Failing YOLO8 to RVC2 .blob conversion

RickDekker

Hi, while trying to convert a YOLO8 model to an RVC2 .blob (which our current pipeline is set up to use), the conversion failed with this error message:

[2025-09-23 14:45:34.170365+00:00] {pod_manager.py:409} INFO - ::group::Waiting until 600s to get the POD scheduled... [2025-09-23 14:45:34.214408+00:00] {pod_manager.py:430} INFO - Waiting 600s to get the POD running... [2025-09-23 14:45:34.284507+00:00] {pod_manager.py:389} INFO - The Pod has an Event: Successfully assigned hub-ai-prod/yolo-export-i9xvmjkm to gke-hub-prod-gke-np-airflow-cpu-stron-e78c2050-pzvz from None [2025-09-23 14:45:39.247555+00:00] {pod_manager.py:417} INFO - ::endgroup:: [2025-09-23 14:45:39.672122+00:00] {pod_manager.py:535} INFO - [base] INFO Class names: ['class_1', 'class_2'] main.py:113 [2025-09-23 14:45:40.213861+00:00] {pod_manager.py:535} INFO - [base] INFO Loading model... main.py:142 [2025-09-23 14:45:40.214206+00:00] {pod_manager.py:535} INFO - [base] mkdir -p failed for path /home/non-root/.config/matplotlib: [Errno 13] Permission denied: '/home/non-root' [2025-09-23 14:45:41.041705+00:00] {pod_manager.py:535} INFO - [base] Matplotlib created a temporary cache directory at /tmp/matplotlib-nmzh9tmu because there was an issue with the default path (/home/non-root/.config/matplotlib); it is highly recommended to set the MPLCONFIGDIR environment variable to a writable directory, in particular to speed up the import of Matplotlib and to better support multiprocessing. [2025-09-23 14:45:41.048785+00:00] {pod_manager.py:535} INFO - [base] WARNING âš ï¸ user config directory '/home/non-root/.config/Ultralytics' is not writeable, defaulting to '/tmp' or CWD.Alternatively you can define a YOLO_CONFIG_DIR environment variable for this path. [2025-09-23 14:45:41.049124+00:00] {pod_manager.py:535} INFO - [base] Creating new Ultralytics Settings v0.0.6 file âœ… [2025-09-23 14:45:41.049640+00:00] {pod_manager.py:535} INFO - [base] View Ultralytics Settings with 'yolo settings' or at '/tmp/Ultralytics/settings.json' [2025-09-23 14:45:41.946802+00:00] {pod_manager.py:535} INFO - [base] Update Settings with 'yolo settings key=value', i.e. 'yolo settings runs_dir=path/to/dir'. For help see https://docs.ultralytics.com/quickstart/#ultralytics-settings. [2025-09-23 14:45:41.947221+00:00] {pod_manager.py:535} INFO - [base] WARNING âš ï¸ shared_with_container/misc/test.pt appears to require 'dill', which is not in Ultralytics requirements. [2025-09-23 14:45:41.947723+00:00] {pod_manager.py:535} INFO - [base] AutoInstall will run now for 'dill' but this feature will be removed in the future. [2025-09-23 14:45:41.948322+00:00] {pod_manager.py:535} INFO - [base] Recommend fixes are to train a new model using the latest 'ultralytics' package or to run a command with an official Ultralytics model, i.e. 'yolo predict model=yolov8n.pt' [2025-09-23 14:45:44.700531+00:00] {pod_manager.py:535} INFO - [base] [31m[1mrequirements:[0m Ultralytics requirement ['dill'] not found, attempting AutoUpdate... [2025-09-23 14:45:44.702029+00:00] {pod_manager.py:535} INFO - [base] ERROR: Could not install packages due to an OSError: [Errno 13] Permission denied: '/home/non-root' [2025-09-23 14:45:44.769955+00:00] {pod_manager.py:535} INFO - [base] Check the permissions. [2025-09-23 14:45:44.770384+00:00] {pod_manager.py:535} INFO - [base] [notice] A new release of pip is available: 24.0 -> 25.2 [2025-09-23 14:45:49.483745+00:00] {pod_manager.py:535} INFO - [base] [notice] To update, run: pip install --upgrade pip [2025-09-23 14:45:49.484199+00:00] {pod_manager.py:535} INFO - [base] ERROR: Could not install packages due to an OSError: [Errno 13] Permission denied: '/home/non-root' [2025-09-23 14:45:49.696989+00:00] {pod_manager.py:535} INFO - [base] Check the permissions. [2025-09-23 14:45:49.697394+00:00] {pod_manager.py:535} INFO - [base] [notice] A new release of pip is available: 24.0 -> 25.2 [2025-09-23 14:45:49.813598+00:00] {pod_manager.py:535} INFO - [base] [notice] To update, run: pip install --upgrade pip [2025-09-23 14:45:49.814037+00:00] {pod_manager.py:535} INFO - [base] Retry 1/2 failed: Command 'pip install --no-cache-dir "dill" ' returned non-zero exit status 1. [2025-09-23 14:45:49.814629+00:00] {pod_manager.py:535} INFO - [base] Retry 2/2 failed: Command 'pip install --no-cache-dir "dill" ' returned non-zero exit status 1. [2025-09-23 14:45:49.821643+00:00] {pod_manager.py:535} INFO - [base] [31m[1mrequirements:[0m âŒ Command 'pip install --no-cache-dir "dill" ' returned non-zero exit status 1. [2025-09-23 14:45:51.830258+00:00] {pod_manager.py:554} INFO - [base] ERROR Error creating exporter: No module named 'dill' main.py:185 [2025-09-23 14:45:52.542903+00:00] {pod_manager.py:713} INFO - Pod yolo-export-i9xvmjkm has phase Running

These are the steps I took on LuxonisHub:

Could you please fix this LuxonisHub model conversion failure? It seems like the Docker image doing the conversions in the background misses some installations and has some permission errors. Thanks!

KlemenSkrlj

Hi @RickDekker ,
Could you tell me where you got the YOLOv8 model from and what variant specifically it is? I just tried converting YOLOv8-S from Ultralytics (from the table here) and conversion worked without issues.
Klemen

RickDekker

I can indeed run it with the models in the table. I think the issue here is related to this issue. Basically we only finetune a YOLOv8-N detection model from that table with ultralytics 8.3.111, could it be that your server is running a different version of ultralytics?

KlemenSkrlj

Oh I see, in that case it is an issue with the difference of ultralytics versions and one that our backend uses. Would you be able to share a .pt model that is crashing for you so I can use it to test out the fix? It can be random weights, as long as you observe the crash with it. You can share them through email: klemen.skrlj@luxonis.com
Thanks

RickDekker

Awesome! I'll send you an email!

jakob

Hi Rick,
the issue has been fixed, but the update will take some time before it is published to HubAI. If you’d like to proceed without waiting, you can run the conversion as follows:

Convert .pt → .onnx locally:

git clone --recursive https://github.com/luxonis/tools.git
cd tools
pip install .
tools <your_model_name>.pt --imgsz "<your_input_size>"

Once the process completes, you’ll find a .tar.xz file of your model in:
tools/shared_with_container/outputs

Convert .onnx → .blob on HubAI:
Upload the .tar.xz file to HubAI and convert it to RVC2 there

RickDekker

Thanks for the fix! This worked for me!