update code

.gitattributes

@@ -25,6 +25,7 @@
 *.safetensors filter=lfs diff=lfs merge=lfs -text
 saved_model/**/* filter=lfs diff=lfs merge=lfs -text
 *.tar.* filter=lfs diff=lfs merge=lfs -text
+*.tar filter=lfs diff=lfs merge=lfs -text
 *.tflite filter=lfs diff=lfs merge=lfs -text
 *.tgz filter=lfs diff=lfs merge=lfs -text
 *.wasm filter=lfs diff=lfs merge=lfs -text
@@ -32,4 +33,4 @@ saved_model/**/* filter=lfs diff=lfs merge=lfs -text
 *.zip filter=lfs diff=lfs merge=lfs -text
 *.zst filter=lfs diff=lfs merge=lfs -text
 *tfevents* filter=lfs diff=lfs merge=lfs -text
-scale-hf-logo.png filter=lfs diff=lfs merge=lfs -text
+test_imgs/*.png filter=lfs diff=lfs merge=lfs -text

.gitignore

@@ -1,13 +0,0 @@
-auto_evals/
-venv/
-__pycache__/
-.env
-.ipynb_checkpoints
-*ipynb
-.vscode/
-
-eval-queue/
-eval-results/
-eval-queue-bk/
-eval-results-bk/
-logs/

.pre-commit-config.yaml

@@ -1,53 +0,0 @@
-# Copyright (c) 2022, NVIDIA CORPORATION.  All rights reserved.
-#
-# Licensed under the Apache License, Version 2.0 (the "License");
-# you may not use this file except in compliance with the License.
-# You may obtain a copy of the License at
-#
-#     http://www.apache.org/licenses/LICENSE-2.0
-#
-# Unless required by applicable law or agreed to in writing, software
-# distributed under the License is distributed on an "AS IS" BASIS,
-# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
-# See the License for the specific language governing permissions and
-# limitations under the License.
-
-default_language_version:
-  python: python3
-
-ci:
-  autofix_prs: true
-  autoupdate_commit_msg: '[pre-commit.ci] pre-commit suggestions'
-  autoupdate_schedule: quarterly
-
-repos:
-  - repo: https://github.com/pre-commit/pre-commit-hooks
-    rev: v4.3.0
-    hooks:
-      - id: check-yaml
-      - id: check-case-conflict
-      - id: detect-private-key
-      - id: check-added-large-files
-        args: ['--maxkb=1000']
-      - id: requirements-txt-fixer
-      - id: end-of-file-fixer
-      - id: trailing-whitespace
-
-  - repo: https://github.com/PyCQA/isort
-    rev: 5.12.0
-    hooks:
-      - id: isort
-        name: Format imports
-
-  - repo: https://github.com/psf/black
-    rev: 22.12.0
-    hooks:
-      - id: black
-        name: Format code
-        additional_dependencies: ['click==8.0.2']
-
-  - repo: https://github.com/charliermarsh/ruff-pre-commit
-    # Ruff version.
-    rev: 'v0.0.267'
-    hooks:
-      - id: ruff

Makefile

@@ -1,13 +0,0 @@
-.PHONY: style format
-
-
-style:
-	python -m black --line-length 119 .
-	python -m isort .
-	ruff check --fix .
-
-
-quality:
-	python -m black --check --line-length 119 .
-	python -m isort --check-only .
-	ruff check .

README.md

@@ -1,46 +1,14 @@
 ---
-title: Happykala Ghibli
-emoji: 🥇
-colorFrom: green
-colorTo: indigo
+title: EasyControl Ghibli
+emoji: 🦀
+colorFrom: blue
+colorTo: blue
 sdk: gradio
+sdk_version: 5.23.2
 app_file: app.py
-pinned: true
+pinned: false
 license: apache-2.0
-short_description: Duplicate this leaderboard to initialize your own!
-sdk_version: 5.19.0
+short_description: New Ghibli EasyControl model is now released!!
 ---
 
-# Start the configuration
-
-Most of the variables to change for a default leaderboard are in `src/env.py` (replace the path for your leaderboard) and `src/about.py` (for tasks).
-
-Results files should have the following format and be stored as json files:
-```json
-{
-    "config": {
-        "model_dtype": "torch.float16", # or torch.bfloat16 or 8bit or 4bit
-        "model_name": "path of the model on the hub: org/model",
-        "model_sha": "revision on the hub",
-    },
-    "results": {
-        "task_name": {
-            "metric_name": score,
-        },
-        "task_name2": {
-            "metric_name": score,
-        }
-    }
-}
-```
-
-Request files are created automatically by this tool.
-
-If you encounter problem on the space, don't hesitate to restart it to remove the create eval-queue, eval-queue-bk, eval-results and eval-results-bk created folder.
-
-# Code logic for more complex edits
-
-You'll find 
-- the main table' columns names and properties in `src/display/utils.py`
-- the logic to read all results and request files, then convert them in dataframe lines, in `src/leaderboard/read_evals.py`, and `src/populate.py`
-- the logic to allow or filter submissions in `src/submission/submit.py` and `src/submission/check_validity.py`
+Check out the configuration reference at https://huggingface.co/docs/hub/spaces-config-reference

app.py

@@ -1,204 +1,110 @@
+import spaces
+import os
+import json
+import time
+import torch
+from PIL import Image
+from tqdm import tqdm
 import gradio as gr
-from gradio_leaderboard import Leaderboard, ColumnFilter, SelectColumns
-import pandas as pd
-from apscheduler.schedulers.background import BackgroundScheduler
-from huggingface_hub import snapshot_download
 
-from src.about import (
-    CITATION_BUTTON_LABEL,
-    CITATION_BUTTON_TEXT,
-    EVALUATION_QUEUE_TEXT,
-    INTRODUCTION_TEXT,
-    LLM_BENCHMARKS_TEXT,
-    TITLE,
-)
-from src.display.css_html_js import custom_css
-from src.display.utils import (
-    BENCHMARK_COLS,
-    COLS,
-    EVAL_COLS,
-    EVAL_TYPES,
-    AutoEvalColumn,
-    ModelType,
-    fields,
-    WeightType,
-    Precision
-)
-from src.envs import API, EVAL_REQUESTS_PATH, EVAL_RESULTS_PATH, QUEUE_REPO, REPO_ID, RESULTS_REPO, TOKEN
-from src.populate import get_evaluation_queue_df, get_leaderboard_df
-from src.submission.submit import add_new_eval
-
-
-def restart_space():
-    API.restart_space(repo_id=REPO_ID)
-
-### Space initialisation
-try:
-    print(EVAL_REQUESTS_PATH)
-    snapshot_download(
-        repo_id=QUEUE_REPO, local_dir=EVAL_REQUESTS_PATH, repo_type="dataset", tqdm_class=None, etag_timeout=30, token=TOKEN
-    )
-except Exception:
-    restart_space()
-try:
-    print(EVAL_RESULTS_PATH)
-    snapshot_download(
-        repo_id=RESULTS_REPO, local_dir=EVAL_RESULTS_PATH, repo_type="dataset", tqdm_class=None, etag_timeout=30, token=TOKEN
-    )
-except Exception:
-    restart_space()
-
-
-LEADERBOARD_DF = get_leaderboard_df(EVAL_RESULTS_PATH, EVAL_REQUESTS_PATH, COLS, BENCHMARK_COLS)
-
-(
-    finished_eval_queue_df,
-    running_eval_queue_df,
-    pending_eval_queue_df,
-) = get_evaluation_queue_df(EVAL_REQUESTS_PATH, EVAL_COLS)
-
-def init_leaderboard(dataframe):
-    if dataframe is None or dataframe.empty:
-        raise ValueError("Leaderboard DataFrame is empty or None.")
-    return Leaderboard(
-        value=dataframe,
-        datatype=[c.type for c in fields(AutoEvalColumn)],
-        select_columns=SelectColumns(
-            default_selection=[c.name for c in fields(AutoEvalColumn) if c.displayed_by_default],
-            cant_deselect=[c.name for c in fields(AutoEvalColumn) if c.never_hidden],
-            label="Select Columns to Display:",
-        ),
-        search_columns=[AutoEvalColumn.model.name, AutoEvalColumn.license.name],
-        hide_columns=[c.name for c in fields(AutoEvalColumn) if c.hidden],
-        filter_columns=[
-            ColumnFilter(AutoEvalColumn.model_type.name, type="checkboxgroup", label="Model types"),
-            ColumnFilter(AutoEvalColumn.precision.name, type="checkboxgroup", label="Precision"),
-            ColumnFilter(
-                AutoEvalColumn.params.name,
-                type="slider",
-                min=0.01,
-                max=150,
-                label="Select the number of parameters (B)",
-            ),
-            ColumnFilter(
-                AutoEvalColumn.still_on_hub.name, type="boolean", label="Deleted/incomplete", default=True
-            ),
-        ],
-        bool_checkboxgroup_label="Hide models",
-        interactive=False,
-    )
-
-
-demo = gr.Blocks(css=custom_css)
-with demo:
-    gr.HTML(TITLE)
-    gr.Markdown(INTRODUCTION_TEXT, elem_classes="markdown-text")
-
-    with gr.Tabs(elem_classes="tab-buttons") as tabs:
-        with gr.TabItem("🏅 LLM Benchmark", elem_id="llm-benchmark-tab-table", id=0):
-            leaderboard = init_leaderboard(LEADERBOARD_DF)
-
-        with gr.TabItem("📝 About", elem_id="llm-benchmark-tab-table", id=2):
-            gr.Markdown(LLM_BENCHMARKS_TEXT, elem_classes="markdown-text")
-
-        with gr.TabItem("🚀 Submit here! ", elem_id="llm-benchmark-tab-table", id=3):
+from safetensors.torch import save_file
+from src.pipeline import FluxPipeline
+from src.transformer_flux import FluxTransformer2DModel
+from src.lora_helper import set_single_lora, set_multi_lora, unset_lora
+
+# Initialize the image processor
+base_path = "black-forest-labs/FLUX.1-dev"    
+lora_base_path = "./models"
+
+
+pipe = FluxPipeline.from_pretrained(base_path, torch_dtype=torch.bfloat16)
+transformer = FluxTransformer2DModel.from_pretrained(base_path, subfolder="transformer", torch_dtype=torch.bfloat16)
+pipe.transformer = transformer
+pipe.to("cuda")
+
+def clear_cache(transformer):
+    for name, attn_processor in transformer.attn_processors.items():
+        attn_processor.bank_kv.clear()
+
+# Define the Gradio interface
+@spaces.GPU()
+def single_condition_generate_image(prompt, spatial_img, height, width, seed, control_type):
+    # Set the control type
+    if control_type == "Ghibli":
+        lora_path = os.path.join(lora_base_path, "Ghibli.safetensors")
+    set_single_lora(pipe.transformer, lora_path, lora_weights=[1], cond_size=512)
+    
+    # Process the image
+    spatial_imgs = [spatial_img] if spatial_img else []
+    image = pipe(
+        prompt,
+        height=int(height),
+        width=int(width),
+        guidance_scale=3.5,
+        num_inference_steps=25,
+        max_sequence_length=512,
+        generator=torch.Generator("cpu").manual_seed(seed), 
+        subject_images=[],
+        spatial_images=spatial_imgs,
+        cond_size=512,
+    ).images[0]
+    clear_cache(pipe.transformer)
+    return image
+
+# Define the Gradio interface components
+control_types = ["Ghibli"]
+
+# Example data
+single_examples = [
+    ["Ghibli Studio style, Charming hand-drawn anime-style illustration", Image.open("./test_imgs/00.png"), 680, 1024, 5, "Ghibli"],
+    ["Ghibli Studio style, Charming hand-drawn anime-style illustration", Image.open("./test_imgs/02.png"), 560, 1024, 42, "Ghibli"],
+    ["Ghibli Studio style, Charming hand-drawn anime-style illustration", Image.open("./test_imgs/03.png"), 568, 1024, 1, "Ghibli"],
+    ["Ghibli Studio style, Charming hand-drawn anime-style illustration", Image.open("./test_imgs/04.png"), 768, 672, 1, "Ghibli"],
+    ["Ghibli Studio style, Charming hand-drawn anime-style illustration", Image.open("./test_imgs/06.png"), 896, 1024, 1, "Ghibli"],
+    ["Ghibli Studio style, Charming hand-drawn anime-style illustration", Image.open("./test_imgs/07.png"), 528, 800, 1, "Ghibli"],
+    ["Ghibli Studio style, Charming hand-drawn anime-style illustration", Image.open("./test_imgs/08.png"), 696, 1024, 1, "Ghibli"],
+    ["Ghibli Studio style, Charming hand-drawn anime-style illustration", Image.open("./test_imgs/09.png"), 896, 1024, 1, "Ghibli"],
+]
+
+
+# Create the Gradio Blocks interface
+with gr.Blocks() as demo:
+    gr.Markdown("# Ghibli Studio Control Image Generation with EasyControl")
+    gr.Markdown("The model is trained on **only 100 real Asian faces** paired with **GPT-4o-generated Ghibli-style counterparts**, and it preserves facial features while applying the iconic anime aesthetic.")
+    gr.Markdown("Generate images using EasyControl with Ghibli control LoRAs.（Due to hardware constraints, only low-resolution images can be generated. For high-resolution (1024+), please set up your own environment.）")
+    
+    gr.Markdown("**[Attention!!]**：The recommended prompts for using Ghibli Control LoRA should include the trigger words: `Ghibli Studio style, Charming hand-drawn anime-style illustration`")
+    # gr.Markdown("😊😊If you like this demo, please give us a star (github: [EasyControl](https://github.com/Xiaojiu-z/EasyControl))")
+
+    with gr.Tab("Ghibli Condition Generation"):
+        with gr.Row():
             with gr.Column():
-                with gr.Row():
-                    gr.Markdown(EVALUATION_QUEUE_TEXT, elem_classes="markdown-text")
-
-                with gr.Column():
-                    with gr.Accordion(
-                        f"✅ Finished Evaluations ({len(finished_eval_queue_df)})",
-                        open=False,
-                    ):
-                        with gr.Row():
-                            finished_eval_table = gr.components.Dataframe(
-                                value=finished_eval_queue_df,
-                                headers=EVAL_COLS,
-                                datatype=EVAL_TYPES,
-                                row_count=5,
-                            )
-                    with gr.Accordion(
-                        f"🔄 Running Evaluation Queue ({len(running_eval_queue_df)})",
-                        open=False,
-                    ):
-                        with gr.Row():
-                            running_eval_table = gr.components.Dataframe(
-                                value=running_eval_queue_df,
-                                headers=EVAL_COLS,
-                                datatype=EVAL_TYPES,
-                                row_count=5,
-                            )
-
-                    with gr.Accordion(
-                        f"⏳ Pending Evaluation Queue ({len(pending_eval_queue_df)})",
-                        open=False,
-                    ):
-                        with gr.Row():
-                            pending_eval_table = gr.components.Dataframe(
-                                value=pending_eval_queue_df,
-                                headers=EVAL_COLS,
-                                datatype=EVAL_TYPES,
-                                row_count=5,
-                            )
-            with gr.Row():
-                gr.Markdown("# ✉️✨ Submit your model here!", elem_classes="markdown-text")
-
-            with gr.Row():
-                with gr.Column():
-                    model_name_textbox = gr.Textbox(label="Model name")
-                    revision_name_textbox = gr.Textbox(label="Revision commit", placeholder="main")
-                    model_type = gr.Dropdown(
-                        choices=[t.to_str(" : ") for t in ModelType if t != ModelType.Unknown],
-                        label="Model type",
-                        multiselect=False,
-                        value=None,
-                        interactive=True,
-                    )
-
-                with gr.Column():
-                    precision = gr.Dropdown(
-                        choices=[i.value.name for i in Precision if i != Precision.Unknown],
-                        label="Precision",
-                        multiselect=False,
-                        value="float16",
-                        interactive=True,
-                    )
-                    weight_type = gr.Dropdown(
-                        choices=[i.value.name for i in WeightType],
-                        label="Weights type",
-                        multiselect=False,
-                        value="Original",
-                        interactive=True,
-                    )
-                    base_model_name_textbox = gr.Textbox(label="Base model (for delta or adapter weights)")
-
-            submit_button = gr.Button("Submit Eval")
-            submission_result = gr.Markdown()
-            submit_button.click(
-                add_new_eval,
-                [
-                    model_name_textbox,
-                    base_model_name_textbox,
-                    revision_name_textbox,
-                    precision,
-                    weight_type,
-                    model_type,
-                ],
-                submission_result,
-            )
-
-    with gr.Row():
-        with gr.Accordion("📙 Citation", open=False):
-            citation_button = gr.Textbox(
-                value=CITATION_BUTTON_TEXT,
-                label=CITATION_BUTTON_LABEL,
-                lines=20,
-                elem_id="citation-button",
-                show_copy_button=True,
-            )
+                prompt = gr.Textbox(label="Prompt", value="Ghibli Studio style, Charming hand-drawn anime-style illustration")
+                spatial_img = gr.Image(label="Ghibli Image", type="pil")  # 上传图像文件
+                height = gr.Slider(minimum=256, maximum=1024, step=64, label="Height", value=768)
+                width = gr.Slider(minimum=256, maximum=1024, step=64, label="Width", value=768)
+                seed = gr.Number(label="Seed", value=42)
+                control_type = gr.Dropdown(choices=control_types, label="Control Type")
+                single_generate_btn = gr.Button("Generate Image")
+            with gr.Column():
+                single_output_image = gr.Image(label="Generated Image")
+
+        # Add examples for Single Condition Generation
+        gr.Examples(
+            examples=single_examples,
+            inputs=[prompt, spatial_img, height, width, seed, control_type],
+            outputs=single_output_image,
+            fn=single_condition_generate_image,
+            cache_examples=False,  # 缓存示例结果以加快加载速度
+            label="Single Condition Examples"
+        )
+
+    # Link the buttons to the functions
+    single_generate_btn.click(
+        single_condition_generate_image,
+        inputs=[prompt, spatial_img, height, width, seed, control_type],
+        outputs=single_output_image
+    )
 
