Dependency issue with Colab Demo 2 (and also Demo 1)

[MikeB2019x]

I am getting this error during pip install:

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.
dopamine-rl 4.1.2 requires jax>=0.1.72, which is not installed.
dopamine-rl 4.1.2 requires jaxlib>=0.1.51, which is not installed.
tensorflow 2.18.0 requires tensorboard<2.19,>=2.18, which is not installed.
orbax-checkpoint 0.11.12 requires jax>=0.5.0, which is not installed.
flax 0.10.5 requires jax>=0.5.1, which is not installed.
optax 0.2.4 requires jax>=0.4.27, which is not installed.
optax 0.2.4 requires jaxlib>=0.4.27, which is not installed.
chex 0.1.89 requires jax>=0.4.27, which is not installed.
chex 0.1.89 requires jaxlib>=0.4.27, which is not installed.
google-colab 1.0.0 requires pandas==2.2.2, but you have pandas 2.2.3 which is incompatible.
tensorflow 2.18.0 requires numpy<2.1.0,>=1.26.0, but you have numpy 2.1.3 which is incompatible.
numba 0.60.0 requires numpy<2.1,>=1.22, but you have numpy 2.1.3 which is incompatible.

The requirements file in github is:

gluonts[torch]<=0.14.4
numpy>=1.23.5
torch>=2.0.0
wandb
scipy
pandas
huggingface_hub[cli]
matplotlib

Suggest maybe:

gluonts[torch]<=0.14.4
numpy<2.1.0
torch>=2.0.0
wandb
scipy
pandas==2.2.2
huggingface_hub[cli]
matplotlib
jax>=0.5.1
jaxlib>=0.4.27
tensorboard==2.18

NOTE: the numpy and pandas versions are also issues with Colab Demo 1. Same solution will apply.

[kashif]

thanks @MikeB2019x can you kindly check this PR #114 where I have updated the gluonts requirements and hopefully also fix the numpy issue

[MikeB2019x]

nvm my last. Running Colab Demo 1 gives the following error. I think you will have to set pandas and numpy explicitly to ==2.2.2 and numpy<2.1.0 respectively.

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.
google-colab 1.0.0 requires pandas==2.2.2, but you have pandas 2.2.3 which is incompatible.
tensorflow 2.18.0 requires numpy<2.1.0,>=1.26.0, but you have numpy 2.1.3 which is incompatible.
numba 0.60.0 requires numpy<2.1,>=1.22, but you have numpy 2.1.3 which is incompatible.

Colab Demo 2 errors out with the following. the missing jax/jaxlib might be b/c those are explicitly uninstalled in line #3 [!pip uninstall jax jaxlib tensorboard -y] of the notebook and are not in the requirements.txt.

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.
tensorflow 2.18.0 requires tensorboard<2.19,>=2.18, which is not installed.
dopamine-rl 4.1.2 requires jax>=0.1.72, which is not installed.
dopamine-rl 4.1.2 requires jaxlib>=0.1.51, which is not installed.
flax 0.10.5 requires jax>=0.5.1, which is not installed.
optax 0.2.4 requires jax>=0.4.27, which is not installed.
optax 0.2.4 requires jaxlib>=0.4.27, which is not installed.
chex 0.1.89 requires jax>=0.4.27, which is not installed.
chex 0.1.89 requires jaxlib>=0.4.27, which is not installed.
orbax-checkpoint 0.11.12 requires jax>=0.5.0, which is not installed.
google-colab 1.0.0 requires pandas==2.2.2, but you have pandas 2.2.3 which is incompatible.
tensorflow 2.18.0 requires numpy<2.1.0,>=1.26.0, but you have numpy 2.1.3 which is incompatible.
numba 0.60.0 requires numpy<2.1,>=1.22, but you have numpy 2.1.3 which is incompatible.

[kashif]

ok let me try with explicit versions

[kashif]

@MikeB2019x i've pinned the versions as you suggested... can you kindly test?

[MikeB2019x]

I just tried demo 1 which gives this:

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.
google-colab 1.0.0 requires pandas==2.2.2, but you have pandas 2.2.3 which is incompatible.
torchaudio 2.6.0+cu124 requires torch==2.6.0, but you have torch 2.7.0 which is incompatible.
torchvision 0.21.0+cu124 requires torch==2.6.0, but you have torch 2.7.0 which is incompatible.
fastai 2.7.19 requires torch<2.7,>=1.10, but you have torch 2.7.0 which is incompatible.

This for demo 2:

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.
dopamine-rl 4.1.2 requires jax>=0.1.72, which is not installed.
dopamine-rl 4.1.2 requires jaxlib>=0.1.51, which is not installed.
google-colab 1.0.0 requires pandas==2.2.2, but you have pandas 2.2.3 which is incompatible.
torchaudio 2.6.0+cu124 requires torch==2.6.0, but you have torch 2.7.0 which is incompatible.
torchvision 0.21.0+cu124 requires torch==2.6.0, but you have torch 2.7.0 which is incompatible.
fastai 2.7.19 requires torch<2.7,>=1.10, but you have torch 2.7.0 which is incompatible.

[kashif]

ok let me pin pandas as torch?