feat: latest torch/comfyui; perf improvments; fix: SSL cert issues

.pre-commit-config.yaml

@@ -38,9 +38,9 @@ repos:
         - python-dotenv
         - aiohttp
         - horde_safety==0.2.3
-        - torch==2.3.1
+        - torch==2.4.1
         - ruamel.yaml
-        - horde_engine==2.15.3
-        - horde_sdk==0.14.11
-        - horde_model_reference==0.9.0
+        - horde_engine==2.16.0
+        - horde_sdk==0.15.1
+        - horde_model_reference==0.9.1
         - semver

Dockerfiles/Dockerfile.12.1.1-22.04

@@ -8,20 +8,21 @@ RUN apt-get update && \
     apt-get install -y --no-install-recommends \
     python3.11 \
     python3-pip \
-    python3-venv \
+    python3.11-dev \
+    python3.11-venv \
+    python3.11-distutils \
+    libgl1 \
     git
 
 RUN git clone https://github.com/Haidra-Org/horde-worker-reGen.git && \
     cd /horde-worker-reGen && \
-    python3.11 -m pip install --upgrade pip && \
-    python3 -m venv venv && \
+    python3.11 -m venv venv && \
     . venv/bin/activate && \
     python -m pip install --upgrade pip && \
-    python -m pip install -r /horde-worker-reGen/requirements.txt -U --extra-index-url https://download.pytorch.org/whl/cu121 && \
+    python -m pip install -r /horde-worker-reGen/requirements.txt -U --extra-index-url https://download.pytorch.org/whl/cu124 && \
+    python -m pip install opencv-python-headless -U && \
     python -m pip cache purge
 
-RUN apt-get update && apt-get install libgl1 -y
-
 CMD cd /horde-worker-reGen && \
     git pull && \
     . venv/bin/activate && \

Dockerfiles/Dockerfile.12.2.2-22.04

@@ -8,20 +8,21 @@ RUN apt-get update && \
     apt-get install -y --no-install-recommends \
     python3.11 \
     python3-pip \
-    python3-venv \
+    python3.11-dev \
+    python3.11-venv \
+    python3.11-distutils \
+    libgl1 \
     git
 
 RUN git clone https://github.com/Haidra-Org/horde-worker-reGen.git && \
     cd /horde-worker-reGen && \
-    python3.11 -m pip install --upgrade pip && \
-    python3 -m venv venv && \
+    python3.11 -m venv venv && \
     . venv/bin/activate && \
     python -m pip install --upgrade pip && \
-    python -m pip install -r /horde-worker-reGen/requirements.txt -U --extra-index-url https://download.pytorch.org/whl/cu121 && \
+    python -m pip install -r /horde-worker-reGen/requirements.txt -U --extra-index-url https://download.pytorch.org/whl/cu124 && \
+    python -m pip install opencv-python-headless -U && \
     python -m pip cache purge
 
-RUN apt-get update && apt-get install libgl1 -y
-
 CMD cd /horde-worker-reGen && \
     git pull && \
     . venv/bin/activate && \

Dockerfiles/Dockerfile.12.3.2-22.04

@@ -8,20 +8,21 @@ RUN apt-get update && \
     apt-get install -y --no-install-recommends \
     python3.11 \
     python3-pip \
-    python3-venv \
+    python3.11-dev \
+    python3.11-venv \
+    python3.11-distutils \
+    libgl1 \
     git
 
 RUN git clone https://github.com/Haidra-Org/horde-worker-reGen.git && \
     cd /horde-worker-reGen && \
-    python3.11 -m pip install --upgrade pip && \
-    python3 -m venv venv && \
+    python3.11 -m venv venv && \
     . venv/bin/activate && \
     python -m pip install --upgrade pip && \
-    python -m pip install -r /horde-worker-reGen/requirements.txt -U --extra-index-url https://download.pytorch.org/whl/cu121 && \
+    python -m pip install -r /horde-worker-reGen/requirements.txt -U --extra-index-url https://download.pytorch.org/whl/cu124 && \
+    python -m pip install opencv-python-headless -U && \
     python -m pip cache purge
 
-RUN apt-get update && apt-get install libgl1 -y
-
 CMD cd /horde-worker-reGen && \
     git pull && \
     . venv/bin/activate && \

Dockerfiles/Dockerfile.12.4-22.04

@@ -0,0 +1,31 @@
+FROM nvidia/cuda:12.4.0-runtime-ubuntu22.04
+
+ARG DEBIAN_FRONTEND=noninteractive
+
+RUN apt-get update && \
+    apt-get install -y software-properties-common && \
+    add-apt-repository ppa:deadsnakes/ppa && \
+    apt-get install -y --no-install-recommends \
+    python3.11 \
+    python3-pip \
+    python3.11-dev \
+    python3.11-venv \
+    python3.11-distutils \
+    libgl1 \
+    git
+
+RUN git clone https://github.com/Haidra-Org/horde-worker-reGen.git && \
+    cd /horde-worker-reGen && \
+    python3.11 -m venv venv && \
+    . venv/bin/activate && \
+    python -m pip install --upgrade pip && \
+    python -m pip install -r /horde-worker-reGen/requirements.txt -U --extra-index-url https://download.pytorch.org/whl/cu124 && \
+    python -m pip install opencv-python-headless -U && \
+    python -m pip cache purge
+
+CMD cd /horde-worker-reGen && \
+    git pull && \
+    . venv/bin/activate && \
+    python -m pip install -r requirements.txt -U && \
+    python download_models.py -e && \
+    python run_worker.py -e