-scheduler = BackgroundScheduler()
-scheduler.add_job(restart_space, "interval", seconds=1800)
-scheduler.start()
-demo.queue(default_concurrency_limit=40).launch()
+# Launch the Gradio app
+demo.queue().launch()

models/Ghibli.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:a5204551bd9cb587d659fe1cc50cf524b6339348bc5b1c3ea3b4efe71eb5e753
+size 298895992

pyproject.toml

@@ -1,13 +0,0 @@
-[tool.ruff]
-# Enable pycodestyle (`E`) and Pyflakes (`F`) codes by default.
-select = ["E", "F"]
-ignore = ["E501"] # line too long (black is taking care of this)
-line-length = 119
-fixable = ["A", "B", "C", "D", "E", "F", "G", "I", "N", "Q", "S", "T", "W", "ANN", "ARG", "BLE", "COM", "DJ", "DTZ", "EM", "ERA", "EXE", "FBT", "ICN", "INP", "ISC", "NPY", "PD", "PGH", "PIE", "PL", "PT", "PTH", "PYI", "RET", "RSE", "RUF", "SIM", "SLF", "TCH", "TID", "TRY", "UP", "YTT"]
-
-[tool.isort]
-profile = "black"
-line_length = 119
-
-[tool.black]
-line-length = 119

requirements.txt

@@ -1,16 +1,14 @@
-APScheduler
-black
-datasets
-gradio
-gradio[oauth]
-gradio_leaderboard==0.0.13
-gradio_client
-huggingface-hub>=0.18.0
-matplotlib
-numpy
-pandas
-python-dateutil
-tqdm
-transformers
-tokenizers>=0.15.0
-sentencepiece
+--extra-index-url https://download.pytorch.org/whl/cu114
+torch
+torchvision
+torchaudio==2.3.1
+diffusers==0.32.2
+easydict==1.13
+einops==0.8.1
+peft==0.14.0
+pillow==11.0.0
+protobuf==5.29.3
+requests==2.32.3
+safetensors==0.5.2
+sentencepiece==0.2.0
+transformers==4.49.0

src/about.py

@@ -1,72 +0,0 @@
-from dataclasses import dataclass
-from enum import Enum
-
-@dataclass
-class Task:
-    benchmark: str
-    metric: str
-    col_name: str
-
-
-# Select your tasks here
-# ---------------------------------------------------
-class Tasks(Enum):
-    # task_key in the json file, metric_key in the json file, name to display in the leaderboard 
-    task0 = Task("anli_r1", "acc", "ANLI")
-    task1 = Task("logiqa", "acc_norm", "LogiQA")
-
-NUM_FEWSHOT = 0 # Change with your few shot
-# ---------------------------------------------------
-
-
-
-# Your leaderboard name
-TITLE = """<h1 align="center" id="space-title">Demo leaderboard</h1>"""
-
-# What does your leaderboard evaluate?
-INTRODUCTION_TEXT = """
-Intro text
-"""
-
-# Which evaluations are you running? how can people reproduce what you have?
-LLM_BENCHMARKS_TEXT = f"""
-## How it works
-
-## Reproducibility
-To reproduce our results, here is the commands you can run:
-
-"""
-
-EVALUATION_QUEUE_TEXT = """
-## Some good practices before submitting a model
-
-### 1) Make sure you can load your model and tokenizer using AutoClasses:
-```python
-from transformers import AutoConfig, AutoModel, AutoTokenizer
-config = AutoConfig.from_pretrained("your model name", revision=revision)
-model = AutoModel.from_pretrained("your model name", revision=revision)
-tokenizer = AutoTokenizer.from_pretrained("your model name", revision=revision)
-```
-If this step fails, follow the error messages to debug your model before submitting it. It's likely your model has been improperly uploaded.
-
-Note: make sure your model is public!
-Note: if your model needs `use_remote_code=True`, we do not support this option yet but we are working on adding it, stay posted!
-
-### 2) Convert your model weights to [safetensors](https://huggingface.co/docs/safetensors/index)
-It's a new format for storing weights which is safer and faster to load and use. It will also allow us to add the number of parameters of your model to the `Extended Viewer`!
-
-### 3) Make sure your model has an open license!
-This is a leaderboard for Open LLMs, and we'd love for as many people as possible to know they can use your model 🤗
-
-### 4) Fill up your model card
-When we add extra information about models to the leaderboard, it will be automatically taken from the model card
-
-## In case of model failure
-If your model is displayed in the `FAILED` category, its execution stopped.
-Make sure you have followed the above steps first.
-If everything is done, check you can launch the EleutherAIHarness on your model locally, using the above command without modifications (you can add `--limit` to limit the number of examples per task).
-"""
-
-CITATION_BUTTON_LABEL = "Copy the following snippet to cite these results"
-CITATION_BUTTON_TEXT = r"""
-"""

src/display/css_html_js.py

@@ -1,105 +0,0 @@
-custom_css = """
-
-.markdown-text {
-    font-size: 16px !important;
-}
-
-#models-to-add-text {
-    font-size: 18px !important;
-}
-
-#citation-button span {
-    font-size: 16px !important;
-}
-
-#citation-button textarea {
-    font-size: 16px !important;
-}
-
-#citation-button > label > button {
-    margin: 6px;
-    transform: scale(1.3);
-}
-
-#leaderboard-table {
-    margin-top: 15px
-}
-
-#leaderboard-table-lite {
-    margin-top: 15px
-}
-
-#search-bar-table-box > div:first-child {
-    background: none;
-    border: none;
-}
- 
-#search-bar {
-    padding: 0px;
-}
-
-/* Limit the width of the first AutoEvalColumn so that names don't expand too much */
-#leaderboard-table td:nth-child(2),
-#leaderboard-table th:nth-child(2) {
-    max-width: 400px;
-    overflow: auto;
-    white-space: nowrap;
-}
-
-.tab-buttons button {
-    font-size: 20px;
-}
-
-#scale-logo {
-    border-style: none !important;
-    box-shadow: none;
-    display: block;
-    margin-left: auto;
-    margin-right: auto;
-    max-width: 600px;
-}
-
-#scale-logo .download {
-    display: none;
-}
-#filter_type{
-    border: 0;
-    padding-left: 0;
-    padding-top: 0;
-}
-#filter_type label {
-    display: flex;
-}
-#filter_type label > span{
-    margin-top: var(--spacing-lg);
-    margin-right: 0.5em;
-}
-#filter_type label > .wrap{
-    width: 103px;
-}
-#filter_type label > .wrap .wrap-inner{  
-    padding: 2px;
-}
-#filter_type label > .wrap .wrap-inner input{
-    width: 1px
-}
-#filter-columns-type{
-    border:0;
-    padding:0.5;
-}
-#filter-columns-size{
-    border:0;
-    padding:0.5;
-}
-#box-filter > .form{
-    border: 0
-}
-"""
-
-get_window_url_params = """
-    function(url_params) {
-        const params = new URLSearchParams(window.location.search);
-        url_params = Object.fromEntries(params);
-        return url_params;
-    }
-    """

src/display/formatting.py

@@ -1,27 +0,0 @@
-def model_hyperlink(link, model_name):
-    return f'<a target="_blank" href="{link}" style="color: var(--link-text-color); text-decoration: underline;text-decoration-style: dotted;">{model_name}</a>'
-
-
-def make_clickable_model(model_name):
-    link = f"https://huggingface.co/{model_name}"
-    return model_hyperlink(link, model_name)
-
-
-def styled_error(error):
-    return f"<p style='color: red; font-size: 20px; text-align: center;'>{error}</p>"
-
-
-def styled_warning(warn):
-    return f"<p style='color: orange; font-size: 20px; text-align: center;'>{warn}</p>"
-
-
-def styled_message(message):
-    return f"<p style='color: green; font-size: 20px; text-align: center;'>{message}</p>"
-
-
-def has_no_nan_values(df, columns):
-    return df[columns].notna().all(axis=1)
-
-
-def has_nan_values(df, columns):
-    return df[columns].isna().any(axis=1)

src/display/utils.py

@@ -1,110 +0,0 @@
-from dataclasses import dataclass, make_dataclass
-from enum import Enum
-
-import pandas as pd
-
-from src.about import Tasks
-
-def fields(raw_class):
-    return [v for k, v in raw_class.__dict__.items() if k[:2] != "__" and k[-2:] != "__"]
-
-
-# These classes are for user facing column names,
-# to avoid having to change them all around the code
-# when a modif is needed
-@dataclass
-class ColumnContent:
-    name: str
-    type: str
-    displayed_by_default: bool
-    hidden: bool = False
-    never_hidden: bool = False
-
-## Leaderboard columns
-auto_eval_column_dict = []
-# Init
-auto_eval_column_dict.append(["model_type_symbol", ColumnContent, ColumnContent("T", "str", True, never_hidden=True)])
-auto_eval_column_dict.append(["model", ColumnContent, ColumnContent("Model", "markdown", True, never_hidden=True)])
-#Scores
-auto_eval_column_dict.append(["average", ColumnContent, ColumnContent("Average ⬆️", "number", True)])
-for task in Tasks:
-    auto_eval_column_dict.append([task.name, ColumnContent, ColumnContent(task.value.col_name, "number", True)])
-# Model information
-auto_eval_column_dict.append(["model_type", ColumnContent, ColumnContent("Type", "str", False)])
-auto_eval_column_dict.append(["architecture", ColumnContent, ColumnContent("Architecture", "str", False)])
-auto_eval_column_dict.append(["weight_type", ColumnContent, ColumnContent("Weight type", "str", False, True)])
-auto_eval_column_dict.append(["precision", ColumnContent, ColumnContent("Precision", "str", False)])
-auto_eval_column_dict.append(["license", ColumnContent, ColumnContent("Hub License", "str", False)])
-auto_eval_column_dict.append(["params", ColumnContent, ColumnContent("#Params (B)", "number", False)])
-auto_eval_column_dict.append(["likes", ColumnContent, ColumnContent("Hub ❤️", "number", False)])
-auto_eval_column_dict.append(["still_on_hub", ColumnContent, ColumnContent("Available on the hub", "bool", False)])
-auto_eval_column_dict.append(["revision", ColumnContent, ColumnContent("Model sha", "str", False, False)])
-
-# We use make dataclass to dynamically fill the scores from Tasks
-AutoEvalColumn = make_dataclass("AutoEvalColumn", auto_eval_column_dict, frozen=True)
-
-## For the queue columns in the submission tab
-@dataclass(frozen=True)
-class EvalQueueColumn:  # Queue column
-    model = ColumnContent("model", "markdown", True)
-    revision = ColumnContent("revision", "str", True)
-    private = ColumnContent("private", "bool", True)
-    precision = ColumnContent("precision", "str", True)
-    weight_type = ColumnContent("weight_type", "str", "Original")
-    status = ColumnContent("status", "str", True)
-
-## All the model information that we might need
-@dataclass
-class ModelDetails:
-    name: str
-    display_name: str = ""
-    symbol: str = "" # emoji
-
-
-class ModelType(Enum):
-    PT = ModelDetails(name="pretrained", symbol="🟢")
-    FT = ModelDetails(name="fine-tuned", symbol="🔶")
-    IFT = ModelDetails(name="instruction-tuned", symbol="⭕")
-    RL = ModelDetails(name="RL-tuned", symbol="🟦")
-    Unknown = ModelDetails(name="", symbol="?")
-
-    def to_str(self, separator=" "):
-        return f"{self.value.symbol}{separator}{self.value.name}"
-
-    @staticmethod
-    def from_str(type):
-        if "fine-tuned" in type or "🔶" in type:
-            return ModelType.FT
-        if "pretrained" in type or "🟢" in type:
-            return ModelType.PT
-        if "RL-tuned" in type or "🟦" in type:
-            return ModelType.RL
-        if "instruction-tuned" in type or "⭕" in type:
-            return ModelType.IFT
-        return ModelType.Unknown
-
-class WeightType(Enum):
-    Adapter = ModelDetails("Adapter")
-    Original = ModelDetails("Original")
-    Delta = ModelDetails("Delta")
-
-class Precision(Enum):
-    float16 = ModelDetails("float16")
-    bfloat16 = ModelDetails("bfloat16")
-    Unknown = ModelDetails("?")
-
-    def from_str(precision):
-        if precision in ["torch.float16", "float16"]:
-            return Precision.float16
-        if precision in ["torch.bfloat16", "bfloat16"]:
-            return Precision.bfloat16
-        return Precision.Unknown
-
-# Column selection
-COLS = [c.name for c in fields(AutoEvalColumn) if not c.hidden]
-
-EVAL_COLS = [c.name for c in fields(EvalQueueColumn)]
-EVAL_TYPES = [c.type for c in fields(EvalQueueColumn)]
-
-BENCHMARK_COLS = [t.value.col_name for t in Tasks]
-

src/envs.py

@@ -1,25 +0,0 @@
-import os
-
-from huggingface_hub import HfApi
-
-# Info to change for your repository
-# ----------------------------------
-TOKEN = os.environ.get("HF_TOKEN") # A read/write token for your org
-
-OWNER = "demo-leaderboard-backend" # Change to your org - don't forget to create a results and request dataset, with the correct format!
-# ----------------------------------
-
-REPO_ID = f"{OWNER}/leaderboard"
-QUEUE_REPO = f"{OWNER}/requests"
-RESULTS_REPO = f"{OWNER}/results"
-
-# If you setup a cache later, just change HF_HOME
-CACHE_PATH=os.getenv("HF_HOME", ".")
-
-# Local caches
-EVAL_REQUESTS_PATH = os.path.join(CACHE_PATH, "eval-queue")
-EVAL_RESULTS_PATH = os.path.join(CACHE_PATH, "eval-results")
-EVAL_REQUESTS_PATH_BACKEND = os.path.join(CACHE_PATH, "eval-queue-bk")
-EVAL_RESULTS_PATH_BACKEND = os.path.join(CACHE_PATH, "eval-results-bk")
-
-API = HfApi(token=TOKEN)

