INIT

.gitignore

@@ -0,0 +1,5 @@
+input/image_out/*.png
+input/image_out/*.jpg
+input/assets/*/*
+
+*.DS_Store

Dockerfile

@@ -0,0 +1,38 @@
+################################################################
+#
+#   Project:    TTI Function
+#
+#   podman build -t tti-function .
+#   podman run -v /<path_to_folder>/transcript_in/:/app/input/ -v /<path_to_folder>/transcript_out/:/app/images/ --name tti-function_container --rm -t tti-function
+#
+################################################################
+
+FROM python:3.11-slim
+
+# set the working directory
+WORKDIR /app
+
+# Create image folder
+RUN mkdir /app/images
+RUN mkdir /app/input
+RUN mkdir /app/tasks
+
+RUN mkdir -p /root/.cache/huggingface/hub/models--RunDiffusion--Juggernaut-XL-v8/snapshots/9022a900377ce2d3303d3e6d86b09f6874e1e2a7/text_encoder/
+RUN mkdir -p /root/.cache/huggingface/hub/models--RunDiffusion--Juggernaut-XL-v8/snapshots/9022a900377ce2d3303d3e6d86b09f6874e1e2a7/text_encoder_2/
+RUN mkdir -p /root/.cache/huggingface/hub/models--RunDiffusion--Juggernaut-XL-v8/snapshots/9022a900377ce2d3303d3e6d86b09f6874e1e2a7/vae/
+RUN mkdir -p /root/.cache/huggingface/hub/models--RunDiffusion--Juggernaut-XL-v8/snapshots/9022a900377ce2d3303d3e6d86b09f6874e1e2a7/unet/
+
+# Update Packagelist
+RUN apt-get update
+
+# install dependencies
+COPY ./requirements.txt /app
+RUN pip install --trusted-host pypi.org --trusted-host pypi.python.org --trusted-host files.pythonhosted.org --no-cache-dir --upgrade -r requirements.txt
+
+# copy files to folder
+COPY ./runner.py /app
+COPY ./init.sh /app
+RUN chmod +x /app/init.sh
+COPY ./styles.json /app
+
+CMD ["bash", "init.sh"]

README.md

@@ -0,0 +1,18 @@
+# TTI-Function
+With the Text-to-Image (TTI) Function you can create images from a text prompt.
+Internally the Juggernaut-XL-v8 Modell (https://huggingface.co/RunDiffusion/Juggernaut-XL-v8) is used to generate the image.
+
+
+## Structure
+* The container has two folders attached, the input and output folder.
+The input folder holds .json-files that contain the context of the images, that should be created.
+The output folder is the location, where the generated images will be stored.
+
+
+## Setup
+Make sure [Podman](https://podman.io/docs/installation) or [Docker](https://docs.docker.com/get-docker/) is installed.
+
+```
+./build_image.sh
+podman run -v /path/to/your/input/folder/:/app/input/ -v /path/to/your/output/folder/:/app/images/ --name tti-function_container --rm -t tti-function
+```

build_image.sh

@@ -0,0 +1,19 @@
+#!/bin/bash
+
+# 319MB ???
+mkdir -p ./input/assets/vae/
+wget -nc "https://huggingface.co/RunDiffusion/Juggernaut-X-v10/resolve/main/vae/diffusion_pytorch_model.bin" -O ./input/assets/vae/diffusion_pytorch_model.bin
+
+# Large (9,6GB) Modell?!?
+mkdir -p ./input/assets/unet/
+wget -nc "https://huggingface.co/RunDiffusion/Juggernaut-X-v10/resolve/main/unet/diffusion_pytorch_model.bin" -O ./input/assets/unet/diffusion_pytorch_model.bin
+
+# 2,78GB
+mkdir -p ./input/assets/text_encoder_2/
+wget -nc "https://huggingface.co/RunDiffusion/Juggernaut-X-v10/resolve/main/text_encoder_2/pytorch_model.bin" -O ./input/assets/text_encoder_2/pytorch_model.bin
+
+# 492MB
+mkdir -p ./input/assets/text_encoder/
+wget -nc "https://huggingface.co/RunDiffusion/Juggernaut-X-v10/resolve/main/text_encoder/pytorch_model.bin" -O ./input/assets/text_encoder/pytorch_model.bin
+
+podman build -t tti-function .

init.sh

@@ -0,0 +1,16 @@
+#!/bin/bash
+env >> /etc/environment
+
+# Copy Models/Artefacts to needed locations
+cp /app/input/assets/unet/diffusion_pytorch_model.bin /root/.cache/huggingface/hub/models--RunDiffusion--Juggernaut-XL-v8/snapshots/9022a900377ce2d3303d3e6d86b09f6874e1e2a7/unet/diffusion_pytorch_model.bin
+cp /app/input/assets/vae/diffusion_pytorch_model.bin /root/.cache/huggingface/hub/models--RunDiffusion--Juggernaut-XL-v8/snapshots/9022a900377ce2d3303d3e6d86b09f6874e1e2a7/vae/diffusion_pytorch_model.bin
+
+cp /app/input/assets/text_encoder_2/pytorch_model.bin /root/.cache/huggingface/hub/models--RunDiffusion--Juggernaut-XL-v8/snapshots/9022a900377ce2d3303d3e6d86b09f6874e1e2a7/text_encoder_2/pytorch_model.bin
+cp /app/input/assets/text_encoder/pytorch_model.bin /root/.cache/huggingface/hub/models--RunDiffusion--Juggernaut-XL-v8/snapshots/9022a900377ce2d3303d3e6d86b09f6874e1e2a7/text_encoder/pytorch_model.bin
+
+# Run Application
+for file in /app/input/file_queues/*.json; do
+    cp $file /app/tasks/current_task.json
+    /usr/local/bin/python /app/runner.py
+    rm /app/tasks/current_task.json
+done

