RuntimeError: CUDA error: an illegal memory access was encountered

Search before asking

• I have searched the Roboflow Notebooks issues and found no similar bug report.

Notebook name

how-to-auto-train-yolov8-model-with-autodistill.ipynb

Bug

Train target model - YOLOv8:
Default epochs=50, got error and tried again with epochs=2 and got same error.

New https://pypi.org/project/ultralytics/8.0.189 available 😃 Update with 'pip install -U ultralytics'
Ultralytics YOLOv8.0.81 🚀 Python-3.10.12 torch-2.0.1+cu118 CUDA:0 (Tesla V100-SXM2-16GB, 16151MiB)
/content
---------------------------------------------------------------------------
RuntimeError                              Traceback (most recent call last)
[<ipython-input-22-8c0827b8627e>](https://localhost:8080/#) in <cell line: 6>()
      4 
      5 target_model = YOLOv8("yolov8n.pt")
----> 6 target_model.train(DATA_YAML_PATH, epochs=2)

7 frames
[/usr/local/lib/python3.10/dist-packages/torch/cuda/random.py](https://localhost:8080/#) in cb()
    109         for i in range(device_count()):
    110             default_generator = torch.cuda.default_generators[i]
--> 111             default_generator.manual_seed(seed)
    112 
    113     _lazy_call(cb, seed_all=True)

RuntimeError: CUDA error: an illegal memory access was encountered
CUDA kernel errors might be asynchronously reported at some other API call, so the stacktrace below might be incorrect.
For debugging consider passing CUDA_LAUNCH_BLOCKING=1.
Compile with `TORCH_USE_CUDA_DSA` to enable device-side assertions.

Environment

• google colab V100 51GB RAM

Minimal Reproducible Example

No response

Additional

No response

Are you willing to submit a PR?

• Yes I'd like to help by submitting a PR!

Start with notebooks/how-to-auto-train-yolov8-model-with-autodistill.ipynb and the cell calling target_model.train(DATA_YAML_PATH, epochs=2). Reproduce the traceback in the reported Google Colab V100 environment, then identify the failing training configuration or dependency interaction. Done means the notebook trains without the illegal memory access or documents a confirmed limitation and workaround.