src/layers_cache.py

@@ -0,0 +1,368 @@
+import inspect
+import math
+from typing import Callable, List, Optional, Tuple, Union
+from einops import rearrange
+import torch
+from torch import nn
+import torch.nn.functional as F
+from torch import Tensor
+from diffusers.models.attention_processor import Attention
+    
+class LoRALinearLayer(nn.Module):
+    def __init__(
+        self,
+        in_features: int,
+        out_features: int,
+        rank: int = 4,
+        network_alpha: Optional[float] = None,
+        device: Optional[Union[torch.device, str]] = None,
+        dtype: Optional[torch.dtype] = None,
+        cond_width=512,
+        cond_height=512,
+        number=0,
+        n_loras=1
+    ):
+        super().__init__()
+        self.down = nn.Linear(in_features, rank, bias=False, device=device, dtype=dtype)
+        self.up = nn.Linear(rank, out_features, bias=False, device=device, dtype=dtype)
+        # This value has the same meaning as the `--network_alpha` option in the kohya-ss trainer script.
+        # See https://github.com/darkstorm2150/sd-scripts/blob/main/docs/train_network_README-en.md#execute-learning
+        self.network_alpha = network_alpha
+        self.rank = rank
+        self.out_features = out_features
+        self.in_features = in_features
+
+        nn.init.normal_(self.down.weight, std=1 / rank)
+        nn.init.zeros_(self.up.weight)
+        
+        self.cond_height = cond_height
+        self.cond_width = cond_width
+        self.number = number
+        self.n_loras = n_loras
+
+    def forward(self, hidden_states: torch.Tensor) -> torch.Tensor:
+        orig_dtype = hidden_states.dtype
+        dtype = self.down.weight.dtype
+
+        ####
+        batch_size = hidden_states.shape[0]
+        cond_size = self.cond_width // 8 * self.cond_height // 8 * 16 // 64
+        block_size =  hidden_states.shape[1] - cond_size * self.n_loras
+        shape = (batch_size, hidden_states.shape[1], 3072)
+        mask = torch.ones(shape, device=hidden_states.device, dtype=dtype) 
+        mask[:, :block_size+self.number*cond_size, :] = 0
+        mask[:, block_size+(self.number+1)*cond_size:, :] = 0
+        hidden_states = mask * hidden_states
+        ####
+        
+        down_hidden_states = self.down(hidden_states.to(dtype))
+        up_hidden_states = self.up(down_hidden_states)
+
+        if self.network_alpha is not None:
+            up_hidden_states *= self.network_alpha / self.rank
+
+        return up_hidden_states.to(orig_dtype)
+    
+
+class MultiSingleStreamBlockLoraProcessor(nn.Module):
+    def __init__(self, dim: int, ranks=[], lora_weights=[], network_alphas=[], device=None, dtype=None, cond_width=512, cond_height=512, n_loras=1):
+        super().__init__()
+        # Initialize a list to store the LoRA layers
+        self.n_loras = n_loras
+        self.cond_width = cond_width
+        self.cond_height = cond_height
+        
+        self.q_loras = nn.ModuleList([
+            LoRALinearLayer(dim, dim, ranks[i],network_alphas[i], device=device, dtype=dtype, cond_width=cond_width, cond_height=cond_height, number=i, n_loras=n_loras)
+            for i in range(n_loras)
+        ])
+        self.k_loras = nn.ModuleList([
+            LoRALinearLayer(dim, dim, ranks[i],network_alphas[i], device=device, dtype=dtype, cond_width=cond_width, cond_height=cond_height, number=i, n_loras=n_loras)
+            for i in range(n_loras)
+        ])
+        self.v_loras = nn.ModuleList([
+            LoRALinearLayer(dim, dim, ranks[i],network_alphas[i], device=device, dtype=dtype, cond_width=cond_width, cond_height=cond_height, number=i, n_loras=n_loras)
+            for i in range(n_loras)
+        ])
+        self.lora_weights = lora_weights
+        self.bank_attn = None
+        self.bank_kv = []
+        
+
+    def __call__(self,
+        attn: Attention,
+        hidden_states: torch.FloatTensor,
+        encoder_hidden_states: torch.FloatTensor = None,
+        attention_mask: Optional[torch.FloatTensor] = None,
+        image_rotary_emb: Optional[torch.Tensor] = None,
+        use_cond = False
+    ) -> torch.FloatTensor:
+
+        batch_size, seq_len, _ = hidden_states.shape if encoder_hidden_states is None else encoder_hidden_states.shape
+        scaled_seq_len = hidden_states.shape[1]
+        cond_size = self.cond_width // 8 * self.cond_height // 8 * 16 // 64
+        block_size =  scaled_seq_len - cond_size * self.n_loras
+        scaled_cond_size = cond_size
+        scaled_block_size = block_size
+
+        if len(self.bank_kv)== 0:
+            cache = True
+        else:
+            cache = False
+        
+        if cache:
+            query = attn.to_q(hidden_states) 
+            key = attn.to_k(hidden_states) 
+            value = attn.to_v(hidden_states) 
+            for i in range(self.n_loras):
+                query = query + self.lora_weights[i] * self.q_loras[i](hidden_states)
+                key = key + self.lora_weights[i] * self.k_loras[i](hidden_states)
+                value = value + self.lora_weights[i] * self.v_loras[i](hidden_states)
+
+            inner_dim = key.shape[-1]
+            head_dim = inner_dim // attn.heads
+            
+            query = query.view(batch_size, -1, attn.heads, head_dim).transpose(1, 2)
+            key = key.view(batch_size, -1, attn.heads, head_dim).transpose(1, 2)
+            value = value.view(batch_size, -1, attn.heads, head_dim).transpose(1, 2)
+
+            self.bank_kv.append(key[:, :, scaled_block_size:, :])
+            self.bank_kv.append(value[:, :, scaled_block_size:, :])
+            
+            if attn.norm_q is not None:
+                query = attn.norm_q(query)
+            if attn.norm_k is not None:
+                key = attn.norm_k(key)
+
+            if image_rotary_emb is not None:
+                from diffusers.models.embeddings import apply_rotary_emb
+                query = apply_rotary_emb(query, image_rotary_emb)
+                key = apply_rotary_emb(key, image_rotary_emb)
+        
+            num_cond_blocks = self.n_loras
+            mask = torch.ones((scaled_seq_len, scaled_seq_len), device=hidden_states.device)
+            mask[ :scaled_block_size, :] = 0  # First block_size row
+            for i in range(num_cond_blocks):
+                start = i * scaled_cond_size + scaled_block_size
+                end = (i + 1) * scaled_cond_size + scaled_block_size
+                mask[start:end, start:end] = 0  # Diagonal blocks
+            mask = mask * -1e10
+            mask = mask.to(query.dtype)
+
+            hidden_states = F.scaled_dot_product_attention(query, key, value, dropout_p=0.0, is_causal=False, attn_mask=mask)
+            self.bank_attn = hidden_states[:, :, scaled_block_size:, :]
+            
+        else:
+            query = attn.to_q(hidden_states) 
+            key = attn.to_k(hidden_states)
+            value = attn.to_v(hidden_states)
+
+            inner_dim = query.shape[-1]
+            head_dim = inner_dim // attn.heads
+            
+            query = query.view(batch_size, -1, attn.heads, head_dim).transpose(1, 2)
+            key = key.view(batch_size, -1, attn.heads, head_dim).transpose(1, 2)
+            value = value.view(batch_size, -1, attn.heads, head_dim).transpose(1, 2)
+            
+            key = torch.concat([key[:, :, :scaled_block_size, :], self.bank_kv[0]], dim=-2)
+            value = torch.concat([value[:, :, :scaled_block_size, :], self.bank_kv[1]], dim=-2)
+
+            if attn.norm_q is not None:
+                query = attn.norm_q(query)
+            if attn.norm_k is not None:
+                key = attn.norm_k(key)
+
+            if image_rotary_emb is not None:
+                from diffusers.models.embeddings import apply_rotary_emb
+                query = apply_rotary_emb(query, image_rotary_emb)
+                key = apply_rotary_emb(key, image_rotary_emb)
+            
+            query = query[:, :, :scaled_block_size, :]
+
+            hidden_states = F.scaled_dot_product_attention(query, key, value, dropout_p=0.0, is_causal=False, attn_mask=None)
+            hidden_states = torch.concat([hidden_states, self.bank_attn], dim=-2)
+            
+        hidden_states = hidden_states.transpose(1, 2).reshape(batch_size, -1, attn.heads * head_dim)
+        hidden_states = hidden_states.to(query.dtype)
+
+        cond_hidden_states = hidden_states[:, block_size:,:]
+        hidden_states = hidden_states[:, : block_size,:]
+
+        return hidden_states if not use_cond else (hidden_states, cond_hidden_states)
+
+
+class MultiDoubleStreamBlockLoraProcessor(nn.Module):
+    def __init__(self, dim: int, ranks=[], lora_weights=[], network_alphas=[], device=None, dtype=None, cond_width=512, cond_height=512, n_loras=1):
+        super().__init__()
+        
+        # Initialize a list to store the LoRA layers
+        self.n_loras = n_loras
+        self.cond_width = cond_width
+        self.cond_height = cond_height
+        self.q_loras = nn.ModuleList([
+            LoRALinearLayer(dim, dim, ranks[i],network_alphas[i], device=device, dtype=dtype, cond_width=cond_width, cond_height=cond_height, number=i, n_loras=n_loras)
+            for i in range(n_loras)
+        ])
+        self.k_loras = nn.ModuleList([
+            LoRALinearLayer(dim, dim, ranks[i],network_alphas[i], device=device, dtype=dtype, cond_width=cond_width, cond_height=cond_height, number=i, n_loras=n_loras)
+            for i in range(n_loras)
+        ])
+        self.v_loras = nn.ModuleList([
+            LoRALinearLayer(dim, dim, ranks[i],network_alphas[i], device=device, dtype=dtype, cond_width=cond_width, cond_height=cond_height, number=i, n_loras=n_loras)
+            for i in range(n_loras)
+        ])
+        self.proj_loras = nn.ModuleList([
+            LoRALinearLayer(dim, dim, ranks[i],network_alphas[i], device=device, dtype=dtype, cond_width=cond_width, cond_height=cond_height, number=i, n_loras=n_loras)
+            for i in range(n_loras)
+        ])
+        self.lora_weights = lora_weights
+        self.bank_attn = None
+        self.bank_kv = []
+
+
+    def __call__(self,
+        attn: Attention,
+        hidden_states: torch.FloatTensor,
+        encoder_hidden_states: torch.FloatTensor = None,
+        attention_mask: Optional[torch.FloatTensor] = None,
+        image_rotary_emb: Optional[torch.Tensor] = None,
+        use_cond=False,
+    ) -> torch.FloatTensor:
+        
+        batch_size, _, _ = hidden_states.shape if encoder_hidden_states is None else encoder_hidden_states.shape
+        cond_size = self.cond_width // 8 * self.cond_height // 8 * 16 // 64 
+        block_size =  hidden_states.shape[1] - cond_size * self.n_loras
+        scaled_seq_len = encoder_hidden_states.shape[1] + hidden_states.shape[1]
+        scaled_cond_size = cond_size
+        scaled_block_size = scaled_seq_len - scaled_cond_size * self.n_loras
+
+        # `context` projections.
+        inner_dim = 3072
+        head_dim = inner_dim // attn.heads
+        encoder_hidden_states_query_proj = attn.add_q_proj(encoder_hidden_states) 
+        encoder_hidden_states_key_proj = attn.add_k_proj(encoder_hidden_states)
+        encoder_hidden_states_value_proj = attn.add_v_proj(encoder_hidden_states)
+
+        encoder_hidden_states_query_proj = encoder_hidden_states_query_proj.view(
+            batch_size, -1, attn.heads, head_dim
+        ).transpose(1, 2)
+        encoder_hidden_states_key_proj = encoder_hidden_states_key_proj.view(
+            batch_size, -1, attn.heads, head_dim
+        ).transpose(1, 2)
+        encoder_hidden_states_value_proj = encoder_hidden_states_value_proj.view(
+            batch_size, -1, attn.heads, head_dim
+        ).transpose(1, 2)
+
+        if attn.norm_added_q is not None:
+            encoder_hidden_states_query_proj = attn.norm_added_q(encoder_hidden_states_query_proj)
+        if attn.norm_added_k is not None:
+            encoder_hidden_states_key_proj = attn.norm_added_k(encoder_hidden_states_key_proj)
+        
+        if len(self.bank_kv)== 0:
+            cache = True
+        else:
+            cache = False
+        
+        if cache:
+            
+            query = attn.to_q(hidden_states) 
+            key = attn.to_k(hidden_states) 
+            value = attn.to_v(hidden_states) 
+            for i in range(self.n_loras):
+                query = query + self.lora_weights[i] * self.q_loras[i](hidden_states)
+                key = key + self.lora_weights[i] * self.k_loras[i](hidden_states)
+                value = value + self.lora_weights[i] * self.v_loras[i](hidden_states)
+
+            inner_dim = key.shape[-1]
+            head_dim = inner_dim // attn.heads
+            query = query.view(batch_size, -1, attn.heads, head_dim).transpose(1, 2)
+            key = key.view(batch_size, -1, attn.heads, head_dim).transpose(1, 2)
+            value = value.view(batch_size, -1, attn.heads, head_dim).transpose(1, 2)
+            
+            
+            self.bank_kv.append(key[:, :, block_size:, :])
+            self.bank_kv.append(value[:, :, block_size:, :])
+
+            if attn.norm_q is not None:
+                query = attn.norm_q(query)
+            if attn.norm_k is not None:
+                key = attn.norm_k(key)
+            
+            # attention
+            query = torch.cat([encoder_hidden_states_query_proj, query], dim=2)
+            key = torch.cat([encoder_hidden_states_key_proj, key], dim=2)
+            value = torch.cat([encoder_hidden_states_value_proj, value], dim=2)
+
+            if image_rotary_emb is not None:
+                from diffusers.models.embeddings import apply_rotary_emb
+                query = apply_rotary_emb(query, image_rotary_emb)
+                key = apply_rotary_emb(key, image_rotary_emb)
+            
+            num_cond_blocks = self.n_loras
+            mask = torch.ones((scaled_seq_len, scaled_seq_len), device=hidden_states.device)
+            mask[ :scaled_block_size, :] = 0  # First block_size row
+            for i in range(num_cond_blocks):
+                start = i * scaled_cond_size + scaled_block_size
+                end = (i + 1) * scaled_cond_size + scaled_block_size
+                mask[start:end, start:end] = 0  # Diagonal blocks
+            mask = mask * -1e10
+            mask = mask.to(query.dtype)
+
+            hidden_states = F.scaled_dot_product_attention(query, key, value, dropout_p=0.0, is_causal=False, attn_mask=mask)
+            self.bank_attn = hidden_states[:, :, scaled_block_size:, :]
+        
+        else:
+            query = attn.to_q(hidden_states) 
+            key = attn.to_k(hidden_states)
+            value = attn.to_v(hidden_states)
+    
+            inner_dim = query.shape[-1]
+            head_dim = inner_dim // attn.heads
+            
+            query = query.view(batch_size, -1, attn.heads, head_dim).transpose(1, 2)
+            key = key.view(batch_size, -1, attn.heads, head_dim).transpose(1, 2)
+            value = value.view(batch_size, -1, attn.heads, head_dim).transpose(1, 2)
+            
+            key = torch.concat([key[:, :, :block_size, :], self.bank_kv[0]], dim=-2)
+            value = torch.concat([value[:, :, :block_size, :], self.bank_kv[1]], dim=-2)
+            
+            if attn.norm_q is not None:
+                query = attn.norm_q(query)
+            if attn.norm_k is not None:
+                key = attn.norm_k(key)
+
+            # attention
+            query = torch.cat([encoder_hidden_states_query_proj, query], dim=2)
+            key = torch.cat([encoder_hidden_states_key_proj, key], dim=2)
+            value = torch.cat([encoder_hidden_states_value_proj, value], dim=2)
+
+            if image_rotary_emb is not None:
+                from diffusers.models.embeddings import apply_rotary_emb
+                query = apply_rotary_emb(query, image_rotary_emb)
+                key = apply_rotary_emb(key, image_rotary_emb)
+            
+            query = query[:, :, :scaled_block_size, :]
+
+            hidden_states = F.scaled_dot_product_attention(query, key, value, dropout_p=0.0, is_causal=False, attn_mask=None)
+            hidden_states = torch.concat([hidden_states, self.bank_attn], dim=-2)
+            
+        hidden_states = hidden_states.transpose(1, 2).reshape(batch_size, -1, attn.heads * head_dim)
+        hidden_states = hidden_states.to(query.dtype)
+        
+        encoder_hidden_states, hidden_states = (
+            hidden_states[:, : encoder_hidden_states.shape[1]],
+            hidden_states[:, encoder_hidden_states.shape[1] :],
+        )
+
+        # Linear projection (with LoRA weight applied to each proj layer)
+        hidden_states = attn.to_out[0](hidden_states)
+        for i in range(self.n_loras):
+             hidden_states = hidden_states + self.lora_weights[i] * self.proj_loras[i](hidden_states)
+        # dropout
+        hidden_states = attn.to_out[1](hidden_states)
+        encoder_hidden_states = attn.to_add_out(encoder_hidden_states)
+        
+        cond_hidden_states = hidden_states[:, block_size:,:]
+        hidden_states = hidden_states[:, :block_size,:]
+        
+        return (hidden_states, encoder_hidden_states, cond_hidden_states) if use_cond else (encoder_hidden_states, hidden_states)

