ONNXRuntimeError

[flsabourin]

Immich ML application is starting correctly, with a fresh cache folder, however, as soon as Immich Server triggers the ML engine, I get an error

at the beginning of the execution:

`(.venv) ➜ machine-learning git:(main) ✗ python -m immich_ml
[10/18/25 22:21:15] INFO Starting gunicorn 23.0.0
[10/18/25 22:21:15] INFO Listening at: http://[::]:3004 (43082)
[10/18/25 22:21:15] INFO Using worker: immich_ml.config.CustomUvicornWorker
[10/18/25 22:21:15] INFO Booting worker with pid: 43083

[10/18/25 22:21:16] INFO Started server process [43083]
[10/18/25 22:21:16] INFO Waiting for application startup.
[10/18/25 22:21:16] INFO Created in-memory cache with unloading after 300s of inactivity.
[10/18/25 22:21:16] INFO Initialized request thread pool with 10 threads.
[10/18/25 22:21:16] INFO Application startup complete.
[10/18/25 22:21:35] INFO Downloading detection model 'buffalo_l'. This may take a while.
README.md: 100%|██████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████| 582/582 [00:00<00:00, 3.40MB/s]
.gitattributes: 2.33kB [00:00, 22.2MB/s] | 0.00/582 [00:00<?, ?B/s]
detection/model.onnx: 100%|███████████████████████████████████████████████████████████████████████████████████████████████████████████████████████| 16.9M/16.9M [00:01<00:00, 14.7MB/s]
recognition/model.onnx: 100%|████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████| 174M/174M [00:01<00:00, 128MB/s]
Fetching 4 files: 100%|██████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████| 4/4 [00:01<00:00, 2.33it/s]
[10/18/25 22:21:37] INFO Loading detection model 'buffalo_l' to memory
[10/18/25 22:21:37] INFO Setting execution providers to ['CoreMLExecutionProvider', 'CPUExecutionProvider'], in descending order of preference
2025-10-18 22:21:37.661521 [W:onnxruntime:, coreml_execution_provider.cc:113 GetCapability] CoreMLExecutionProvider::GetCapability, number of partitions supported by CoreML: 4 number of nodes in the graph: 153 number of nodes supported by CoreML: 132
[10/18/25 22:21:38] INFO Loading recognition model 'buffalo_l' to memory
[10/18/25 22:21:38] INFO Setting execution providers to ['CoreMLExecutionProvider', 'CPUExecutionProvider'], in descending order of preference
2025-10-18 22:21:38.328678 [W:onnxruntime:, helper.cc:83 IsInputSupported] CoreML does not support input dim > 16384. Input:681, shape: {-1,25088}
[10/18/25 22:21:39] INFO Setting execution providers to ['CoreMLExecutionProvider', 'CPUExecutionProvider'], in descending order of preference
2025-10-18 22:21:39.436192 [W:onnxruntime:, helper.cc:83 IsInputSupported] CoreML does not support input dim > 16384. Input:681, shape: {-1,25088}`

and at the end of the execution trace

Fail: [ONNXRuntimeError] : 1 : FAIL : Non-zero status code returned while running 11938055035134306363_CoreML_11938055035134306363_0 node. Name:'CoreMLExecutionProvider_11938055035134306363_CoreML_11938055035134306363_0_0' Status Message: Error executing model: Unable to compute the prediction using a neural network model. It can be an invalid input data or broken/unsupported model (error code: -1).

I am running on an Mac mini M4, Sequoia 15.6.1

Any clue how to fix this?

ps: I changed port to 3004 not to mess up with my ML in Docker, using port 3003 (but stopped at the time of execution)

[smsm1]

I'm getting a similar error with my M3 MacBook Pro for the OCR job, so interested if there's a solution.

[KoenM9264]

I also am getting the same error, on my M1 Pro MacBook Pro and M1 Mac Mini. Any solutions?

[AgainstGreaterOdds]

Same here with an M4 Pro, latest Tahoe.

[shtefko]

Is this a dead project?