "flux-Realism",
-    "repo": "XLabs-AI/flux-RealismLora",
-    "trigger_word": ""
-  },
   {
     "image": "https://huggingface.co/multimodalart/vintage-ads-flux/resolve/main/samples/j_XNU6Oe0mgttyvf9uPb3_dc244dd3d6c246b4aff8351444868d66.png",
     "title": "Vintage Ads",
@@ -205,13 +199,6 @@
     "repo": "dataautogpt3/FLUX-SyntheticAnime",
     "trigger_word": "1980s anime screengrab, VHS quality"
   },
-  {
-    "image": "https://github.com/XLabs-AI/x-flux/blob/main/assets/readme/examples/result_14.png?raw=true",
-    "title": "flux-anime",
-    "repo": "XLabs-AI/flux-lora-collection",
-    "weights": "anime_lora.safetensors",
-    "trigger_word": ", anime"
-  },
   {
     "image": "https://replicate.delivery/yhqm/QD8Ioy5NExqSCtBS8hG04XIRQZFaC9pxJemINT1bibyjZfSTA/out-0.webp",
     "title": "80s Cyberpunk",
@@ -225,20 +212,6 @@
     "repo": "kudzueye/boreal-flux-dev-v2",
     "trigger_word": "phone photo"
   },
-  {
-    "image": "https://github.com/XLabs-AI/x-flux/blob/main/assets/readme/examples/result_18.png?raw=true",
-    "title": "flux-disney",
-    "repo": "XLabs-AI/flux-lora-collection",
-    "weights": "disney_lora.safetensors",
-    "trigger_word": ", disney style"
-  },
-  {
-    "image": "https://github.com/XLabs-AI/x-flux/blob/main/assets/readme/examples/result_23.png?raw=true",
-    "title": "flux-art",
-    "repo": "XLabs-AI/flux-lora-collection",
-    "weights": "art_lora.safetensors",
-    "trigger_word": ", art"
-  },
   {
     "image": "https://huggingface.co/martintomov/retrofuturism-flux/resolve/main/images/2e40deba-858e-454f-ae1c-d1ba2adb6a65.jpeg",
     "title": "Retrofuturism Flux",

modutils.py

@@ -136,7 +136,7 @@ def save_gallery_images(images, progress=gr.Progress(track_tqdm=True)):
     dt_now = datetime.now(timezone(timedelta(hours=9)))
     basename = dt_now.strftime('%Y%m%d_%H%M%S_')
     i = 1
-    if not images: return images
+    if not images: return images, gr.update(visible=False)
     output_images = []
     output_paths = []
     for image in images:
@@ -153,7 +153,7 @@ def save_gallery_images(images, progress=gr.Progress(track_tqdm=True)):
             output_paths.append(str(newpath))
             output_images.append((str(newpath), str(filename)))
     progress(1, desc="Gallery updated.")
-    return gr.update(value=output_images), gr.update(value=output_paths), gr.update(visible=True)
+    return gr.update(value=output_images), gr.update(value=output_paths, visible=True)
 
 
 def download_private_repo(repo_id, dir_path, is_replace):

requirements.txt

@@ -1,7 +1,7 @@
-spaces
+spaces>=0.30.3
 torch
 git+https://github.com/huggingface/diffusers.git@1131e3d04e3131f4c24565257665d75364d696d9
-transformers
+git+https://github.com/huggingface/transformers.git
 git+https://github.com/huggingface/peft.git
 sentencepiece
 torchvision