src/leaderboard/read_evals.py

@@ -1,196 +0,0 @@
-import glob
-import json
-import math
-import os
-from dataclasses import dataclass
-
-import dateutil
-import numpy as np
-
-from src.display.formatting import make_clickable_model
-from src.display.utils import AutoEvalColumn, ModelType, Tasks, Precision, WeightType
-from src.submission.check_validity import is_model_on_hub
-
-
-@dataclass
-class EvalResult:
-    """Represents one full evaluation. Built from a combination of the result and request file for a given run.
-    """
-    eval_name: str # org_model_precision (uid)
-    full_model: str # org/model (path on hub)
-    org: str 
-    model: str
-    revision: str # commit hash, "" if main
-    results: dict
-    precision: Precision = Precision.Unknown
-    model_type: ModelType = ModelType.Unknown # Pretrained, fine tuned, ...
-    weight_type: WeightType = WeightType.Original # Original or Adapter
-    architecture: str = "Unknown" 
-    license: str = "?"
-    likes: int = 0
-    num_params: int = 0
-    date: str = "" # submission date of request file
-    still_on_hub: bool = False
-
-    @classmethod
-    def init_from_json_file(self, json_filepath):
-        """Inits the result from the specific model result file"""
-        with open(json_filepath) as fp:
-            data = json.load(fp)
-
-        config = data.get("config")
-
-        # Precision
-        precision = Precision.from_str(config.get("model_dtype"))
-
-        # Get model and org
-        org_and_model = config.get("model_name", config.get("model_args", None))
-        org_and_model = org_and_model.split("/", 1)
-
-        if len(org_and_model) == 1:
-            org = None
-            model = org_and_model[0]
-            result_key = f"{model}_{precision.value.name}"
-        else:
-            org = org_and_model[0]
-            model = org_and_model[1]
-            result_key = f"{org}_{model}_{precision.value.name}"
-        full_model = "/".join(org_and_model)
-
-        still_on_hub, _, model_config = is_model_on_hub(
-            full_model, config.get("model_sha", "main"), trust_remote_code=True, test_tokenizer=False
-        )
-        architecture = "?"
-        if model_config is not None:
-            architectures = getattr(model_config, "architectures", None)
-            if architectures:
-                architecture = ";".join(architectures)
-
-        # Extract results available in this file (some results are split in several files)
-        results = {}
-        for task in Tasks:
-            task = task.value
-
-            # We average all scores of a given metric (not all metrics are present in all files)
-            accs = np.array([v.get(task.metric, None) for k, v in data["results"].items() if task.benchmark == k])
-            if accs.size == 0 or any([acc is None for acc in accs]):
-                continue
-
-            mean_acc = np.mean(accs) * 100.0
-            results[task.benchmark] = mean_acc
-
-        return self(
-            eval_name=result_key,
-            full_model=full_model,
-            org=org,
-            model=model,
-            results=results,
-            precision=precision,  
-            revision= config.get("model_sha", ""),
-            still_on_hub=still_on_hub,
-            architecture=architecture
-        )
-
-    def update_with_request_file(self, requests_path):
-        """Finds the relevant request file for the current model and updates info with it"""
-        request_file = get_request_file_for_model(requests_path, self.full_model, self.precision.value.name)
-
-        try:
-            with open(request_file, "r") as f:
-                request = json.load(f)
-            self.model_type = ModelType.from_str(request.get("model_type", ""))
-            self.weight_type = WeightType[request.get("weight_type", "Original")]
-            self.license = request.get("license", "?")
-            self.likes = request.get("likes", 0)
-            self.num_params = request.get("params", 0)
-            self.date = request.get("submitted_time", "")
-        except Exception:
-            print(f"Could not find request file for {self.org}/{self.model} with precision {self.precision.value.name}")
-
-    def to_dict(self):
-        """Converts the Eval Result to a dict compatible with our dataframe display"""
-        average = sum([v for v in self.results.values() if v is not None]) / len(Tasks)
-        data_dict = {
-            "eval_name": self.eval_name,  # not a column, just a save name,
-            AutoEvalColumn.precision.name: self.precision.value.name,
-            AutoEvalColumn.model_type.name: self.model_type.value.name,
-            AutoEvalColumn.model_type_symbol.name: self.model_type.value.symbol,
-            AutoEvalColumn.weight_type.name: self.weight_type.value.name,
-            AutoEvalColumn.architecture.name: self.architecture,
-            AutoEvalColumn.model.name: make_clickable_model(self.full_model),
-            AutoEvalColumn.revision.name: self.revision,
-            AutoEvalColumn.average.name: average,
-            AutoEvalColumn.license.name: self.license,
-            AutoEvalColumn.likes.name: self.likes,
-            AutoEvalColumn.params.name: self.num_params,
-            AutoEvalColumn.still_on_hub.name: self.still_on_hub,
-        }
-
-        for task in Tasks:
-            data_dict[task.value.col_name] = self.results[task.value.benchmark]
-
-        return data_dict
-
-
-def get_request_file_for_model(requests_path, model_name, precision):
-    """Selects the correct request file for a given model. Only keeps runs tagged as FINISHED"""
-    request_files = os.path.join(
-        requests_path,
-        f"{model_name}_eval_request_*.json",
-    )
-    request_files = glob.glob(request_files)
-
-    # Select correct request file (precision)
-    request_file = ""
-    request_files = sorted(request_files, reverse=True)
-    for tmp_request_file in request_files:
-        with open(tmp_request_file, "r") as f:
-            req_content = json.load(f)
-            if (
-                req_content["status"] in ["FINISHED"]
-                and req_content["precision"] == precision.split(".")[-1]
-            ):
-                request_file = tmp_request_file
-    return request_file
-
-
-def get_raw_eval_results(results_path: str, requests_path: str) -> list[EvalResult]:
-    """From the path of the results folder root, extract all needed info for results"""
-    model_result_filepaths = []
-
-    for root, _, files in os.walk(results_path):
-        # We should only have json files in model results
-        if len(files) == 0 or any([not f.endswith(".json") for f in files]):
-            continue
-
-        # Sort the files by date
-        try:
-            files.sort(key=lambda x: x.removesuffix(".json").removeprefix("results_")[:-7])
-        except dateutil.parser._parser.ParserError:
-            files = [files[-1]]
-
-        for file in files:
-            model_result_filepaths.append(os.path.join(root, file))
-
-    eval_results = {}
-    for model_result_filepath in model_result_filepaths:
-        # Creation of result
-        eval_result = EvalResult.init_from_json_file(model_result_filepath)
-        eval_result.update_with_request_file(requests_path)
-
-        # Store results of same eval together
-        eval_name = eval_result.eval_name
-        if eval_name in eval_results.keys():
-            eval_results[eval_name].results.update({k: v for k, v in eval_result.results.items() if v is not None})
-        else:
-            eval_results[eval_name] = eval_result
-
-    results = []
-    for v in eval_results.values():
-        try:
-            v.to_dict() # we test if the dict version is complete
-            results.append(v)
-        except KeyError:  # not all eval values present
-            continue
-
-    return results

src/lora_helper.py

@@ -0,0 +1,196 @@
+from diffusers.models.attention_processor import FluxAttnProcessor2_0
+from safetensors import safe_open
+import re
+import torch
+from .layers_cache import MultiDoubleStreamBlockLoraProcessor, MultiSingleStreamBlockLoraProcessor
+
+device = "cuda"
+
+def load_safetensors(path):
+    tensors = {}
+    with safe_open(path, framework="pt", device="cpu") as f:
+        for key in f.keys():
+            tensors[key] = f.get_tensor(key)
+    return tensors
+
+def get_lora_rank(checkpoint):
+    for k in checkpoint.keys():
+        if k.endswith(".down.weight"):
+            return checkpoint[k].shape[0]
+
+def load_checkpoint(local_path):
+    if local_path is not None:
+        if '.safetensors' in local_path:
+            print(f"Loading .safetensors checkpoint from {local_path}")
+            checkpoint = load_safetensors(local_path)
+        else:
+            print(f"Loading checkpoint from {local_path}")
+            checkpoint = torch.load(local_path, map_location='cpu')
+    return checkpoint
+
+def update_model_with_lora(checkpoint, lora_weights, transformer, cond_size):
+        number = len(lora_weights)
+        ranks = [get_lora_rank(checkpoint) for _ in range(number)]
+        lora_attn_procs = {}
+        double_blocks_idx = list(range(19))
+        single_blocks_idx = list(range(38))
+        for name, attn_processor in transformer.attn_processors.items():
+            match = re.search(r'\.(\d+)\.', name)
+            if match:
+                layer_index = int(match.group(1))
+            
+            if name.startswith("transformer_blocks") and layer_index in double_blocks_idx:
+                
+                lora_state_dicts = {}
+                for key, value in checkpoint.items():
+                    # Match based on the layer index in the key (assuming the key contains layer index)
+                    if re.search(r'\.(\d+)\.', key):
+                        checkpoint_layer_index = int(re.search(r'\.(\d+)\.', key).group(1))
+                        if checkpoint_layer_index == layer_index and key.startswith("transformer_blocks"):
+                            lora_state_dicts[key] = value
+                
+                lora_attn_procs[name] = MultiDoubleStreamBlockLoraProcessor(
+                    dim=3072, ranks=ranks, network_alphas=ranks, lora_weights=lora_weights, device=device, dtype=torch.bfloat16, cond_width=cond_size, cond_height=cond_size, n_loras=number
+                )
+                
+                # Load the weights from the checkpoint dictionary into the corresponding layers
+                for n in range(number):
+                    lora_attn_procs[name].q_loras[n].down.weight.data = lora_state_dicts.get(f'{name}.q_loras.{n}.down.weight', None)
+                    lora_attn_procs[name].q_loras[n].up.weight.data = lora_state_dicts.get(f'{name}.q_loras.{n}.up.weight', None)
+                    lora_attn_procs[name].k_loras[n].down.weight.data = lora_state_dicts.get(f'{name}.k_loras.{n}.down.weight', None)
+                    lora_attn_procs[name].k_loras[n].up.weight.data = lora_state_dicts.get(f'{name}.k_loras.{n}.up.weight', None)
+                    lora_attn_procs[name].v_loras[n].down.weight.data = lora_state_dicts.get(f'{name}.v_loras.{n}.down.weight', None)
+                    lora_attn_procs[name].v_loras[n].up.weight.data = lora_state_dicts.get(f'{name}.v_loras.{n}.up.weight', None)
+                    lora_attn_procs[name].proj_loras[n].down.weight.data = lora_state_dicts.get(f'{name}.proj_loras.{n}.down.weight', None)
+                    lora_attn_procs[name].proj_loras[n].up.weight.data = lora_state_dicts.get(f'{name}.proj_loras.{n}.up.weight', None)
+                    lora_attn_procs[name].to(device)
+                
+            elif name.startswith("single_transformer_blocks") and layer_index in single_blocks_idx:
+                
+                lora_state_dicts = {}
+                for key, value in checkpoint.items():
+                    # Match based on the layer index in the key (assuming the key contains layer index)
+                    if re.search(r'\.(\d+)\.', key):
+                        checkpoint_layer_index = int(re.search(r'\.(\d+)\.', key).group(1))
+                        if checkpoint_layer_index == layer_index and key.startswith("single_transformer_blocks"):
+                            lora_state_dicts[key] = value
+                
+                lora_attn_procs[name] = MultiSingleStreamBlockLoraProcessor(
+                    dim=3072, ranks=ranks, network_alphas=ranks, lora_weights=lora_weights, device=device, dtype=torch.bfloat16, cond_width=cond_size, cond_height=cond_size, n_loras=number
+                )
+                # Load the weights from the checkpoint dictionary into the corresponding layers
+                for n in range(number):
+                    lora_attn_procs[name].q_loras[n].down.weight.data = lora_state_dicts.get(f'{name}.q_loras.{n}.down.weight', None)
+                    lora_attn_procs[name].q_loras[n].up.weight.data = lora_state_dicts.get(f'{name}.q_loras.{n}.up.weight', None)
+                    lora_attn_procs[name].k_loras[n].down.weight.data = lora_state_dicts.get(f'{name}.k_loras.{n}.down.weight', None)
+                    lora_attn_procs[name].k_loras[n].up.weight.data = lora_state_dicts.get(f'{name}.k_loras.{n}.up.weight', None)
+                    lora_attn_procs[name].v_loras[n].down.weight.data = lora_state_dicts.get(f'{name}.v_loras.{n}.down.weight', None)
+                    lora_attn_procs[name].v_loras[n].up.weight.data = lora_state_dicts.get(f'{name}.v_loras.{n}.up.weight', None)
+                    lora_attn_procs[name].to(device)
+            else:
+                lora_attn_procs[name] = FluxAttnProcessor2_0()
+
+        transformer.set_attn_processor(lora_attn_procs)
+        
+
+def update_model_with_multi_lora(checkpoints, lora_weights, transformer, cond_size):
+        ck_number = len(checkpoints)
+        cond_lora_number = [len(ls) for ls in lora_weights]
+        cond_number = sum(cond_lora_number)
+        ranks = [get_lora_rank(checkpoint) for checkpoint in checkpoints]
+        multi_lora_weight = []
+        for ls in lora_weights:
+            for n in ls:
+                multi_lora_weight.append(n)
+        
+        lora_attn_procs = {}
+        double_blocks_idx = list(range(19))
+        single_blocks_idx = list(range(38))
+        for name, attn_processor in transformer.attn_processors.items():
+            match = re.search(r'\.(\d+)\.', name)
+            if match:
+                layer_index = int(match.group(1))
+            
+            if name.startswith("transformer_blocks") and layer_index in double_blocks_idx:
+                lora_state_dicts = [{} for _ in range(ck_number)]
+                for idx, checkpoint in enumerate(checkpoints):
+                    for key, value in checkpoint.items():
+                        # Match based on the layer index in the key (assuming the key contains layer index)
+                        if re.search(r'\.(\d+)\.', key):
+                            checkpoint_layer_index = int(re.search(r'\.(\d+)\.', key).group(1))
+                            if checkpoint_layer_index == layer_index and key.startswith("transformer_blocks"):
+                                lora_state_dicts[idx][key] = value
+                
+                lora_attn_procs[name] = MultiDoubleStreamBlockLoraProcessor(
+                    dim=3072, ranks=ranks, network_alphas=ranks, lora_weights=multi_lora_weight, device=device, dtype=torch.bfloat16, cond_width=cond_size, cond_height=cond_size, n_loras=cond_number
+                )
+                
+                # Load the weights from the checkpoint dictionary into the corresponding layers
+                num = 0
+                for idx in range(ck_number):
+                    for n in range(cond_lora_number[idx]):
+                        lora_attn_procs[name].q_loras[num].down.weight.data = lora_state_dicts[idx].get(f'{name}.q_loras.{n}.down.weight', None)
+                        lora_attn_procs[name].q_loras[num].up.weight.data = lora_state_dicts[idx].get(f'{name}.q_loras.{n}.up.weight', None)
+                        lora_attn_procs[name].k_loras[num].down.weight.data = lora_state_dicts[idx].get(f'{name}.k_loras.{n}.down.weight', None)
+                        lora_attn_procs[name].k_loras[num].up.weight.data = lora_state_dicts[idx].get(f'{name}.k_loras.{n}.up.weight', None)
+                        lora_attn_procs[name].v_loras[num].down.weight.data = lora_state_dicts[idx].get(f'{name}.v_loras.{n}.down.weight', None)
+                        lora_attn_procs[name].v_loras[num].up.weight.data = lora_state_dicts[idx].get(f'{name}.v_loras.{n}.up.weight', None)
+                        lora_attn_procs[name].proj_loras[num].down.weight.data = lora_state_dicts[idx].get(f'{name}.proj_loras.{n}.down.weight', None)
+                        lora_attn_procs[name].proj_loras[num].up.weight.data = lora_state_dicts[idx].get(f'{name}.proj_loras.{n}.up.weight', None)
+                        lora_attn_procs[name].to(device)
+                        num += 1
+                
+            elif name.startswith("single_transformer_blocks") and layer_index in single_blocks_idx:
+                
+                lora_state_dicts = [{} for _ in range(ck_number)]
+                for idx, checkpoint in enumerate(checkpoints):
+                    for key, value in checkpoint.items():
+                        # Match based on the layer index in the key (assuming the key contains layer index)
+                        if re.search(r'\.(\d+)\.', key):
+                            checkpoint_layer_index = int(re.search(r'\.(\d+)\.', key).group(1))
+                            if checkpoint_layer_index == layer_index and key.startswith("single_transformer_blocks"):
+                                lora_state_dicts[idx][key] = value
+                
+                lora_attn_procs[name] = MultiSingleStreamBlockLoraProcessor(
+                    dim=3072, ranks=ranks, network_alphas=ranks, lora_weights=multi_lora_weight, device=device, dtype=torch.bfloat16, cond_width=cond_size, cond_height=cond_size, n_loras=cond_number
+                )
+                # Load the weights from the checkpoint dictionary into the corresponding layers
+                num = 0
+                for idx in range(ck_number):
+                    for n in range(cond_lora_number[idx]):
+                        lora_attn_procs[name].q_loras[num].down.weight.data = lora_state_dicts[idx].get(f'{name}.q_loras.{n}.down.weight', None)
+                        lora_attn_procs[name].q_loras[num].up.weight.data = lora_state_dicts[idx].get(f'{name}.q_loras.{n}.up.weight', None)
+                        lora_attn_procs[name].k_loras[num].down.weight.data = lora_state_dicts[idx].get(f'{name}.k_loras.{n}.down.weight', None)
+                        lora_attn_procs[name].k_loras[num].up.weight.data = lora_state_dicts[idx].get(f'{name}.k_loras.{n}.up.weight', None)
+                        lora_attn_procs[name].v_loras[num].down.weight.data = lora_state_dicts[idx].get(f'{name}.v_loras.{n}.down.weight', None)
+                        lora_attn_procs[name].v_loras[num].up.weight.data = lora_state_dicts[idx].get(f'{name}.v_loras.{n}.up.weight', None)
+                        lora_attn_procs[name].to(device)
+                        num += 1
+
+            else:
+                lora_attn_procs[name] = FluxAttnProcessor2_0()
+
+        transformer.set_attn_processor(lora_attn_procs)
+
+
+def set_single_lora(transformer, local_path, lora_weights=[], cond_size=512):
+    checkpoint = load_checkpoint(local_path)
+    update_model_with_lora(checkpoint, lora_weights, transformer, cond_size)
+   
+def set_multi_lora(transformer, local_paths, lora_weights=[[]], cond_size=512):
+    checkpoints = [load_checkpoint(local_path) for local_path in local_paths]
+    update_model_with_multi_lora(checkpoints, lora_weights, transformer, cond_size)
+
+def unset_lora(transformer):
+    lora_attn_procs = {}
+    for name, attn_processor in transformer.attn_processors.items():
+        lora_attn_procs[name] = FluxAttnProcessor2_0()
+    transformer.set_attn_processor(lora_attn_procs)
+
+
+'''
+unset_lora(pipe.transformer)
+lora_path = "./lora.safetensors"
+lora_weights = [1, 1]
+set_lora(pipe.transformer, local_path=lora_path, lora_weights=lora_weights, cond_size=512)
+'''

