Pin ComfyUI to 0.29.0, build with CUDA 12.8, expand allowed CUDA versions to 12.8–13.3

.changeset/pin-comfyui-cuda-12-8.md

@@ -0,0 +1,5 @@
+---
+"worker-comfyui": minor
+---
+
+Pin ComfyUI to 0.29.0, build on CUDA 12.8, and pin the runtime PyTorch to 2.11.0+cu128 so the image runs on every host allowed by the Hub config. `allowedCudaVersions` in `.runpod/hub.json` moves from 12.6/12.7 to 12.8–13.0 (driver >= 570). Existing Hub endpoints keep the old release and its near-exhausted 12.6/12.7 host pool — update your endpoint to this release to pick up the new host range.

.runpod/hub.json

@@ -9,7 +9,7 @@
     "containerDiskInGb": 20,
     "gpuIds": "ADA_24",
     "gpuCount": 1,
-    "allowedCudaVersions": ["12.7", "12.6"],
+    "allowedCudaVersions": ["12.8", "12.9", "13.0"],
     "env": [
       {
         "key": "COMFY_ORG_API_KEY",

.runpod/tests_.json

@@ -131,6 +131,6 @@
         "value": "false"
       }
     ],
-    "allowedCudaVersions": ["12.7", "12.6"]
+    "allowedCudaVersions": ["12.8", "12.9", "13.0"]
   }
 }

Dockerfile

@@ -1,12 +1,12 @@
 # Build argument for base image selection
-ARG BASE_IMAGE=nvidia/cuda:12.6.3-cudnn-runtime-ubuntu24.04
+ARG BASE_IMAGE=nvidia/cuda:12.8.1-cudnn-runtime-ubuntu24.04
 
 # Stage 1: Base image with common dependencies
 FROM ${BASE_IMAGE} AS base
 
 # Build arguments for this stage with sensible defaults for standalone builds
-ARG COMFYUI_VERSION=latest
-ARG CUDA_VERSION_FOR_COMFY
+ARG COMFYUI_VERSION=0.29.0
+ARG CUDA_VERSION_FOR_COMFY=12.8
 ARG ENABLE_PYTORCH_UPGRADE=false
 ARG PYTORCH_INDEX_URL
 
@@ -48,7 +48,9 @@ RUN wget -qO- https://astral.sh/uv/install.sh | sh \
 ENV PATH="/opt/venv/bin:${PATH}"
 
 # Install comfy-cli + dependencies needed by it to install ComfyUI
-RUN uv pip install comfy-cli pip setuptools wheel
+# comfy-cli is pinned: its install/torch-index behavior decides what lands in
+# the workspace venv, so an unpinned version makes builds non-reproducible.
+RUN uv pip install comfy-cli==1.13.0 pip setuptools wheel
 
 # Install ComfyUI
 RUN if [ -n "${CUDA_VERSION_FOR_COMFY}" ]; then \
@@ -76,7 +78,17 @@ RUN if [ "$ENABLE_PYTORCH_UPGRADE" = "true" ]; then \
 # breaking API changes also crash ComfyUI at startup. Pinning them in the same
 # RUN downgrades within one layer, so the unwanted versions aren't left behind
 # bloating the image.
-RUN uv pip install -r /comfyui/requirements.txt \
+#
+# torch is installed FIRST, pinned to +cu128 builds: ComfyUI's requirements.txt
+# declares a bare `torch`, and default PyPI serves CUDA 13 builds (torch's PyPI
+# wheels depend on nvidia-*-cu13 since 2.11) that require driver >= 580. Hosts
+# allowed in .runpod/hub.json advertise CUDA 12.8/12.9 (driver 570/575), where
+# a cu13 torch fails CUDA init at startup. cu128 builds run on driver >= 570,
+# i.e. every allowed host. Installing torch first satisfies the bare `torch`
+# requirement so the PyPI pass doesn't touch it.
+RUN uv pip install torch==2.11.0 torchvision==0.26.0 torchaudio==2.11.0 \
+      --index-url https://download.pytorch.org/whl/cu128 \
+    && uv pip install -r /comfyui/requirements.txt \
     && for r in /comfyui/custom_nodes/*/requirements.txt; do \
          [ -f "$r" ] && uv pip install -r "$r" || true; \
        done \

docker-bake.hcl

@@ -11,16 +11,16 @@ variable "RELEASE_VERSION" {
 }
 
 variable "COMFYUI_VERSION" {
-  default = "latest"
+  default = "0.29.0"
 }
 
-# Global defaults for standard CUDA 12.6.3 images
+# Global defaults for standard CUDA 12.8.1 images
 variable "BASE_IMAGE" {
-  default = "nvidia/cuda:12.6.3-cudnn-runtime-ubuntu24.04"
+  default = "nvidia/cuda:12.8.1-cudnn-runtime-ubuntu24.04"
 }
 
 variable "CUDA_VERSION_FOR_COMFY" {
-  default = "12.6"
+  default = "12.8"
 }
 
 variable "ENABLE_PYTORCH_UPGRADE" {