bschifferer

@jperez999 (+ not sure who else can help). For movielens-1m notebooks, the tests expect following data structure on the CI machine: ``` /raid/data/movielens/ ├── ml-1m │ ├── README │ ├──...

@jperez999 have you had a chance to add the dataset to the CI environment? Is there something we can help with?

I do not understand this Roadmap ticket. I think the ticket should be `setup and cleanup integration tests for EXISTING Merlin example notebooks`. The ticket contains too many requests/examples, which...

Are we adding model to the ranking notebook? https://github.com/NVIDIA-Merlin/models/blob/main/examples/03-Exploring-different-models.ipynb

I think we discussed that TensorFlow 2.8 will set `cuda_malloc_async` by default. @rnyak shared, if we remove `os.environ["TF_GPU_ALLOCATOR"]="cuda_malloc_async"`, that TensorFlow2.8 will consume the full GPU memory. I reproduced the behavior...

Let's test 2.7 if `cuda_malloc_async` works

Tensorflow 2.6 behavior (21.12 container) Set nothing: 31/32GB Set cuda_malloc_async: 0.5/32GB Set TF_MEMORY_ALLOCATION=0.5: 31/32GB Use configure_tensorflow: is not availble in the docker container

Tensorflow 2.7 behavior (22.02 container) Set nothing: 31/32GB Set cuda_malloc_async: 31/32GB Set TF_MEMORY_ALLOCATION=0.5: 31/32GB Use configure_tensorflow: is not availble in the docker container

@rnyak @jperez999 I am not sure, how we should continue with the TensorFlow allocation logic :) . I do not understand, why it works for 2.6 and it does not...

Yes, we can do it. But I wonder, why cuda_malloc_async works for TF2.6 but it does not work for TF2.7?