src/pipeline.py

@@ -0,0 +1,745 @@
+import inspect
+from typing import Any, Callable, Dict, List, Optional, Union
+
+import numpy as np
+import torch
+from transformers import CLIPTextModel, CLIPTokenizer, T5EncoderModel, T5TokenizerFast
+
+from diffusers.image_processor import (VaeImageProcessor)
+from diffusers.loaders import FluxLoraLoaderMixin, FromSingleFileMixin
+from diffusers.models.autoencoders import AutoencoderKL
+from diffusers.schedulers import FlowMatchEulerDiscreteScheduler
+from diffusers.utils import (
+    USE_PEFT_BACKEND,
+    is_torch_xla_available,
+    logging,
+    scale_lora_layers,
+    unscale_lora_layers,
+)
+from diffusers.utils.torch_utils import randn_tensor
+from diffusers.pipelines.pipeline_utils import DiffusionPipeline
+from diffusers.pipelines.flux.pipeline_output import FluxPipelineOutput
+from torchvision.transforms.functional import pad
+from .transformer_flux import FluxTransformer2DModel
+
+if is_torch_xla_available():
+    import torch_xla.core.xla_model as xm
+
+    XLA_AVAILABLE = True
+else:
+    XLA_AVAILABLE = False
+
+logger = logging.get_logger(__name__)  # pylint: disable=invalid-name
+
+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 prepare_latent_image_ids_(height, width, device, dtype):
+    latent_image_ids = torch.zeros(height//2, width//2, 3, device=device, dtype=dtype)
+    latent_image_ids[..., 1] = latent_image_ids[..., 1] + torch.arange(height//2, device=device)[:, None]  # y
+    latent_image_ids[..., 2] = latent_image_ids[..., 2] + torch.arange(width//2, device=device)[None, :]   # x
+    return latent_image_ids
+
+def prepare_latent_subject_ids(height, width, device, dtype):
+    latent_image_ids = torch.zeros(height // 2, width // 2, 3, device=device, dtype=dtype)
+    latent_image_ids[..., 1] = latent_image_ids[..., 1] + torch.arange(height // 2, device=device)[:, None]
+    latent_image_ids[..., 2] = latent_image_ids[..., 2] + torch.arange(width // 2, device=device)[None, :]
+    latent_image_id_height, latent_image_id_width, latent_image_id_channels = latent_image_ids.shape
+    latent_image_ids = latent_image_ids.reshape(
+        latent_image_id_height * latent_image_id_width, latent_image_id_channels
+    )
+    return latent_image_ids.to(device=device, dtype=dtype)
+
+def resize_position_encoding(batch_size, original_height, original_width, target_height, target_width, device, dtype):
+    latent_image_ids = prepare_latent_image_ids_(original_height, original_width, device, dtype)
+    latent_image_id_height, latent_image_id_width, latent_image_id_channels = latent_image_ids.shape
+    latent_image_ids = latent_image_ids.reshape(
+        latent_image_id_height * latent_image_id_width, latent_image_id_channels
+    )
+    
+    scale_h = original_height / target_height
+    scale_w = original_width / target_width
+    latent_image_ids_resized = torch.zeros(target_height//2, target_width//2, 3, device=device, dtype=dtype)
+    latent_image_ids_resized[..., 1] = latent_image_ids_resized[..., 1] + torch.arange(target_height//2, device=device)[:, None] * scale_h
+    latent_image_ids_resized[..., 2] = latent_image_ids_resized[..., 2] + torch.arange(target_width//2, device=device)[None, :] * scale_w
+    
+    cond_latent_image_id_height, cond_latent_image_id_width, cond_latent_image_id_channels = latent_image_ids_resized.shape
+    cond_latent_image_ids = latent_image_ids_resized.reshape(
+            cond_latent_image_id_height * cond_latent_image_id_width, cond_latent_image_id_channels
+        )
+    return latent_image_ids, cond_latent_image_ids 
+    
+# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_img2img.retrieve_latents
+def retrieve_latents(
+        encoder_output: torch.Tensor, generator: Optional[torch.Generator] = None, sample_mode: str = "sample"
+):
+    if hasattr(encoder_output, "latent_dist") and sample_mode == "sample":
+        return encoder_output.latent_dist.sample(generator)
+    elif hasattr(encoder_output, "latent_dist") and sample_mode == "argmax":
+        return encoder_output.latent_dist.mode()
+    elif hasattr(encoder_output, "latents"):
+        return encoder_output.latents
+    else:
+        raise AttributeError("Could not access latents of provided encoder_output")
+
+
+# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.retrieve_timesteps
+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:
+        accepts_timesteps = "timesteps" in set(inspect.signature(scheduler.set_timesteps).parameters.keys())
+        if not accepts_timesteps:
+            raise ValueError(
+                f"The current scheduler class {scheduler.__class__}'s `set_timesteps` does not support custom"
+                f" timestep schedules. Please check whether you are using the correct scheduler."
+            )
+        scheduler.set_timesteps(timesteps=timesteps, device=device, **kwargs)
+        timesteps = scheduler.timesteps
+        num_inference_steps = len(timesteps)
+    elif sigmas is not None:
+        accept_sigmas = "sigmas" in set(inspect.signature(scheduler.set_timesteps).parameters.keys())
+        if not accept_sigmas:
+            raise ValueError(
+                f"The current scheduler class {scheduler.__class__}'s `set_timesteps` does not support custom"
+                f" sigmas schedules. Please check whether you are using the correct scheduler."
+            )
+        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
+
+
+class FluxPipeline(DiffusionPipeline, FluxLoraLoaderMixin, FromSingleFileMixin):
+    def __init__(
+            self,
+            scheduler: FlowMatchEulerDiscreteScheduler,
+            vae: AutoencoderKL,
+            text_encoder: CLIPTextModel,
+            tokenizer: CLIPTokenizer,
+            text_encoder_2: T5EncoderModel,
+            tokenizer_2: T5TokenizerFast,
+            transformer: FluxTransformer2DModel,
+    ):
+        super().__init__()
+
+        self.register_modules(
+            vae=vae,
+            text_encoder=text_encoder,
+            text_encoder_2=text_encoder_2,
+            tokenizer=tokenizer,
+            tokenizer_2=tokenizer_2,
+            transformer=transformer,
+            scheduler=scheduler,
+        )
+        self.vae_scale_factor = (
+            2 ** (len(self.vae.config.block_out_channels)) if hasattr(self, "vae") and self.vae is not None else 16
+        )
+        self.image_processor = VaeImageProcessor(vae_scale_factor=self.vae_scale_factor)
+        self.tokenizer_max_length = (
+            self.tokenizer.model_max_length if hasattr(self, "tokenizer") and self.tokenizer is not None else 77
+        )
+        self.default_sample_size = 64
+
+    def _get_t5_prompt_embeds(
+            self,
+            prompt: Union[str, List[str]] = None,
+            num_images_per_prompt: int = 1,
+            max_sequence_length: int = 512,
+            device: Optional[torch.device] = None,
+            dtype: Optional[torch.dtype] = None,
+    ):
+        device = device or self._execution_device
+        dtype = dtype or self.text_encoder.dtype
+
+        prompt = [prompt] if isinstance(prompt, str) else prompt
+        batch_size = len(prompt)
+
+        text_inputs = self.tokenizer_2(
+            prompt,
+            padding="max_length",
+            max_length=max_sequence_length,
+            truncation=True,
+            return_length=False,
+            return_overflowing_tokens=False,
+            return_tensors="pt",
+        )
+        text_input_ids = text_inputs.input_ids
+        untruncated_ids = self.tokenizer_2(prompt, padding="longest", return_tensors="pt").input_ids
+
+        if untruncated_ids.shape[-1] >= text_input_ids.shape[-1] and not torch.equal(text_input_ids, untruncated_ids):
+            removed_text = self.tokenizer_2.batch_decode(untruncated_ids[:, self.tokenizer_max_length - 1: -1])
+            logger.warning(
+                "The following part of your input was truncated because `max_sequence_length` is set to "
+                f" {max_sequence_length} tokens: {removed_text}"
+            )
+
+        prompt_embeds = self.text_encoder_2(text_input_ids.to(device), output_hidden_states=False)[0]
+
+        dtype = self.text_encoder_2.dtype
+        prompt_embeds = prompt_embeds.to(dtype=dtype, device=device)
+
+        _, seq_len, _ = prompt_embeds.shape
+
+        # duplicate text embeddings and attention mask for each generation per prompt, using mps friendly method
+        prompt_embeds = prompt_embeds.repeat(1, num_images_per_prompt, 1)
+        prompt_embeds = prompt_embeds.view(batch_size * num_images_per_prompt, seq_len, -1)
+
+        return prompt_embeds
+
+    def _get_clip_prompt_embeds(
+            self,
+            prompt: Union[str, List[str]],
+            num_images_per_prompt: int = 1,
+            device: Optional[torch.device] = None,
+    ):
+        device = device or self._execution_device
+
+        prompt = [prompt] if isinstance(prompt, str) else prompt
+        batch_size = len(prompt)
+
+        text_inputs = self.tokenizer(
+            prompt,
+            padding="max_length",
+            max_length=self.tokenizer_max_length,
+            truncation=True,
+            return_overflowing_tokens=False,
+            return_length=False,
+            return_tensors="pt",
+        )
+
+        text_input_ids = text_inputs.input_ids
+        untruncated_ids = self.tokenizer(prompt, padding="longest", return_tensors="pt").input_ids
+        if untruncated_ids.shape[-1] >= text_input_ids.shape[-1] and not torch.equal(text_input_ids, untruncated_ids):
+            removed_text = self.tokenizer.batch_decode(untruncated_ids[:, self.tokenizer_max_length - 1: -1])
+            logger.warning(
+                "The following part of your input was truncated because CLIP can only handle sequences up to"
+                f" {self.tokenizer_max_length} tokens: {removed_text}"
+            )
+        prompt_embeds = self.text_encoder(text_input_ids.to(device), output_hidden_states=False)
+
+        # Use pooled output of CLIPTextModel
+        prompt_embeds = prompt_embeds.pooler_output
+        prompt_embeds = prompt_embeds.to(dtype=self.text_encoder.dtype, device=device)
+
+        # duplicate text embeddings for each generation per prompt, using mps friendly method
+        prompt_embeds = prompt_embeds.repeat(1, num_images_per_prompt)
+        prompt_embeds = prompt_embeds.view(batch_size * num_images_per_prompt, -1)
+
+        return prompt_embeds
+
+    def encode_prompt(
+            self,
+            prompt: Union[str, List[str]],
+            prompt_2: Union[str, List[str]],
+            device: Optional[torch.device] = None,
+            num_images_per_prompt: int = 1,
+            prompt_embeds: Optional[torch.FloatTensor] = None,
+            pooled_prompt_embeds: Optional[torch.FloatTensor] = None,
+            max_sequence_length: int = 512,
+            lora_scale: Optional[float] = None,
+    ):
+        device = device or self._execution_device
+
+        # set lora scale so that monkey patched LoRA
+        # function of text encoder can correctly access it
+        if lora_scale is not None and isinstance(self, FluxLoraLoaderMixin):
+            self._lora_scale = lora_scale
+
+            # dynamically adjust the LoRA scale
+            if self.text_encoder is not None and USE_PEFT_BACKEND:
+                scale_lora_layers(self.text_encoder, lora_scale)
+            if self.text_encoder_2 is not None and USE_PEFT_BACKEND:
+                scale_lora_layers(self.text_encoder_2, lora_scale)
+
+        prompt = [prompt] if isinstance(prompt, str) else prompt
+
+        if prompt_embeds is None:
+            prompt_2 = prompt_2 or prompt
+            prompt_2 = [prompt_2] if isinstance(prompt_2, str) else prompt_2
+
+            # We only use the pooled prompt output from the CLIPTextModel
+            pooled_prompt_embeds = self._get_clip_prompt_embeds(
+                prompt=prompt,
+                device=device,
+                num_images_per_prompt=num_images_per_prompt,
+            )
+            prompt_embeds = self._get_t5_prompt_embeds(
+                prompt=prompt_2,
+                num_images_per_prompt=num_images_per_prompt,
+                max_sequence_length=max_sequence_length,
+                device=device,
+            )
+
+        if self.text_encoder is not None:
+            if isinstance(self, FluxLoraLoaderMixin) and USE_PEFT_BACKEND:
+                # Retrieve the original scale by scaling back the LoRA layers
+                unscale_lora_layers(self.text_encoder, lora_scale)
+
+        if self.text_encoder_2 is not None:
+            if isinstance(self, FluxLoraLoaderMixin) and USE_PEFT_BACKEND:
+                # Retrieve the original scale by scaling back the LoRA layers
+                unscale_lora_layers(self.text_encoder_2, lora_scale)
+
+        dtype = self.text_encoder.dtype if self.text_encoder is not None else self.transformer.dtype
+        text_ids = torch.zeros(prompt_embeds.shape[1], 3).to(device=device, dtype=dtype)
+
+        return prompt_embeds, pooled_prompt_embeds, text_ids
+
+    # Copied from diffusers.pipelines.stable_diffusion_3.pipeline_stable_diffusion_3_inpaint.StableDiffusion3InpaintPipeline._encode_vae_image
+    def _encode_vae_image(self, image: torch.Tensor, generator: torch.Generator):
+        if isinstance(generator, list):
+            image_latents = [
+                retrieve_latents(self.vae.encode(image[i: i + 1]), generator=generator[i])
+                for i in range(image.shape[0])
+            ]
+            image_latents = torch.cat(image_latents, dim=0)
+        else:
+            image_latents = retrieve_latents(self.vae.encode(image), generator=generator)
+
+        image_latents = (image_latents - self.vae.config.shift_factor) * self.vae.config.scaling_factor
+
+        return image_latents
+
+    def check_inputs(
+            self,
+            prompt,
+            prompt_2,
+            height,
+            width,
+            prompt_embeds=None,
+            pooled_prompt_embeds=None,
+            callback_on_step_end_tensor_inputs=None,
+            max_sequence_length=None,
+    ):
+        if height % 8 != 0 or width % 8 != 0:
+            raise ValueError(f"`height` and `width` have to be divisible by 8 but are {height} and {width}.")
+
+        if prompt is not None and prompt_embeds is not None:
+            raise ValueError(
+                f"Cannot forward both `prompt`: {prompt} and `prompt_embeds`: {prompt_embeds}. Please make sure to"
+                " only forward one of the two."
+            )
+        elif prompt_2 is not None and prompt_embeds is not None:
+            raise ValueError(
+                f"Cannot forward both `prompt_2`: {prompt_2} and `prompt_embeds`: {prompt_embeds}. Please make sure to"
+                " only forward one of the two."
+            )
+        elif prompt is None and prompt_embeds is None:
+            raise ValueError(
+                "Provide either `prompt` or `prompt_embeds`. Cannot leave both `prompt` and `prompt_embeds` undefined."
+            )
+        elif prompt is not None and (not isinstance(prompt, str) and not isinstance(prompt, list)):
+            raise ValueError(f"`prompt` has to be of type `str` or `list` but is {type(prompt)}")
+        elif prompt_2 is not None and (not isinstance(prompt_2, str) and not isinstance(prompt_2, list)):
+            raise ValueError(f"`prompt_2` has to be of type `str` or `list` but is {type(prompt_2)}")
+
+        if prompt_embeds is not None and pooled_prompt_embeds is None:
+            raise ValueError(
+                "If `prompt_embeds` are provided, `pooled_prompt_embeds` also have to be passed. Make sure to generate `pooled_prompt_embeds` from the same text encoder that was used to generate `prompt_embeds`."
+            )
+
+        if max_sequence_length is not None and max_sequence_length > 512:
+            raise ValueError(f"`max_sequence_length` cannot be greater than 512 but is {max_sequence_length}")
+
+    @staticmethod
+    def _prepare_latent_image_ids(batch_size, height, width, device, dtype):
+        latent_image_ids = torch.zeros(height // 2, width // 2, 3)
+        latent_image_ids[..., 1] = latent_image_ids[..., 1] + torch.arange(height // 2)[:, None]
+        latent_image_ids[..., 2] = latent_image_ids[..., 2] + torch.arange(width // 2)[None, :]
+        latent_image_id_height, latent_image_id_width, latent_image_id_channels = latent_image_ids.shape
+        latent_image_ids = latent_image_ids.reshape(
+            latent_image_id_height * latent_image_id_width, latent_image_id_channels
+        )
+        return latent_image_ids.to(device=device, dtype=dtype)
+
+    @staticmethod
+    def _pack_latents(latents, batch_size, num_channels_latents, height, width):
+        latents = latents.view(batch_size, num_channels_latents, height // 2, 2, width // 2, 2)
+        latents = latents.permute(0, 2, 4, 1, 3, 5)
+        latents = latents.reshape(batch_size, (height // 2) * (width // 2), num_channels_latents * 4)
+        return latents
+
+    @staticmethod
+    def _unpack_latents(latents, height, width, vae_scale_factor):
+        batch_size, num_patches, channels = latents.shape
+
+        height = height // vae_scale_factor
+        width = width // vae_scale_factor
+
+        latents = latents.view(batch_size, height, width, channels // 4, 2, 2)
+        latents = latents.permute(0, 3, 1, 4, 2, 5)
+
+        latents = latents.reshape(batch_size, channels // (2 * 2), height * 2, width * 2)
+
+        return latents
+
+    def enable_vae_slicing(self):
+        r"""
+        Enable sliced VAE decoding. When this option is enabled, the VAE will split the input tensor in slices to
+        compute decoding in several steps. This is useful to save some memory and allow larger batch sizes.
+        """
+        self.vae.enable_slicing()
+
+    def disable_vae_slicing(self):
+        r"""
+        Disable sliced VAE decoding. If `enable_vae_slicing` was previously enabled, this method will go back to
+        computing decoding in one step.
+        """
+        self.vae.disable_slicing()
+
+    def enable_vae_tiling(self):
+        r"""
+        Enable tiled VAE decoding. When this option is enabled, the VAE will split the input tensor into tiles to
+        compute decoding and encoding in several steps. This is useful for saving a large amount of memory and to allow
+        processing larger images.
+        """
+        self.vae.enable_tiling()
+
+    def disable_vae_tiling(self):
+        r"""
+        Disable tiled VAE decoding. If `enable_vae_tiling` was previously enabled, this method will go back to
+        computing decoding in one step.
+        """
+        self.vae.disable_tiling()
+
+    def prepare_latents(
+            self,
+            batch_size,
+            num_channels_latents,
+            height,
+            width,
+            dtype,
+            device,
+            generator,
+            subject_image,
+            condition_image,
+            latents=None,
+            cond_number=1,
+            sub_number=1
+    ):
+        height_cond = 2 * (self.cond_size // self.vae_scale_factor)
+        width_cond = 2 * (self.cond_size // self.vae_scale_factor)
+        height = 2 * (int(height) // self.vae_scale_factor)  
+        width = 2 * (int(width) // self.vae_scale_factor)
+
+        shape = (batch_size, num_channels_latents, height, width)  # 1 16 106 80
+        noise_latents = randn_tensor(shape, generator=generator, device=device, dtype=dtype)  
+        noise_latents = self._pack_latents(noise_latents, batch_size, num_channels_latents, height, width)
+        noise_latent_image_ids, cond_latent_image_ids = resize_position_encoding(
+                batch_size,
+                height,
+                width,
+                height_cond,
+                width_cond,
+                device,
+                dtype,
+            )
+        
+        latents_to_concat = []
+        latents_ids_to_concat = [noise_latent_image_ids]
+        
+        # subject
+        if subject_image is not None:
+            shape_subject = (batch_size, num_channels_latents, height_cond*sub_number, width_cond)  
+            subject_image = subject_image.to(device=device, dtype=dtype)
+            subject_image_latents = self._encode_vae_image(image=subject_image, generator=generator)
+            subject_latents = self._pack_latents(subject_image_latents, batch_size, num_channels_latents, height_cond*sub_number, width_cond)
+            mask2 = torch.zeros(shape_subject, device=device, dtype=dtype)
+            mask2 = self._pack_latents(mask2, batch_size, num_channels_latents, height_cond*sub_number, width_cond)
+            latent_subject_ids = prepare_latent_subject_ids(height_cond, width_cond, device, dtype)
+            latent_subject_ids[:, 1] += 64  # fixed offset
+            subject_latent_image_ids = torch.concat([latent_subject_ids for _ in range(sub_number)], dim=-2)
+            latents_to_concat.append(subject_latents)
+            latents_ids_to_concat.append(subject_latent_image_ids)
+            
+        # spatial
+        if condition_image is not None:
+            shape_cond = (batch_size, num_channels_latents, height_cond*cond_number, width_cond)  
+            condition_image = condition_image.to(device=device, dtype=dtype)
+            image_latents = self._encode_vae_image(image=condition_image, generator=generator)
+            cond_latents = self._pack_latents(image_latents, batch_size, num_channels_latents, height_cond*cond_number, width_cond)
+            mask3 = torch.zeros(shape_cond, device=device, dtype=dtype)
+            mask3 = self._pack_latents(mask3, batch_size, num_channels_latents, height_cond*cond_number, width_cond) 
+            cond_latent_image_ids = cond_latent_image_ids
+            cond_latent_image_ids = torch.concat([cond_latent_image_ids for _ in range(cond_number)], dim=-2)
+            latents_ids_to_concat.append(cond_latent_image_ids)
+            latents_to_concat.append(cond_latents)
+
+        cond_latents = torch.concat(latents_to_concat, dim=-2)
+        latent_image_ids = torch.concat(latents_ids_to_concat, dim=-2)
+        return cond_latents, latent_image_ids, noise_latents
+
+    @property
+    def guidance_scale(self):
+        return self._guidance_scale
+
+    @property
+    def joint_attention_kwargs(self):
+        return self._joint_attention_kwargs
+
+    @property
+    def num_timesteps(self):
+        return self._num_timesteps
+
+    @property
+    def interrupt(self):
+        return self._interrupt
+
+    @torch.no_grad()
+    def __call__(
+            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,
+            callback_on_step_end: Optional[Callable[[int, int, Dict], None]] = None,
+            callback_on_step_end_tensor_inputs: List[str] = ["latents"],
+            max_sequence_length: int = 512,
+            spatial_images=None,
+            subject_images=None,
+            cond_size=512,
+    ):
+
+        height = height or self.default_sample_size * self.vae_scale_factor
+        width = width or self.default_sample_size * self.vae_scale_factor
+        self.cond_size = cond_size
+        
+        # 1. Check inputs. Raise error if not correct
+        self.check_inputs(
+            prompt,
+            prompt_2,
+            height,
+            width,
+            prompt_embeds=prompt_embeds,
+            pooled_prompt_embeds=pooled_prompt_embeds,
+            callback_on_step_end_tensor_inputs=callback_on_step_end_tensor_inputs,
+            max_sequence_length=max_sequence_length,
+        )
+
+        self._guidance_scale = guidance_scale
+        self._joint_attention_kwargs = joint_attention_kwargs
+        self._interrupt = False
+        
+        cond_number = len(spatial_images)
+        sub_number = len(subject_images)
+        
+        if sub_number > 0:
+            subject_image_ls = []
+            for subject_image in subject_images:
+                w, h = subject_image.size[:2]
+                scale = self.cond_size / max(h, w)
+                new_h, new_w = int(h * scale), int(w * scale)
+                subject_image = self.image_processor.preprocess(subject_image, height=new_h, width=new_w)
+                subject_image = subject_image.to(dtype=torch.float32)
+                pad_h = cond_size - subject_image.shape[-2]
+                pad_w = cond_size - subject_image.shape[-1]
+                subject_image = pad(
+                    subject_image,
+                    padding=(int(pad_w / 2), int(pad_h / 2), int(pad_w / 2), int(pad_h / 2)),
+                    fill=0
+                )
+                subject_image_ls.append(subject_image)
+            subject_image = torch.concat(subject_image_ls, dim=-2)
+        else:
+            subject_image = None
+        
+        if cond_number > 0:
+            condition_image_ls = []
+            for img in spatial_images:
+                condition_image = self.image_processor.preprocess(img, height=self.cond_size, width=self.cond_size)
+                condition_image = condition_image.to(dtype=torch.float32)
+                condition_image_ls.append(condition_image)
+            condition_image = torch.concat(condition_image_ls, dim=-2)
+        else:
+            condition_image = None
+        
+        # 2. Define call parameters
+        if prompt is not None and isinstance(prompt, str):
+            batch_size = 1
+        elif prompt is not None and isinstance(prompt, list):
+            batch_size = len(prompt)
+        else:
+            batch_size = prompt_embeds.shape[0]
+
+        device = self._execution_device
+
+        lora_scale = (
+            self.joint_attention_kwargs.get("scale", None) if self.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  # 16
+        cond_latents, latent_image_ids, noise_latents = self.prepare_latents(
+            batch_size * num_images_per_prompt,
+            num_channels_latents,
+            height,
+            width,
+            prompt_embeds.dtype,
+            device,
+            generator,
+            subject_image,
+            condition_image,
+            latents,
+            cond_number,
+            sub_number
+        )
+        latents = noise_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,
+        )
+        num_warmup_steps = max(len(timesteps) - num_inference_steps * self.scheduler.order, 0)
+        self._num_timesteps = len(timesteps)
+
+        # handle guidance
+        if self.transformer.config.guidance_embeds:
+            guidance = torch.full([1], guidance_scale, device=device, dtype=torch.float32)
+            guidance = guidance.expand(latents.shape[0])
+        else:
+            guidance = None
+        
+        ## Caching conditions
+        # clean the cache
+        for name, attn_processor in self.transformer.attn_processors.items():
+            attn_processor.bank_kv.clear()
+        # cache with warmup latents
+        start_idx = latents.shape[1] - 32
+        warmup_latents = latents[:, start_idx:, :]
+        warmup_latent_ids = latent_image_ids[start_idx:, :]
+        t = torch.tensor([timesteps[0]], device=device)
+        timestep = t.expand(warmup_latents.shape[0]).to(latents.dtype)
+        _ = self.transformer(
+                    hidden_states=warmup_latents,  
+                    cond_hidden_states=cond_latents,
+                    timestep=timestep/ 1000,
+                    guidance=guidance,
+                    pooled_projections=pooled_prompt_embeds,
+                    encoder_hidden_states=prompt_embeds,
+                    txt_ids=text_ids,
+                    img_ids=warmup_latent_ids,
+                    joint_attention_kwargs=self.joint_attention_kwargs,
+                    return_dict=False,
+                )[0]
+        
+        # 6. Denoising loop
+        with self.progress_bar(total=num_inference_steps) as progress_bar:
+            for i, t in enumerate(timesteps):
+                if self.interrupt:
+                    continue
+
+                # broadcast to batch dimension in a way that's compatible with ONNX/Core ML
+                timestep = t.expand(latents.shape[0]).to(latents.dtype)
+                noise_pred = self.transformer(
+                    hidden_states=latents,  
+                    cond_hidden_states=cond_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]
+
+                # compute the previous noisy sample x_t -> x_t-1
+                latents_dtype = latents.dtype
+                latents = self.scheduler.step(noise_pred, t, latents, return_dict=False)[0]
+                latents = latents
+
+                if latents.dtype != latents_dtype:
+                    if torch.backends.mps.is_available():
+                        # some platforms (eg. apple mps) misbehave due to a pytorch bug: https://github.com/pytorch/pytorch/pull/99272
+                        latents = latents.to(latents_dtype)
+
+                if callback_on_step_end is not None:
+                    callback_kwargs = {}
+                    for k in callback_on_step_end_tensor_inputs:
+                        callback_kwargs[k] = locals()[k]
+                    callback_outputs = callback_on_step_end(self, i, t, callback_kwargs)
+
+                    latents = callback_outputs.pop("latents", latents)
+                    prompt_embeds = callback_outputs.pop("prompt_embeds", prompt_embeds)
+
+                # call the callback, if provided
+                if i == len(timesteps) - 1 or ((i + 1) > num_warmup_steps and (i + 1) % self.scheduler.order == 0):
+                    progress_bar.update()
+
+                if XLA_AVAILABLE:
+                    xm.mark_step()
+
+        if output_type == "latent":
+            image = latents
+
+        else:
+            latents = self._unpack_latents(latents, height, width, self.vae_scale_factor)
+            latents = (latents / self.vae.config.scaling_factor) + self.vae.config.shift_factor
+            image = self.vae.decode(latents.to(dtype=self.vae.dtype), return_dict=False)[0]
+            image = self.image_processor.postprocess(image, output_type=output_type)
+
+        # Offload all models
+        self.maybe_free_model_hooks()
+
+        if not return_dict:
+            return (image,)
+
+        return FluxPipelineOutput(images=image)

