I am using RL environments in Google Colab with the torchrl library. The following code used to work:
!pip3 install torchrl
Now, I get the following error from running that line at the beginning of the notebook:
Installing collected packages: triton, nvidia-nvtx-cu12, nvidia-nvjitlink-cu12, nvidia-nccl-cu12, nvidia-curand-cu12, nvidia-cufft-cu12, nvidia-cuda-runtime-cu12, nvidia-cuda-nvrtc-cu12, nvidia-cuda-cupti-cu12, nvidia-cublas-cu12, nvidia-cusparse-cu12, nvidia-cudnn-cu12, nvidia-cusolver-cu12, torch, tensordict, torchrl
Attempting uninstall: triton
Found existing installation: triton 2.1.0
Uninstalling triton-2.1.0:
Successfully uninstalled triton-2.1.0
Attempting uninstall: torch
Found existing installation: torch 2.1.0+cu121
Uninstalling torch-2.1.0+cu121:
Successfully uninstalled torch-2.1.0+cu121
ERROR: pip's dependency resolver does not currently take into account all the packages that are installed. This behaviour is the source of the following dependency conflicts.
torchaudio 2.1.0+cu121 requires torch==2.1.0, but you have torch 2.2.1 which is incompatible.
torchdata 0.7.0 requires torch==2.1.0, but you have torch 2.2.1 which is incompatible.
torchtext 0.16.0 requires torch==2.1.0, but you have torch 2.2.1 which is incompatible.
torchvision 0.16.0+cu121 requires torch==2.1.0, but you have torch 2.2.1 which is incompatible.
The installation also runs much more slowly. I think that the issue has something to do with the fact that torch 2.2.1 was released recently, but I am not sure how to resolve it (running !pip3 install torch==2.1.0 first does not work as it is uninstalled for some reason). Any insight is appreciated. Thanks so much!
That's odd, pip3 install torchrl works on my end.
The issues seems to come from audio/text/vision which have been installed previously
Does
!pip3 install torchrl torchvision torchaudio torchtext torchdata -U
Solves your issue? You could also consider removing a couple of these if you're not using them