Broken Docker example

[sergree]

Hello! Thank you for your awesome work.

I tried to install your Docker example and have this error:

22.70 ERROR: Could not find a version that satisfies the requirement megatron-lm==2.2.0 (from nemo-toolkit==1.0.2) (from versions: none)
22.70 ERROR: No matching distribution found for megatron-lm==2.2.0 (from nemo-toolkit==1.0.2)
------
Dockerfile:13
--------------------
  11 |
  12 |     WORKDIR /workspace/NeMo
  13 | >>> RUN ./reinstall.sh
  14 |     RUN conda install -c anaconda swig
  15 |     WORKDIR /workspace/NeMo/scripts/asr_language_modeling/ngram_lm
--------------------

Thank you in advance.

[IuliiaZaitova]

I am getting the same error! Have you managed to resolve it?

[mutiev]

How to Start Working on This Issue:

1. Comment Out Lines in Dockerfile:
   Comment out the lines in the Dockerfile starting from and including RUN ./reinstall.sh down to but excluding CMD ["bash"] (keep this line).

2. Launch the Container:
   After commenting out the lines, launch the container.

3. Manually Handle Dependencies:
   Manually work with the dependencies in requirements/requirements_nlp.txt.

4. Address the Main Issue:
   The primary issue is that megatron-lm has been dropped from PyPI.

My Approach:

1. Commented Out Installation Line:
   I commented out the line for installing megatron-lm in requirements_nlp.txt.

2. Added Manual Installation:
   I added the installation of the package from the GitHub repository after installing requirements_nlp.txt in reinstall.sh:

   pip install git+https://github.com/NVIDIA/Megatron-LM

3. Testing
   In progress. AFK