InvalidArchiveError when installing TensorFlow 2.7.0 from sleap/label/dev channel

[boyuan99]

Bug description

Unable to install TensorFlow 2.7.0 from sleap/label/dev channel due to corrupted package archive. The installation process fails with InvalidArchiveError.

Expected behaviour

Successfully install TensorFlow 2.7.0 from the sleap/label/dev channel as part of SLEAP environment setup.

Actual behaviour

When trying to install TensorFlow 2.7.0, the download/extraction process fails with an InvalidArchiveError. Previously also encountered "package appears to be corrupted" errors indicating missing files in the package manifest.

Your personal set up

• OS: Windows 11
• Version(s): SLEAP (attempting to install), Python 3.7.12, conda 24.11.2 (miniforge3)
• SLEAP installation method:
  • [Conda from package](https://sleap.ai/installation.html#conda-package)

Environment packages

# Environment: sleap
# Channels:
#  - sleap/label/dev
#  - nvidia
#  - conda-forge
# Platform: win-64

ca-certificates    conda-forge/win-64::ca-certificates-2025.1.31-h56e8100_0
libsqlite          conda-forge/win-64::libsqlite-3.49.1-h67fdade_1
openssl            conda-forge/win-64::openssl-3.4.1-ha4e3fda_0
pip                conda-forge/noarch::pip-24.0-pyhd8ed1ab_0
python             conda-forge/win-64::python-3.7.12-h900ac77_100_cpython
setuptools         conda-forge/noarch::setuptools-69.0.3-pyhd8ed1ab_0
sqlite             conda-forge/win-64::sqlite-3.49.1-h2466b09_1
ucrt               conda-forge/win-64::ucrt-10.0.22621.0-h57928b3_1
vc                 conda-forge/win-64::vc-14.3-h5fd82a7_24
vc14_runtime       conda-forge/win-64::vc14_runtime-14.42.34433-h6356254_24
vs2015_runtime     conda-forge/win-64::vs2015_runtime-14.42.34433-hfef2bbc_24
wheel              conda-forge/noarch::wheel-0.42.0-pyhd8ed1ab_0

Logs

(sleap) > conda install -c sleap/label/dev tensorflow=2.7.0
Channels:
 - sleap/label/dev
 - nvidia
 - conda-forge
Platform: win-64
Collecting package metadata (repodata.json): done
Solving environment: done

## Package Plan ##

  environment location: [path to environment]

  added / updated specs:
    - tensorflow=2.7.0


The following packages will be downloaded:

    package                    |            build
    ---------------------------|-----------------
    hdf5-1.12.2                |nompi_h57737ce_101        11.1 MB  conda-forge
    libcurl-8.12.1             |       h88aaa65_0         342 KB  conda-forge
    libprotobuf-3.21.8         |       h12be248_0         2.4 MB  conda-forge
    protobuf-4.21.8            |   py37h7f67f24_0         692 KB  conda-forge
    tensorflow-2.7.0           |   py37h5685391_4       387.9 MB  sleap/label/dev
    ------------------------------------------------------------
                                           Total:       402.4 MB

[package list continues...]

Downloading and Extracting Packages:

InvalidArchiveError("Error with archive [path to tensorflow package]")

Previously also encountered:

The package for tensorflow located at [path to tensorflow package]
appears to be corrupted. The path 'Lib/site-packages/tensorflow/include/external/cudnn_frontend_archive/_virtual_includ'
specified in the package manifest cannot be found.
This transaction has incompatible packages due to a shared path.
  packages: anaconda/win-64::protobuf-3.20.3-py37hd77b12b_0, sleap/label/dev/win-64::tensorflow-2.7.0-py37h5685391_4
  path: 'lib/site-packages/protobuf-3.20.3-py3.7-nspkg.pth'

How to reproduce

1. First, I tried the recommended installation method:

   mamba create -y -n sleap -c conda-forge -c nvidia -c sleap/label/dev -c sleap -c anaconda sleap=1.4.1
   

   But this still had problems with the TensorFlow package.

2. Then I tried creating a new conda environment with Python 3.7:

   conda create -n sleap python=3.7
   conda activate sleap
   

3. Added the required channels:

   conda config --env --add channels conda-forge
   conda config --env --add channels nvidia
   conda config --env --add channels sleap/label/dev
   

4. Tried installing TensorFlow 2.7.0:

   conda install -c sleap/label/dev tensorflow=2.7.0
   

5. Observed the InvalidArchiveError during download/extraction

I've tried multiple approaches including:

• Using the environment.yml file directly
• Creating a clean environment and installing packages incrementally
• Cleaning conda cache with conda clean -a
• Attempting installation on different days

The TensorFlow 2.7.0 package seems to be consistently corrupted or having issues in the sleap/label/dev channel.

[gitttt-1234]

Hi @boyuan99 !

Sorry, I'm not able to replicate this error. I just tested the command once given here and it worked. There might be an issue with the conda installation. Could you try installing/ re-installing miniforge to not disturb your current conda?

Thanks,

Divya