input/file_queues/sample.json

@@ -0,0 +1,6 @@
+{
+	"prompt": "Portrait of a person in business clothes, Looking at the camera, Smiling, Centered, Symmetric, Office Background, looking engaged, dynamic, casual chic",
+	"neg_prompt": "",
+	"style": "Fooocus Photograph",
+	"aspect_ratio": "square"
+}

requirements.txt

@@ -0,0 +1,4 @@
+diffusers
+transformers
+accelerate
+torch

runner.py

@@ -0,0 +1,131 @@
+import sys
+import os
+import uuid
+import json
+import logging
+import traceback
+from datetime import datetime as dt
+
+import torch
+from diffusers import DiffusionPipeline
+
+# Setup Logging
+logging.basicConfig(
+    level=logging.DEBUG,
+    # level=logging.INFO,
+    format="%(asctime)s [%(levelname)s] %(message)s",
+    handlers=[
+        logging.FileHandler("/var/log/" + str(dt.today().strftime('%Y-%m-%d')) + "_-_cron.log"),
+        logging.StreamHandler(sys.stdout)
+    ]
+)
+
+try:
+    for root, dirs, files in os.walk('/app/tasks'):
+        for file in files:
+            try:
+                with open(os.path.join(root, file)) as f:
+                    d = json.load(f)
+                    print(json.dumps(d, indent=4, sort_keys=True))
+
+                    prompt = d["prompt"]
+                    neg_prompt = d["neg_prompt"]
+                    style = d["style"]
+
+                    if "iterations" in d:
+                        iterations = int(d["iterations"])
+                    else:
+                        iterations = 40
+
+                    if "format" in d:
+                        format = d["format"]
+                    else:
+                        format = "square"
+
+                    match format.lower():
+                        case "portrait":
+                            format_width = 915
+                            format_height = 1144
+                        case "widescreen":
+                            format_width = 1344
+                            format_height = 768
+                        case "photo":  # Photo (4x3)
+                            format_width = 1182
+                            format_height = 886
+                        case "square":
+                            format_width = 1024
+                            format_height = 1024
+
+                    # For Debugging decrease image size
+                    # scale_factor = 0.70   # Error: Image shredded
+                    # scale_factor = 0.68   # Error: Image shredded
+                    scale_factor = 0.65  # Working on MacBook Pro (M3 - Nov. 2023)
+
+                    format_width = int(round(round(format_width * scale_factor) / 8) * 8)
+                    format_height = int(round(round(format_height * scale_factor) / 8) * 8)
+
+                with open('/app/styles.json') as f:
+                    for entry in json.load(f):
+                        if entry['style'] == style:
+                            prompt += ", " + entry['prompt_extension']
+                            neg_prompt += "extra fingers, mutated hands, poorly drawn hands, poorly drawn face, deformed, ugly, blurry, bad anatomy, bad proportions, extra limbs, glitchy, double torso, extra arms, extra hands, mangled fingers, missing lips, ugly face, distorted face, extra legs"
+                            neg_prompt += entry['negative_prompt']
+
+                base = DiffusionPipeline.from_pretrained(
+                    "RunDiffusion/Juggernaut-XL-v8"
+                ).to("cpu")
+                refiner = DiffusionPipeline.from_pretrained(
+                    "RunDiffusion/Juggernaut-XL-v8",
+                    text_encoder_2=base.text_encoder_2,
+                    vae=base.vae,
+                ).to("cpu")
+
+                n_steps = iterations
+                high_noise_frac = 0.8
+
+                image = base(
+                    prompt=prompt,
+                    generator=torch.Generator(device="cpu").manual_seed(torch.Generator(device="cpu").seed()),
+                    # generator=torch.Generator(device="cpu").manual_seed(420),
+                    negative_prompt=neg_prompt,
+                    num_inference_steps=n_steps,
+                    denoising_end=high_noise_frac,
+                    width=format_width,
+                    height=format_height,
+                    original_size=(format_width, format_height),
+                    target_size=(format_width, format_height),
+                    output_type="latent",
+                ).images
+                image = refiner(
+                    prompt=prompt,
+                    generator=torch.Generator(device="cpu").manual_seed(torch.Generator(device="cpu").seed()),
+                    # generator=torch.Generator(device="cpu").manual_seed(420),
+                    negative_prompt=neg_prompt,
+                    num_inference_steps=n_steps,
+                    denoising_start=high_noise_frac,
+                    width=format_width,
+                    height=format_height,
+                    original_size=(format_width, format_height),
+                    target_size=(format_width, format_height),
+                    image=image,
+                ).images[0]
+
+                img_uuid = str(uuid.uuid4())
+                file_name = '/app/images/' + img_uuid + '.png'
+                image.save(file_name)
+
+                logging.debug("Image generated and written to file: " + file_name)
+                sys.exit(0)
+
+            except Exception as e:
+                logging.debug("There was an error: " + str(e))
+                logging.debug("Stacktrace: " + str(traceback.format_exc()))
+                logging.debug("Skipping iteration!!!")
+                continue
+
+except Exception as e:
+    logging.debug("There was an error: " + str(e))
+    logging.debug("Stacktrace: " + str(traceback.format_exc()))
+
+finally:
+    exit(0)