Help! I always get this error: onnxruntime CUDA error cudaErrorInvalidPtx:a PTX JIT compilation failed

[jrgane1102]

Hello,
I have a NVIDIA GeForce RTX 5070 Ti and used to be able to run reactor with no issues. However, after a recent update to comfyui, my reactor stopped working. I reinstalled multiple versions of comfyui with pytorch version: 2.7.1+cu128 and I always get this error. Any help would be appreciated

[ReActor] 09:00:17 - STATUS - Checking for any unsafe content...
[ReActor] 09:00:17 - STATUS - Working: source face index [0], target face index [0]
[ReActor] 09:00:17 - STATUS - Analyzing Source Image...
2025-10-04 09:00:19.3641915 [E:onnxruntime:, sequential_executor.cc:572 onnxruntime::ExecuteKernel] Non-zero status code returned while running Relu node. Name:'Relu_2' Status Message: CUDA error cudaErrorInvalidPtx:a PTX JIT compilation failed
[ReActor] 09:00:19 - ERROR - No faces found
[ReActor] 09:00:19 - STATUS - Trying to halve 'det_size' parameter
2025-10-04 09:00:20.7025126 [E:onnxruntime:, sequential_executor.cc:572 onnxruntime::ExecuteKernel] Non-zero status code returned while running Relu node. Name:'Relu_2' Status Message: CUDA error cudaErrorInvalidPtx:a PTX JIT compilation failed
[ReActor] 09:00:21 - ERROR - No faces found
[ReActor] 09:00:21 - STATUS - Analyzing Target Image...
2025-10-04 09:00:21.1318757 [E:onnxruntime:, sequential_executor.cc:572 onnxruntime::ExecuteKernel] Non-zero status code returned while running Relu node. Name:'Relu_2' Status Message: CUDA error cudaErrorInvalidPtx:a PTX JIT compilation failed
[ReActor] 09:00:21 - ERROR - No faces found
[ReActor] 09:00:21 - STATUS - Trying to halve 'det_size' parameter
2025-10-04 09:00:21.4184794 [E:onnxruntime:, sequential_executor.cc:572 onnxruntime::ExecuteKernel] Non-zero status code returned while running Relu node. Name:'Relu_2' Status Message: CUDA error cudaErrorInvalidPtx:a PTX JIT compilation failed
[ReActor] 09:00:21 - ERROR - No faces found
[ReActor] 09:00:21 - STATUS - Cannot detect any Target, skipping swapping...
[ReActor] 09:00:21 - STATUS - --Done!--
[ReActor] 09:00:21 - STATUS - Restoring with GPEN-BFR-512.onnx | Face Size is set to 512
FETCH ComfyRegistry Data: 65/99
2025-10-04 09:00:22.9322023 [E:onnxruntime:, sequential_executor.cc:572 onnxruntime::ExecuteKernel] Non-zero status code returned while running LeakyRelu node. Name:'/ecd0/ecd0.0/ecd0.0.1/LeakyRelu' Status Message: CUDA error cudaErrorInvalidPtx:a PTX JIT compilation failed
Failed inference: [ONNXRuntimeError] : 1 : FAIL : Non-zero status code returned while running LeakyRelu node. Name:'/ecd0/ecd0.0/ecd0.0.1/LeakyRelu' Status Message: CUDA error cudaErrorInvalidPtx:a PTX JIT compilation failed

[poisas]

:) and yesterday there was another update that force pytorch update to 2.9.1 +cu130 :D so if your env was 2.7.1 you need to reinstall everything new pytorch an cu130. Chatgpt is your best friend :)