google-deepmind / google-deepmind/ai-foundations

Endless dependency resolution

Open
#39 0 comments 0 reactions 0 assignees View on GitHub
Dominant language
Jupyter Notebook
Stars
305
Forks
153
PR merge metrics
No merged PRs in 30d

Description

While installing package in Colab (e.g. for Lab 1.1), it never succeeds to resolve dependency resolution, stays frozen here (could see it after removing `%%capture`) :

```
Requirement already satisfied: keras_hub==0.26.0 in /usr/local/lib/python3.13/dist-packages (from ai_foundations==0.1.0) (0.26.0)
Collecting numpy==2.0.2 (from ai_foundations==0.1.0)
Using cached numpy-2.0.2.tar.gz (18.9 MB)
Installing build dependencies ... done
Getting requirements to build wheel ... done
Installing backend dependencies ... done
Preparing metadata (pyproject.toml) ...
```

I manage to import it and use it without dependency resolution:

```
# 1. Clean out the frozen pip state completely
!pip cache purge

# 2. Install kauldron strictly using the DeepMind-approved version
!pip install --no-deps kauldron==1.3.0

# 3. Force-install the DeepMind lab repository instantly without letting pip resolve dependencies
!pip install --no-deps "git+https://github.com/google-deepmind/ai-foundations.git@main"
```

Contributor guide

Open the contributing guide

Research direction

Start by reproducing the Lab 1.1 Colab installation with output visible instead of using %%capture, focusing on the numpy==2.0.2 resolution shown in the log. Compare the normal installation with the provided --no-deps commands for kauldron==1.3.0 and ai-foundations from main. Done means the package installs and the lab can import it without bypassing dependency resolution.

Written by the indexing model from the issue text.

Assessment

Tech stack
numpy, python
Domain
build-system
Issue type
Bug
Difficulty
3/5
Estimated time
1-2 days
Activity status
Active
Clarity
Needs clarification
Newbie friendliness
48/100

Get new issues in your inbox

A short digest of beginner-friendly GitHub issues.