src/populate.py

@@ -1,58 +0,0 @@
-import json
-import os
-
-import pandas as pd
-
-from src.display.formatting import has_no_nan_values, make_clickable_model
-from src.display.utils import AutoEvalColumn, EvalQueueColumn
-from src.leaderboard.read_evals import get_raw_eval_results
-
-
-def get_leaderboard_df(results_path: str, requests_path: str, cols: list, benchmark_cols: list) -> pd.DataFrame:
-    """Creates a dataframe from all the individual experiment results"""
-    raw_data = get_raw_eval_results(results_path, requests_path)
-    all_data_json = [v.to_dict() for v in raw_data]
-
-    df = pd.DataFrame.from_records(all_data_json)
-    df = df.sort_values(by=[AutoEvalColumn.average.name], ascending=False)
-    df = df[cols].round(decimals=2)
-
-    # filter out if any of the benchmarks have not been produced
-    df = df[has_no_nan_values(df, benchmark_cols)]
-    return df
-
-
-def get_evaluation_queue_df(save_path: str, cols: list) -> list[pd.DataFrame]:
-    """Creates the different dataframes for the evaluation queues requestes"""
-    entries = [entry for entry in os.listdir(save_path) if not entry.startswith(".")]
-    all_evals = []
-
-    for entry in entries:
-        if ".json" in entry:
-            file_path = os.path.join(save_path, entry)
-            with open(file_path) as fp:
-                data = json.load(fp)
-
-            data[EvalQueueColumn.model.name] = make_clickable_model(data["model"])
-            data[EvalQueueColumn.revision.name] = data.get("revision", "main")
-
-            all_evals.append(data)
-        elif ".md" not in entry:
-            # this is a folder
-            sub_entries = [e for e in os.listdir(f"{save_path}/{entry}") if os.path.isfile(e) and not e.startswith(".")]
-            for sub_entry in sub_entries:
-                file_path = os.path.join(save_path, entry, sub_entry)
-                with open(file_path) as fp:
-                    data = json.load(fp)
-
-                data[EvalQueueColumn.model.name] = make_clickable_model(data["model"])
-                data[EvalQueueColumn.revision.name] = data.get("revision", "main")
-                all_evals.append(data)
-
-    pending_list = [e for e in all_evals if e["status"] in ["PENDING", "RERUN"]]
-    running_list = [e for e in all_evals if e["status"] == "RUNNING"]
-    finished_list = [e for e in all_evals if e["status"].startswith("FINISHED") or e["status"] == "PENDING_NEW_EVAL"]
-    df_pending = pd.DataFrame.from_records(pending_list, columns=cols)
-    df_running = pd.DataFrame.from_records(running_list, columns=cols)
-    df_finished = pd.DataFrame.from_records(finished_list, columns=cols)
-    return df_finished[cols], df_running[cols], df_pending[cols]

src/prompt_helper.py

@@ -0,0 +1,205 @@
+import torch
+
+
+def load_text_encoders(args, class_one, class_two):
+    text_encoder_one = class_one.from_pretrained(
+        args.pretrained_model_name_or_path, subfolder="text_encoder", revision=args.revision, variant=args.variant
+    )
+    text_encoder_two = class_two.from_pretrained(
+        args.pretrained_model_name_or_path, subfolder="text_encoder_2", revision=args.revision, variant=args.variant
+    )
+    return text_encoder_one, text_encoder_two
+
+
+def tokenize_prompt(tokenizer, prompt, max_sequence_length):
+    text_inputs = tokenizer(
+        prompt,
+        padding="max_length",
+        max_length=max_sequence_length,
+        truncation=True,
+        return_length=False,
+        return_overflowing_tokens=False,
+        return_tensors="pt",
+    )
+    text_input_ids = text_inputs.input_ids
+    return text_input_ids
+
+
+def tokenize_prompt_clip(tokenizer, prompt):
+    text_inputs = tokenizer(
+        prompt,
+        padding="max_length",
+        max_length=77,
+        truncation=True,
+        return_length=False,
+        return_overflowing_tokens=False,
+        return_tensors="pt",
+    )
+    text_input_ids = text_inputs.input_ids
+    return text_input_ids
+
+
+def tokenize_prompt_t5(tokenizer, prompt):
+    text_inputs = tokenizer(
+        prompt,
+        padding="max_length",
+        max_length=512,
+        truncation=True,
+        return_length=False,
+        return_overflowing_tokens=False,
+        return_tensors="pt",
+    )
+    text_input_ids = text_inputs.input_ids
+    return text_input_ids
+
+
+def _encode_prompt_with_t5(
+        text_encoder,
+        tokenizer,
+        max_sequence_length=512,
+        prompt=None,
+        num_images_per_prompt=1,
+        device=None,
+        text_input_ids=None,
+):
+    prompt = [prompt] if isinstance(prompt, str) else prompt
+    batch_size = len(prompt)
+
+    if tokenizer is not None:
+        text_inputs = tokenizer(
+            prompt,
+            padding="max_length",
+            max_length=max_sequence_length,
+            truncation=True,
+            return_length=False,
+            return_overflowing_tokens=False,
+            return_tensors="pt",
+        )
+        text_input_ids = text_inputs.input_ids
+    else:
+        if text_input_ids is None:
+            raise ValueError("text_input_ids must be provided when the tokenizer is not specified")
+
+    prompt_embeds = text_encoder(text_input_ids.to(device))[0]
+
+    dtype = text_encoder.dtype
+    prompt_embeds = prompt_embeds.to(dtype=dtype, device=device)
+
+    _, seq_len, _ = prompt_embeds.shape
+
+    # duplicate text embeddings and attention mask for each generation per prompt, using mps friendly method
+    prompt_embeds = prompt_embeds.repeat(1, num_images_per_prompt, 1)
+    prompt_embeds = prompt_embeds.view(batch_size * num_images_per_prompt, seq_len, -1)
+
+    return prompt_embeds
+
+
+def _encode_prompt_with_clip(
+        text_encoder,
+        tokenizer,
+        prompt: str,
+        device=None,
+        text_input_ids=None,
+        num_images_per_prompt: int = 1,
+):
+    prompt = [prompt] if isinstance(prompt, str) else prompt
+    batch_size = len(prompt)
+
+    if tokenizer is not None:
+        text_inputs = tokenizer(
+            prompt,
+            padding="max_length",
+            max_length=77,
+            truncation=True,
+            return_overflowing_tokens=False,
+            return_length=False,
+            return_tensors="pt",
+        )
+
+        text_input_ids = text_inputs.input_ids
+    else:
+        if text_input_ids is None:
+            raise ValueError("text_input_ids must be provided when the tokenizer is not specified")
+
+    prompt_embeds = text_encoder(text_input_ids.to(device), output_hidden_states=False)
+
+    # Use pooled output of CLIPTextModel
+    prompt_embeds = prompt_embeds.pooler_output
+    prompt_embeds = prompt_embeds.to(dtype=text_encoder.dtype, device=device)
+
+    # duplicate text embeddings for each generation per prompt, using mps friendly method
+    prompt_embeds = prompt_embeds.repeat(1, num_images_per_prompt, 1)
+    prompt_embeds = prompt_embeds.view(batch_size * num_images_per_prompt, -1)
+
+    return prompt_embeds
+
+
+def encode_prompt(
+        text_encoders,
+        tokenizers,
+        prompt: str,
+        max_sequence_length,
+        device=None,
+        num_images_per_prompt: int = 1,
+        text_input_ids_list=None,
+):
+    prompt = [prompt] if isinstance(prompt, str) else prompt
+    dtype = text_encoders[0].dtype
+
+    pooled_prompt_embeds = _encode_prompt_with_clip(
+        text_encoder=text_encoders[0],
+        tokenizer=tokenizers[0],
+        prompt=prompt,
+        device=device if device is not None else text_encoders[0].device,
+        num_images_per_prompt=num_images_per_prompt,
+        text_input_ids=text_input_ids_list[0] if text_input_ids_list else None,
+    )
+
+    prompt_embeds = _encode_prompt_with_t5(
+        text_encoder=text_encoders[1],
+        tokenizer=tokenizers[1],
+        max_sequence_length=max_sequence_length,
+        prompt=prompt,
+        num_images_per_prompt=num_images_per_prompt,
+        device=device if device is not None else text_encoders[1].device,
+        text_input_ids=text_input_ids_list[1] if text_input_ids_list else None,
+    )
+
+    text_ids = torch.zeros(prompt_embeds.shape[1], 3).to(device=device, dtype=dtype)
+
+    return prompt_embeds, pooled_prompt_embeds, text_ids
+
+
+def encode_token_ids(text_encoders, tokens, accelerator, num_images_per_prompt=1, device=None):
+    text_encoder_clip = text_encoders[0]
+    text_encoder_t5 = text_encoders[1]
+    tokens_clip, tokens_t5 = tokens[0], tokens[1]
+    batch_size = tokens_clip.shape[0]
+
+    if device == "cpu":
+        device = "cpu"
+    else:
+        device = accelerator.device
+
+    # clip
+    prompt_embeds = text_encoder_clip(tokens_clip.to(device), output_hidden_states=False)
+    # Use pooled output of CLIPTextModel
+    prompt_embeds = prompt_embeds.pooler_output
+    prompt_embeds = prompt_embeds.to(dtype=text_encoder_clip.dtype, device=accelerator.device)
+    # duplicate text embeddings for each generation per prompt, using mps friendly method
+    prompt_embeds = prompt_embeds.repeat(1, num_images_per_prompt, 1)
+    pooled_prompt_embeds = prompt_embeds.view(batch_size * num_images_per_prompt, -1)
+    pooled_prompt_embeds = pooled_prompt_embeds.to(dtype=text_encoder_clip.dtype, device=accelerator.device)
+
+    # t5
+    prompt_embeds = text_encoder_t5(tokens_t5.to(device))[0]
+    dtype = text_encoder_t5.dtype
+    prompt_embeds = prompt_embeds.to(dtype=dtype, device=accelerator.device)
+    _, seq_len, _ = prompt_embeds.shape
+    # duplicate text embeddings and attention mask for each generation per prompt, using mps friendly method
+    prompt_embeds = prompt_embeds.repeat(1, num_images_per_prompt, 1)
+    prompt_embeds = prompt_embeds.view(batch_size * num_images_per_prompt, seq_len, -1)
+
+    text_ids = torch.zeros(prompt_embeds.shape[1], 3).to(device=accelerator.device, dtype=dtype)
+
+    return prompt_embeds, pooled_prompt_embeds, text_ids

