nick/comfyui-serverless
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

Disable torch.compile to fix RTX 5090 CUDA error
All checks were successful
Build and Push Docker Image / build (push) Successful in 31m16s
Details

Browse Source 
The inductor backend was causing cudaErrorInvalidValue during Triton
kernel autotuning on Blackwell architecture. Setting backend to empty
string disables torch.compile entirely.

🤖 Generated with [Claude Code](https://claude.com/claude-code)

Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
This commit is contained in:
Debian
2026-01-06 00:22:52 +00:00
parent 2b64ac96d2
commit ab5b853521
3 changed files with 3 additions and 3 deletions
Download Patch File Download Diff File Expand all files Collapse all files

2
i2v-workflow.json
View File

@@ -36,7 +36,7 @@
"Node name for S&R": "WanVideoTorchCompileSettings" "Node name for S&R": "WanVideoTorchCompileSettings"
}, },
"widgets_values": [ "widgets_values": [
"inductor", "",
false, false,
"default", "default",
false, false,

2
workflows/Wan22-I2V-Remix-API.json
View File

@@ -47,7 +47,7 @@
}, },
"127": { "127": {
"inputs": { "inputs": {
"backend": "inductor", "backend": "",
"fullgraph": false, "fullgraph": false,
"mode": "default", "mode": "default",
"dynamic": false, "dynamic": false,

2
workflows/Wan22-I2V-Remix.json
View File

@@ -36,7 +36,7 @@
"Node name for S&R": "WanVideoTorchCompileSettings" "Node name for S&R": "WanVideoTorchCompileSettings"
}, },
"widgets_values": [ "widgets_values": [
"inductor", "",
false, false,
"default", "default",
false, false,