CUDA CCCL notebook is not defaulting to the T4 runtime in Colab

[brycelelbach]

Hi @brycelelbach - many thanks for you prompt solution!

The imports are now executing fine. Unfortunately, now the example fails because of incompatibility between CUDA driver and version. As I run the third cell:

---------------------------------------------------------------------------

CUDARuntimeError                          Traceback (most recent call last)

[/tmp/ipython-input-3076553934.py](https://localhost:8080/#) in <cell line: 0>()
      3 """
      4 
----> 5 d_input = cp.array([-2, 3, 5, 1, 7, -6, 8, -4], dtype=np.int32)
      6 d_output = cp.empty(1, dtype=np.int32)
      7 

[/usr/local/lib/python3.12/dist-packages/cupy/_creation/from_data.py](https://localhost:8080/#) in array(obj, dtype, copy, order, subok, ndmin, blocking)
     54 
     55     """
---> 56     return _core.array(obj, dtype, copy, order, subok, ndmin, blocking)
     57 
     58 

cupy/_core/core.pyx in cupy._core.core.array()

cupy/_core/core.pyx in cupy._core.core.array()

cupy/_core/core.pyx in cupy._core.core._array_default()

cupy/_core/core.pyx in cupy._core.core.ndarray.__new__()

cupy/_core/core.pyx in cupy._core.core._ndarray_base._init()

cupy/cuda/memory.pyx in cupy.cuda.memory.alloc()

cupy/cuda/memory.pyx in cupy.cuda.memory.MemoryPool.malloc()

cupy/cuda/memory.pyx in cupy.cuda.memory.MemoryPool.malloc()

cupy/cuda/memory.pxd in cupy.cuda.memory.MemoryPool.device_pool()

cupy/cuda/memory.pyx in cupy.cuda.memory.MemoryPool._ensure_pools_and_return_device_pool()

cupy_backends/cuda/api/runtime.pyx in cupy_backends.cuda.api.runtime.getDeviceCount()

cupy_backends/cuda/api/runtime.pyx in cupy_backends.cuda.api.runtime.check_status()

CUDARuntimeError: cudaErrorInsufficientDriver: CUDA driver version is insufficient for CUDA runtime version

Similar errors show up when running the next cell with cp.asarray.

Could you please reopen this issue or create a new one if appropriate? Thanks again.

Originally posted by @ynikitenko in #135

I tested and the notebook runs fine when connected to a T4 instance in Colab. When I open the notebook in Colab, it does not default to the T4 instance. I believe this still works for other notebooks. The metadata for the notebook seems to be missing the "accelerator": "GPU" line.

We need to validate notebook metadata and formats in CI.

[brycelelbach]

Will be fixed by #148