src/submission/check_validity.py

@@ -1,99 +0,0 @@
-import json
-import os
-import re
-from collections import defaultdict
-from datetime import datetime, timedelta, timezone
-
-import huggingface_hub
-from huggingface_hub import ModelCard
-from huggingface_hub.hf_api import ModelInfo
-from transformers import AutoConfig
-from transformers.models.auto.tokenization_auto import AutoTokenizer
-
-def check_model_card(repo_id: str) -> tuple[bool, str]:
-    """Checks if the model card and license exist and have been filled"""
-    try:
-        card = ModelCard.load(repo_id)
-    except huggingface_hub.utils.EntryNotFoundError:
-        return False, "Please add a model card to your model to explain how you trained/fine-tuned it."
-
-    # Enforce license metadata
-    if card.data.license is None:
-        if not ("license_name" in card.data and "license_link" in card.data):
-            return False, (
-                "License not found. Please add a license to your model card using the `license` metadata or a"
-                " `license_name`/`license_link` pair."
-            )
-
-    # Enforce card content
-    if len(card.text) < 200:
-        return False, "Please add a description to your model card, it is too short."
-
-    return True, ""
-
-def is_model_on_hub(model_name: str, revision: str, token: str = None, trust_remote_code=False, test_tokenizer=False) -> tuple[bool, str]:
-    """Checks if the model model_name is on the hub, and whether it (and its tokenizer) can be loaded with AutoClasses."""
-    try:
-        config = AutoConfig.from_pretrained(model_name, revision=revision, trust_remote_code=trust_remote_code, token=token)
-        if test_tokenizer:
-            try:
-                tk = AutoTokenizer.from_pretrained(model_name, revision=revision, trust_remote_code=trust_remote_code, token=token)
-            except ValueError as e:
-                return (
-                    False,
-                    f"uses a tokenizer which is not in a transformers release: {e}",
-                    None
-                )
-            except Exception as e:
-                return (False, "'s tokenizer cannot be loaded. Is your tokenizer class in a stable transformers release, and correctly configured?", None)
-        return True, None, config
-
-    except ValueError:
-        return (
-            False,
-            "needs to be launched with `trust_remote_code=True`. For safety reason, we do not allow these models to be automatically submitted to the leaderboard.",
-            None
-        )
-
-    except Exception as e:
-        return False, "was not found on hub!", None
-
-
-def get_model_size(model_info: ModelInfo, precision: str):
-    """Gets the model size from the configuration, or the model name if the configuration does not contain the information."""
-    try:
-        model_size = round(model_info.safetensors["total"] / 1e9, 3)
-    except (AttributeError, TypeError):
-        return 0  # Unknown model sizes are indicated as 0, see NUMERIC_INTERVALS in app.py
-
-    size_factor = 8 if (precision == "GPTQ" or "gptq" in model_info.modelId.lower()) else 1
-    model_size = size_factor * model_size
-    return model_size
-
-def get_model_arch(model_info: ModelInfo):
-    """Gets the model architecture from the configuration"""
-    return model_info.config.get("architectures", "Unknown")
-
-def already_submitted_models(requested_models_dir: str) -> set[str]:
-    """Gather a list of already submitted models to avoid duplicates"""
-    depth = 1
-    file_names = []
-    users_to_submission_dates = defaultdict(list)
-
-    for root, _, files in os.walk(requested_models_dir):
-        current_depth = root.count(os.sep) - requested_models_dir.count(os.sep)
-        if current_depth == depth:
-            for file in files:
-                if not file.endswith(".json"):
-                    continue
-                with open(os.path.join(root, file), "r") as f:
-                    info = json.load(f)
-                    file_names.append(f"{info['model']}_{info['revision']}_{info['precision']}")
-
-                    # Select organisation
-                    if info["model"].count("/") == 0 or "submitted_time" not in info:
-                        continue
-                    organisation, _ = info["model"].split("/")
-                    users_to_submission_dates[organisation].append(info["submitted_time"])
-
-    return set(file_names), users_to_submission_dates

src/submission/submit.py

@@ -1,119 +0,0 @@
-import json
-import os
-from datetime import datetime, timezone
-
-from src.display.formatting import styled_error, styled_message, styled_warning
-from src.envs import API, EVAL_REQUESTS_PATH, TOKEN, QUEUE_REPO
-from src.submission.check_validity import (
-    already_submitted_models,
-    check_model_card,
-    get_model_size,
-    is_model_on_hub,
-)
-
-REQUESTED_MODELS = None
-USERS_TO_SUBMISSION_DATES = None
-
-def add_new_eval(
-    model: str,
-    base_model: str,
-    revision: str,
-    precision: str,
-    weight_type: str,
-    model_type: str,
-):
-    global REQUESTED_MODELS
-    global USERS_TO_SUBMISSION_DATES
-    if not REQUESTED_MODELS:
-        REQUESTED_MODELS, USERS_TO_SUBMISSION_DATES = already_submitted_models(EVAL_REQUESTS_PATH)
-
-    user_name = ""
-    model_path = model
-    if "/" in model:
-        user_name = model.split("/")[0]
-        model_path = model.split("/")[1]
-
-    precision = precision.split(" ")[0]
-    current_time = datetime.now(timezone.utc).strftime("%Y-%m-%dT%H:%M:%SZ")
-
-    if model_type is None or model_type == "":
-        return styled_error("Please select a model type.")
-
-    # Does the model actually exist?
-    if revision == "":
-        revision = "main"
-
-    # Is the model on the hub?
-    if weight_type in ["Delta", "Adapter"]:
-        base_model_on_hub, error, _ = is_model_on_hub(model_name=base_model, revision=revision, token=TOKEN, test_tokenizer=True)
-        if not base_model_on_hub:
-            return styled_error(f'Base model "{base_model}" {error}')
-
-    if not weight_type == "Adapter":
-        model_on_hub, error, _ = is_model_on_hub(model_name=model, revision=revision, token=TOKEN, test_tokenizer=True)
-        if not model_on_hub:
-            return styled_error(f'Model "{model}" {error}')
-
-    # Is the model info correctly filled?
-    try:
-        model_info = API.model_info(repo_id=model, revision=revision)
-    except Exception:
-        return styled_error("Could not get your model information. Please fill it up properly.")
-
-    model_size = get_model_size(model_info=model_info, precision=precision)
-
-    # Were the model card and license filled?
-    try:
-        license = model_info.cardData["license"]
-    except Exception:
-        return styled_error("Please select a license for your model")
-
-    modelcard_OK, error_msg = check_model_card(model)
-    if not modelcard_OK:
-        return styled_error(error_msg)
-
-    # Seems good, creating the eval
-    print("Adding new eval")
-
-    eval_entry = {
-        "model": model,
-        "base_model": base_model,
-        "revision": revision,
-        "precision": precision,
-        "weight_type": weight_type,
-        "status": "PENDING",
-        "submitted_time": current_time,
-        "model_type": model_type,
-        "likes": model_info.likes,
-        "params": model_size,
-        "license": license,
-        "private": False,
-    }
-
-    # Check for duplicate submission
-    if f"{model}_{revision}_{precision}" in REQUESTED_MODELS:
-        return styled_warning("This model has been already submitted.")
-
-    print("Creating eval file")
-    OUT_DIR = f"{EVAL_REQUESTS_PATH}/{user_name}"
-    os.makedirs(OUT_DIR, exist_ok=True)
-    out_path = f"{OUT_DIR}/{model_path}_eval_request_False_{precision}_{weight_type}.json"
-
-    with open(out_path, "w") as f:
-        f.write(json.dumps(eval_entry))
-
-    print("Uploading eval file")
-    API.upload_file(
-        path_or_fileobj=out_path,
-        path_in_repo=out_path.split("eval-queue/")[1],
-        repo_id=QUEUE_REPO,
-        repo_type="dataset",
-        commit_message=f"Add {model} to eval queue",
-    )
-
-    # Remove the local file
-    os.remove(out_path)
-
-    return styled_message(
-        "Your request has been submitted to the evaluation queue!\nPlease wait for up to an hour for the model to show in the PENDING list."
-    )

