search

NVIDIA-Merlin / Merlin

NVIDIA Merlin is an open source library providing end-to-end GPU-accelerated recommender systems, from feature engineering and preprocessing to training deep learning models and running inference in production.
Apache License 2.0
759 stars 113 forks source link

[QST] Adding Google Cloud SDK to Merlin-Pytorch:22.12 #810

Closed ldane closed 1 year ago

ldane commented 1 year ago

❓ Questions & Help

Incompatible python libraries while building Vertex AI container.

Details

We're following this repository for creating our development environment for Merlin-Pytorch:22.12. While creating the container image, we are getting these messages about missing dependencies or incompatible versions.

Is this expected result?

fastai 2.7.10 requires spacy<4, which is not installed. fastai 2.7.10 requires torchvision>=0.8.2, which is not installed. dask-cudf 22.8.0a0+304.g6ca81bbc78.dirty requires cupy-cuda118<12,>=9.5.0, which is not installed. cudf 22.8.0a0+304.g6ca81bbc78.dirty requires cupy-cuda118<12,>=9.5.0, which is not installed. merlin-core 0.10.0 requires fsspec==2022.5.0, but you have fsspec 2023.1.0 which is incompatible. jupyterlab-server 2.17.0 requires jsonschema>=4.17.3, but you have jsonschema 3.2.0 which is incompatible. jupyter-events 0.5.0 requires jsonschema[format-nongpl]>=4.3.0, but you have jsonschema 3.2.0 which is incompatible. grpcio-channelz 1.51.1 requires protobuf>=4.21.6, but you have protobuf 3.20.3 which is incompatible. distributed 2022.7.1 requires tornado<6.2,>=6.0.3, but you have tornado 6.2 which is incompatible. cudf 22.8.0a0+304.g6ca81bbc78.dirty requires cuda-python<11.7.1,>=11.5, but you have cuda-python 11.8.1 which is incompatible.

rnyak commented 1 year ago

@ldane these warnings should be fine. are you able to do the following? 

import nvtabular
import cudf, dask_cudf
import torch
ldane commented 1 year ago

@rnyak Yes, I can import them. I wanted to have this messages visible in case they are related to the issue in this ticket