README_advanced.md

@@ -39,7 +39,7 @@
 ### Get worker files and install dependencies
 - `git clone https://github.com/Haidra-Org/horde-worker-reGen.git`
 - `cd .\horde-worker-reGen\`
-- `pip install -r requirements.txt --extra-index-url https://download.pytorch.org/whl/cu121`
+- `pip install -r requirements.txt --extra-index-url https://download.pytorch.org/whl/cu124`
 
 ### Run worker
 - Set your config now, copying `bridgeData_template.yaml` to `bridgeData.yaml`, being sure to set an API key and worker name at a minimum

bridgeData_template.yaml

@@ -114,6 +114,10 @@ unload_models_from_vram_often: true
 #   12GB-16GB VRAM: false
 #   8GB-10GB VRAM: true
 
+# Normally only one model will load off disk at a time. Set to true to load multiple models at once.
+# This requires a very fast disk. You will see a sharp increase in disk usage, especially with SDXL/Cascade/Flux/other large models.
+very_fast_disk_mode: false
+
 # List of words to reject if they appear in the prompt.
 blacklist: []
 

horde-bridge.cmd

@@ -5,7 +5,7 @@ cd /d %~dp0
 call runtime python -s -m pip -V
 
 call python -s -m pip uninstall hordelib
-call python -s -m pip install horde_sdk~=0.14.11 horde_model_reference~=0.9.0 horde_engine~=2.15.3 horde_safety~=0.2.3 -U
+call python -s -m pip install horde_sdk~=0.15.1 horde_model_reference~=0.9.1 horde_engine~=2.16.0 horde_safety~=0.2.3 -U
 
 if %ERRORLEVEL% NEQ 0 (
     echo "Please run update-runtime.cmd."

horde_worker_regen/__init__.py

@@ -8,7 +8,7 @@
 
 ASSETS_FOLDER_PATH = Path(__file__).parent / "assets"
 
-__version__ = "9.0.7"
+__version__ = "9.1.1"
 
 
 import pkg_resources  # noqa: E402

horde_worker_regen/_version_meta.json

@@ -1,5 +1,5 @@
 {
-    "recommended_version": "9.0.7",
+    "recommended_version": "9.1.1",
     "required_min_version": "9.0.2",
     "required_min_version_update_date": "2024-09-26",
     "required_min_version_info": {

horde_worker_regen/bridge_data/data_model.py

@@ -68,6 +68,9 @@ class reGenBridgeData(CombinedHordeBridgeData):
     moderate_performance_mode: bool = Field(default=False)
     """If you have a 3080 or better, set this to true to enable moderate performance mode."""
 
+    very_fast_disk_mode: bool = Field(default=False)
+    """If you have a very fast disk, set this to true to concurrently load more models at a time from disk."""
+
     post_process_job_overlap: bool = Field(default=False)
     """High and moderate performance modes will skip post processing if this is set to true."""
 

horde_worker_regen/load_env_vars.py

@@ -72,6 +72,11 @@ def load_env_vars_from_config() -> None:  # FIXME: there is a dynamic way to do
             )
             os.environ["AI_HORDE_URL"] = custom_horde_url
 
+    if "load_large_models" in config and os.getenv("AI_HORDE_MODEL_META_LARGE_MODELS") is None:
+        config_value = config["load_large_models"]
+        if config_value is True:
+            os.environ["AI_HORDE_MODEL_META_LARGE_MODELS"] = "1"
+
 
 if __name__ == "__main__":
     load_env_vars_from_config()

horde_worker_regen/process_management/inference_process.py

@@ -4,7 +4,6 @@
 
 import base64
 import contextlib
-import io
 import sys
 import time
 
@@ -352,13 +351,14 @@ def preload_model(
             seamless_tiling_enabled (bool): Whether or not seamless tiling is enabled.
             job_info (ImageGenerateJobPopResponse): The job to preload the model for.
         """
+        logger.debug(f"Currently active model is {self._active_model_name}. Requested model is {horde_model_name}")
+
         if self._active_model_name == horde_model_name:
             return
 
         if self._is_busy:
             logger.warning("Cannot preload model while busy")
 
-        logger.debug(f"Currently active model is {self._active_model_name}")
         logger.debug(f"Preloading model {horde_model_name}")
 
         if self._active_model_name is not None:
@@ -585,13 +585,11 @@ def send_inference_result_message(
 
         if results is not None:
             for result in results:
-                buffered_image = io.BytesIO()
-                if result.image is None:
+                if result.rawpng is None:
                     logger.critical("Result or result image is None")
                     continue
 
-                result.image.save(buffered_image, format="PNG")
-                image_base64 = base64.b64encode(buffered_image.getvalue()).decode("utf-8")
+                image_base64 = base64.b64encode(result.rawpng.getvalue()).decode("utf-8")
                 all_image_results.append(
                     HordeImageResult(
                         image_base64=image_base64,