src/transformer_flux.py

@@ -0,0 +1,583 @@
+from typing import Any, Dict, Optional, Tuple, Union
+
+import numpy as np
+import torch
+import torch.nn as nn
+import torch.nn.functional as F
+
+from diffusers.configuration_utils import ConfigMixin, register_to_config
+from diffusers.loaders import FluxTransformer2DLoadersMixin, FromOriginalModelMixin, PeftAdapterMixin
+from diffusers.models.attention import FeedForward
+from diffusers.models.attention_processor import (
+    Attention,
+    AttentionProcessor,
+    FluxAttnProcessor2_0,
+    FluxAttnProcessor2_0_NPU,
+    FusedFluxAttnProcessor2_0,
+)
+from diffusers.models.modeling_utils import ModelMixin
+from diffusers.models.normalization import AdaLayerNormContinuous, AdaLayerNormZero, AdaLayerNormZeroSingle
+from diffusers.utils import USE_PEFT_BACKEND, is_torch_version, logging, scale_lora_layers, unscale_lora_layers
+from diffusers.utils.import_utils import is_torch_npu_available
+from diffusers.utils.torch_utils import maybe_allow_in_graph
+from diffusers.models.embeddings import CombinedTimestepGuidanceTextProjEmbeddings, CombinedTimestepTextProjEmbeddings, FluxPosEmbed
+from diffusers.models.modeling_outputs import Transformer2DModelOutput
+
+logger = logging.get_logger(__name__)  # pylint: disable=invalid-name
+
+@maybe_allow_in_graph
+class FluxSingleTransformerBlock(nn.Module):
+
+    def __init__(self, dim, num_attention_heads, attention_head_dim, mlp_ratio=4.0):
+        super().__init__()
+        self.mlp_hidden_dim = int(dim * mlp_ratio)
+
+        self.norm = AdaLayerNormZeroSingle(dim)
+        self.proj_mlp = nn.Linear(dim, self.mlp_hidden_dim)
+        self.act_mlp = nn.GELU(approximate="tanh")
+        self.proj_out = nn.Linear(dim + self.mlp_hidden_dim, dim)
+
+        if is_torch_npu_available():
+            processor = FluxAttnProcessor2_0_NPU()
+        else:
+            processor = FluxAttnProcessor2_0()
+        self.attn = Attention(
+            query_dim=dim,
+            cross_attention_dim=None,
+            dim_head=attention_head_dim,
+            heads=num_attention_heads,
+            out_dim=dim,
+            bias=True,
+            processor=processor,
+            qk_norm="rms_norm",
+            eps=1e-6,
+            pre_only=True,
+        )
+
+    def forward(
+        self,
+        hidden_states: torch.Tensor,
+        cond_hidden_states: torch.Tensor,
+        temb: torch.Tensor,
+        cond_temb: torch.Tensor,
+        image_rotary_emb: Optional[Tuple[torch.Tensor, torch.Tensor]] = None,
+        joint_attention_kwargs: Optional[Dict[str, Any]] = None,
+    ) -> torch.Tensor:
+        use_cond = cond_hidden_states is not None
+        
+        residual = hidden_states
+        norm_hidden_states, gate = self.norm(hidden_states, emb=temb)
+        mlp_hidden_states = self.act_mlp(self.proj_mlp(norm_hidden_states))
+        
+        if use_cond:
+            residual_cond = cond_hidden_states
+            norm_cond_hidden_states, cond_gate = self.norm(cond_hidden_states, emb=cond_temb)
+            mlp_cond_hidden_states = self.act_mlp(self.proj_mlp(norm_cond_hidden_states))
+        
+        norm_hidden_states_concat = torch.concat([norm_hidden_states, norm_cond_hidden_states], dim=-2)
+        
+        joint_attention_kwargs = joint_attention_kwargs or {}
+        attn_output = self.attn(
+            hidden_states=norm_hidden_states_concat,
+            image_rotary_emb=image_rotary_emb,
+            use_cond=use_cond,
+            **joint_attention_kwargs,
+        )
+        if use_cond:
+            attn_output, cond_attn_output = attn_output
+
+        hidden_states = torch.cat([attn_output, mlp_hidden_states], dim=2)
+        gate = gate.unsqueeze(1)
+        hidden_states = gate * self.proj_out(hidden_states)
+        hidden_states = residual + hidden_states
+        
+        if use_cond:
+            condition_latents = torch.cat([cond_attn_output, mlp_cond_hidden_states], dim=2)
+            cond_gate = cond_gate.unsqueeze(1)
+            condition_latents = cond_gate * self.proj_out(condition_latents)
+            condition_latents = residual_cond + condition_latents
+        
+        if hidden_states.dtype == torch.float16:
+            hidden_states = hidden_states.clip(-65504, 65504)
+
+        return hidden_states, condition_latents if use_cond else None
+
+
+@maybe_allow_in_graph
+class FluxTransformerBlock(nn.Module):
+    def __init__(
+        self, dim: int, num_attention_heads: int, attention_head_dim: int, qk_norm: str = "rms_norm", eps: float = 1e-6
+    ):
+        super().__init__()
+
+        self.norm1 = AdaLayerNormZero(dim)
+
+        self.norm1_context = AdaLayerNormZero(dim)
+
+        if hasattr(F, "scaled_dot_product_attention"):
+            processor = FluxAttnProcessor2_0()
+        else:
+            raise ValueError(
+                "The current PyTorch version does not support the `scaled_dot_product_attention` function."
+            )
+        self.attn = Attention(
+            query_dim=dim,
+            cross_attention_dim=None,
+            added_kv_proj_dim=dim,
+            dim_head=attention_head_dim,
+            heads=num_attention_heads,
+            out_dim=dim,
+            context_pre_only=False,
+            bias=True,
+            processor=processor,
+            qk_norm=qk_norm,
+            eps=eps,
+        )
+
+        self.norm2 = nn.LayerNorm(dim, elementwise_affine=False, eps=1e-6)
+        self.ff = FeedForward(dim=dim, dim_out=dim, activation_fn="gelu-approximate")
+
+        self.norm2_context = nn.LayerNorm(dim, elementwise_affine=False, eps=1e-6)
+        self.ff_context = FeedForward(dim=dim, dim_out=dim, activation_fn="gelu-approximate")
+
+        # let chunk size default to None
+        self._chunk_size = None
+        self._chunk_dim = 0
+
+    def forward(
+        self,
+        hidden_states: torch.Tensor,
+        cond_hidden_states: torch.Tensor,
+        encoder_hidden_states: torch.Tensor,
+        temb: torch.Tensor,
+        cond_temb: torch.Tensor,
+        image_rotary_emb: Optional[Tuple[torch.Tensor, torch.Tensor]] = None,
+        joint_attention_kwargs: Optional[Dict[str, Any]] = None,
+    ) -> Tuple[torch.Tensor, torch.Tensor]:
+        use_cond = cond_hidden_states is not None
+        
+        norm_hidden_states, gate_msa, shift_mlp, scale_mlp, gate_mlp = self.norm1(hidden_states, emb=temb)
+        if use_cond:
+                (
+                    norm_cond_hidden_states,
+                    cond_gate_msa,
+                    cond_shift_mlp,
+                    cond_scale_mlp,
+                    cond_gate_mlp,
+                ) = self.norm1(cond_hidden_states, emb=cond_temb)
+                
+        norm_encoder_hidden_states, c_gate_msa, c_shift_mlp, c_scale_mlp, c_gate_mlp = self.norm1_context(
+            encoder_hidden_states, emb=temb
+        )
+        
+        norm_hidden_states = torch.concat([norm_hidden_states, norm_cond_hidden_states], dim=-2)
+        
+        joint_attention_kwargs = joint_attention_kwargs or {}
+        # Attention.
+        attention_outputs = self.attn(
+            hidden_states=norm_hidden_states,
+            encoder_hidden_states=norm_encoder_hidden_states,
+            image_rotary_emb=image_rotary_emb,
+            use_cond=use_cond,
+            **joint_attention_kwargs,
+        )
+
+        attn_output, context_attn_output = attention_outputs[:2]
+        cond_attn_output = attention_outputs[2] if use_cond else None
+
+        # Process attention outputs for the `hidden_states`.
+        attn_output = gate_msa.unsqueeze(1) * attn_output
+        hidden_states = hidden_states + attn_output
+
+        if use_cond:
+            cond_attn_output = cond_gate_msa.unsqueeze(1) * cond_attn_output
+            cond_hidden_states = cond_hidden_states + cond_attn_output
+
+        norm_hidden_states = self.norm2(hidden_states)
+        norm_hidden_states = norm_hidden_states * (1 + scale_mlp[:, None]) + shift_mlp[:, None]
+        
+        if use_cond:
+            norm_cond_hidden_states = self.norm2(cond_hidden_states)
+            norm_cond_hidden_states = (
+                norm_cond_hidden_states * (1 + cond_scale_mlp[:, None])
+                + cond_shift_mlp[:, None]
+            )
+
+        ff_output = self.ff(norm_hidden_states)
+        ff_output = gate_mlp.unsqueeze(1) * ff_output
+        hidden_states = hidden_states + ff_output
+        
+        if use_cond:
+            cond_ff_output = self.ff(norm_cond_hidden_states)
+            cond_ff_output = cond_gate_mlp.unsqueeze(1) * cond_ff_output
+            cond_hidden_states = cond_hidden_states + cond_ff_output
+
+        # Process attention outputs for the `encoder_hidden_states`.
+
+        context_attn_output = c_gate_msa.unsqueeze(1) * context_attn_output
+        encoder_hidden_states = encoder_hidden_states + context_attn_output
+
+        norm_encoder_hidden_states = self.norm2_context(encoder_hidden_states)
+        norm_encoder_hidden_states = norm_encoder_hidden_states * (1 + c_scale_mlp[:, None]) + c_shift_mlp[:, None]
+
+        context_ff_output = self.ff_context(norm_encoder_hidden_states)
+        encoder_hidden_states = encoder_hidden_states + c_gate_mlp.unsqueeze(1) * context_ff_output
+        if encoder_hidden_states.dtype == torch.float16:
+            encoder_hidden_states = encoder_hidden_states.clip(-65504, 65504)
+
+        return encoder_hidden_states, hidden_states, cond_hidden_states if use_cond else None
+
+
+class FluxTransformer2DModel(
+    ModelMixin, ConfigMixin, PeftAdapterMixin, FromOriginalModelMixin, FluxTransformer2DLoadersMixin
+):
+    _supports_gradient_checkpointing = True
+    _no_split_modules = ["FluxTransformerBlock", "FluxSingleTransformerBlock"]
+
+    @register_to_config
+    def __init__(
+        self,
+        patch_size: int = 1,
+        in_channels: int = 64,
+        out_channels: Optional[int] = None,
+        num_layers: int = 19,
+        num_single_layers: int = 38,
+        attention_head_dim: int = 128,
+        num_attention_heads: int = 24,
+        joint_attention_dim: int = 4096,
+        pooled_projection_dim: int = 768,
+        guidance_embeds: bool = False,
+        axes_dims_rope: Tuple[int] = (16, 56, 56),
+    ):
+        super().__init__()
+        self.out_channels = out_channels or in_channels
+        self.inner_dim = num_attention_heads * attention_head_dim
+
+        self.pos_embed = FluxPosEmbed(theta=10000, axes_dim=axes_dims_rope)
+
+        text_time_guidance_cls = (
+            CombinedTimestepGuidanceTextProjEmbeddings if guidance_embeds else CombinedTimestepTextProjEmbeddings
+        )
+        self.time_text_embed = text_time_guidance_cls(
+            embedding_dim=self.inner_dim, pooled_projection_dim=pooled_projection_dim
+        )
+
+        self.context_embedder = nn.Linear(joint_attention_dim, self.inner_dim)
+        self.x_embedder = nn.Linear(in_channels, self.inner_dim)
+
+        self.transformer_blocks = nn.ModuleList(
+            [
+                FluxTransformerBlock(
+                    dim=self.inner_dim,
+                    num_attention_heads=num_attention_heads,
+                    attention_head_dim=attention_head_dim,
+                )
+                for _ in range(num_layers)
+            ]
+        )
+
+        self.single_transformer_blocks = nn.ModuleList(
+            [
+                FluxSingleTransformerBlock(
+                    dim=self.inner_dim,
+                    num_attention_heads=num_attention_heads,
+                    attention_head_dim=attention_head_dim,
+                )
+                for _ in range(num_single_layers)
+            ]
+        )
+
+        self.norm_out = AdaLayerNormContinuous(self.inner_dim, self.inner_dim, elementwise_affine=False, eps=1e-6)
+        self.proj_out = nn.Linear(self.inner_dim, patch_size * patch_size * self.out_channels, bias=True)
+
+        self.gradient_checkpointing = False
+
+    @property
+    # Copied from diffusers.models.unets.unet_2d_condition.UNet2DConditionModel.attn_processors
+    def attn_processors(self) -> Dict[str, AttentionProcessor]:
+        r"""
+        Returns:
+            `dict` of attention processors: A dictionary containing all attention processors used in the model with
+            indexed by its weight name.
+        """
+        # set recursively
+        processors = {}
+
+        def fn_recursive_add_processors(name: str, module: torch.nn.Module, processors: Dict[str, AttentionProcessor]):
+            if hasattr(module, "get_processor"):
+                processors[f"{name}.processor"] = module.get_processor()
+
+            for sub_name, child in module.named_children():
+                fn_recursive_add_processors(f"{name}.{sub_name}", child, processors)
+
+            return processors
+
+        for name, module in self.named_children():
+            fn_recursive_add_processors(name, module, processors)
+
+        return processors
+
+    # Copied from diffusers.models.unets.unet_2d_condition.UNet2DConditionModel.set_attn_processor
+    def set_attn_processor(self, processor: Union[AttentionProcessor, Dict[str, AttentionProcessor]]):
+        r"""
+        Sets the attention processor to use to compute attention.
+
+        Parameters:
+            processor (`dict` of `AttentionProcessor` or only `AttentionProcessor`):
+                The instantiated processor class or a dictionary of processor classes that will be set as the processor
+                for **all** `Attention` layers.
+
+                If `processor` is a dict, the key needs to define the path to the corresponding cross attention
+                processor. This is strongly recommended when setting trainable attention processors.
+
+        """
+        count = len(self.attn_processors.keys())
+
+        if isinstance(processor, dict) and len(processor) != count:
+            raise ValueError(
+                f"A dict of processors was passed, but the number of processors {len(processor)} does not match the"
+                f" number of attention layers: {count}. Please make sure to pass {count} processor classes."
+            )
+
+        def fn_recursive_attn_processor(name: str, module: torch.nn.Module, processor):
+            if hasattr(module, "set_processor"):
+                if not isinstance(processor, dict):
+                    module.set_processor(processor)
+                else:
+                    module.set_processor(processor.pop(f"{name}.processor"))
+
+            for sub_name, child in module.named_children():
+                fn_recursive_attn_processor(f"{name}.{sub_name}", child, processor)
+
+        for name, module in self.named_children():
+            fn_recursive_attn_processor(name, module, processor)
+
+    # Copied from diffusers.models.unets.unet_2d_condition.UNet2DConditionModel.fuse_qkv_projections with FusedAttnProcessor2_0->FusedFluxAttnProcessor2_0
+    def fuse_qkv_projections(self):
+        """
+        Enables fused QKV projections. For self-attention modules, all projection matrices (i.e., query, key, value)
+        are fused. For cross-attention modules, key and value projection matrices are fused.
+
+        <Tip warning={true}>
+
+        This API is 🧪 experimental.
+
+        </Tip>
+        """
+        self.original_attn_processors = None
+
+        for _, attn_processor in self.attn_processors.items():
+            if "Added" in str(attn_processor.__class__.__name__):
+                raise ValueError("`fuse_qkv_projections()` is not supported for models having added KV projections.")
+
+        self.original_attn_processors = self.attn_processors
+
+        for module in self.modules():
+            if isinstance(module, Attention):
+                module.fuse_projections(fuse=True)
+
+        self.set_attn_processor(FusedFluxAttnProcessor2_0())
+
+    # Copied from diffusers.models.unets.unet_2d_condition.UNet2DConditionModel.unfuse_qkv_projections
+    def unfuse_qkv_projections(self):
+        """Disables the fused QKV projection if enabled.
+
+        <Tip warning={true}>
+
+        This API is 🧪 experimental.
+
+        </Tip>
+
+        """
+        if self.original_attn_processors is not None:
+            self.set_attn_processor(self.original_attn_processors)
+
+    def _set_gradient_checkpointing(self, module, value=False):
+        if hasattr(module, "gradient_checkpointing"):
+            module.gradient_checkpointing = value
+
+    def forward(
+        self,
+        hidden_states: torch.Tensor,
+        cond_hidden_states: torch.Tensor = None,
+        encoder_hidden_states: torch.Tensor = None,
+        pooled_projections: torch.Tensor = None,
+        timestep: torch.LongTensor = None,
+        img_ids: torch.Tensor = None,
+        txt_ids: torch.Tensor = None,
+        guidance: torch.Tensor = None,
+        joint_attention_kwargs: Optional[Dict[str, Any]] = None,
+        controlnet_block_samples=None,
+        controlnet_single_block_samples=None,
+        return_dict: bool = True,
+        controlnet_blocks_repeat: bool = False,
+    ) -> Union[torch.Tensor, Transformer2DModelOutput]:
+        if cond_hidden_states is not None:
+            use_condition = True
+        else:
+            use_condition = False
+        
+        if joint_attention_kwargs is not None:
+            joint_attention_kwargs = joint_attention_kwargs.copy()
+            lora_scale = joint_attention_kwargs.pop("scale", 1.0)
+        else:
+            lora_scale = 1.0
+
+        if USE_PEFT_BACKEND:
+            # weight the lora layers by setting `lora_scale` for each PEFT layer
+            scale_lora_layers(self, lora_scale)
+        else:
+            if joint_attention_kwargs is not None and joint_attention_kwargs.get("scale", None) is not None:
+                logger.warning(
+                    "Passing `scale` via `joint_attention_kwargs` when not using the PEFT backend is ineffective."
+                )
+
+        hidden_states = self.x_embedder(hidden_states)
+        cond_hidden_states = self.x_embedder(cond_hidden_states)
+
+        timestep = timestep.to(hidden_states.dtype) * 1000
+        if guidance is not None:
+            guidance = guidance.to(hidden_states.dtype) * 1000
+        else:
+            guidance = None
+
+        temb = (
+            self.time_text_embed(timestep, pooled_projections)
+            if guidance is None
+            else self.time_text_embed(timestep, guidance, pooled_projections)
+        )
+        
+        cond_temb = (
+            self.time_text_embed(torch.ones_like(timestep) * 0, pooled_projections)
+                if guidance is None
+                else self.time_text_embed(
+                    torch.ones_like(timestep) * 0, guidance, pooled_projections
+                )
+            )
+        
+        encoder_hidden_states = self.context_embedder(encoder_hidden_states)
+
+        if txt_ids.ndim == 3:
+            logger.warning(
+                "Passing `txt_ids` 3d torch.Tensor is deprecated."
+                "Please remove the batch dimension and pass it as a 2d torch Tensor"
+            )
+            txt_ids = txt_ids[0]
+        if img_ids.ndim == 3:
+            logger.warning(
+                "Passing `img_ids` 3d torch.Tensor is deprecated."
+                "Please remove the batch dimension and pass it as a 2d torch Tensor"
+            )
+            img_ids = img_ids[0]
+
+        ids = torch.cat((txt_ids, img_ids), dim=0)
+        image_rotary_emb = self.pos_embed(ids)
+
+        if joint_attention_kwargs is not None and "ip_adapter_image_embeds" in joint_attention_kwargs:
+            ip_adapter_image_embeds = joint_attention_kwargs.pop("ip_adapter_image_embeds")
+            ip_hidden_states = self.encoder_hid_proj(ip_adapter_image_embeds)
+            joint_attention_kwargs.update({"ip_hidden_states": ip_hidden_states})
+
+        for index_block, block in enumerate(self.transformer_blocks):
+            if torch.is_grad_enabled() and self.gradient_checkpointing:
+
+                def create_custom_forward(module, return_dict=None):
+                    def custom_forward(*inputs):
+                        if return_dict is not None:
+                            return module(*inputs, return_dict=return_dict)
+                        else:
+                            return module(*inputs)
+
+                    return custom_forward
+
+                ckpt_kwargs: Dict[str, Any] = {"use_reentrant": False} if is_torch_version(">=", "1.11.0") else {}
+                encoder_hidden_states, hidden_states = torch.utils.checkpoint.checkpoint(
+                    create_custom_forward(block),
+                    hidden_states,
+                    encoder_hidden_states,
+                    temb,
+                    image_rotary_emb,
+                    cond_temb=cond_temb if use_condition else None,
+                    cond_hidden_states=cond_hidden_states if use_condition else None,
+                    **ckpt_kwargs,
+                )
+
+            else:
+                encoder_hidden_states, hidden_states, cond_hidden_states = block(
+                    hidden_states=hidden_states,
+                    encoder_hidden_states=encoder_hidden_states,
+                    cond_hidden_states=cond_hidden_states if use_condition else None,
+                    temb=temb,
+                    cond_temb=cond_temb if use_condition else None,
+                    image_rotary_emb=image_rotary_emb,
+                    joint_attention_kwargs=joint_attention_kwargs,
+                )
+
+            # controlnet residual
+            if controlnet_block_samples is not None:
+                interval_control = len(self.transformer_blocks) / len(controlnet_block_samples)
+                interval_control = int(np.ceil(interval_control))
+                # For Xlabs ControlNet.
+                if controlnet_blocks_repeat:
+                    hidden_states = (
+                        hidden_states + controlnet_block_samples[index_block % len(controlnet_block_samples)]
+                    )
+                else:
+                    hidden_states = hidden_states + controlnet_block_samples[index_block // interval_control]
+        hidden_states = torch.cat([encoder_hidden_states, hidden_states], dim=1)
+
+        for index_block, block in enumerate(self.single_transformer_blocks):
+            if torch.is_grad_enabled() and self.gradient_checkpointing:
+
+                def create_custom_forward(module, return_dict=None):
+                    def custom_forward(*inputs):
+                        if return_dict is not None:
+                            return module(*inputs, return_dict=return_dict)
+                        else:
+                            return module(*inputs)
+
+                    return custom_forward
+
+                ckpt_kwargs: Dict[str, Any] = {"use_reentrant": False} if is_torch_version(">=", "1.11.0") else {}
+                hidden_states, cond_hidden_states = torch.utils.checkpoint.checkpoint(
+                    create_custom_forward(block),
+                    hidden_states,
+                    temb,
+                    image_rotary_emb,
+                    cond_temb=cond_temb if use_condition else None,
+                    cond_hidden_states=cond_hidden_states if use_condition else None,
+                    **ckpt_kwargs,
+                )
+
+            else:
+                hidden_states, cond_hidden_states = block(
+                    hidden_states=hidden_states,
+                    cond_hidden_states=cond_hidden_states if use_condition else None,
+                    temb=temb,
+                    cond_temb=cond_temb if use_condition else None,
+                    image_rotary_emb=image_rotary_emb,
+                    joint_attention_kwargs=joint_attention_kwargs,
+                )
+
+            # controlnet residual
+            if controlnet_single_block_samples is not None:
+                interval_control = len(self.single_transformer_blocks) / len(controlnet_single_block_samples)
+                interval_control = int(np.ceil(interval_control))
+                hidden_states[:, encoder_hidden_states.shape[1] :, ...] = (
+                    hidden_states[:, encoder_hidden_states.shape[1] :, ...]
+                    + controlnet_single_block_samples[index_block // interval_control]
+                )
+
+        hidden_states = hidden_states[:, encoder_hidden_states.shape[1] :, ...]
+
+        hidden_states = self.norm_out(hidden_states, temb)
+        output = self.proj_out(hidden_states)
+
+        if USE_PEFT_BACKEND:
+            # remove `lora_scale` from each PEFT layer
+            unscale_lora_layers(self, lora_scale)
+
+        if not return_dict:
+            return (output,)
+
+        return Transformer2DModelOutput(sample=output)