Checkpoint of Starcoder2 HF model is not imported into NeMo model

[d-schneid]

Hi,
when trying to import the Starcoder2 model from HF, it is not imported into the NeMo model.
In the log it says the model cannot be instantiated properly.
But, at the end the checkpoint is imported successfully into the specified output path.

I'm following the NeMo documentation, what do I need to change such that the NeMo model is instantiated based on the HF model?

if __name__ == '__main__':
       model = llm.Starcoder2Model(llm.Starcoder2Config3B())
       llm.import_ckpt(model=model, source='hf://bigcode/starcoder2-3b',
                       output_path=Path('./data/model/starcoder2'), overwrite=True)

[NeMo W 2025-04-02 00:11:19 nemo_logging:361] /usr/local/lib/python3.10/dist-packages/lightning/pytorch/
trainer/trainer.py:1090: `trainer.init_module` cannot fully support proper instantiation of your model with
the `MegatronStrategy` strategy. Please instantiate your model inside the`LightningModule.configure_model` hook
instead
    
[NeMo W 2025-04-02 00:11:34 megatron_strategy:329] Could not copy Trainer's 'max_steps' to LR scheduler's
'max_steps'. If you are not using an LR scheduler, this warning can safely be ignored.
[NeMo W 2025-04-02 00:11:50 nemo_logging:361] /usr/local/lib/python3.10/dist-packages/pyannote/core/notebook.py:
134: MatplotlibDeprecationWarning: The get_cmap function was deprecated in Matplotlib 3.7 and will be removed in
3.11. Use ``matplotlib.colormaps[name]`` or ``matplotlib.colormaps.get_cmap()`` or ``pyplot.get_cmap()`` instead.
      cm = get_cmap("Set1")
    
Converted Starcoder2 model to Nemo, model saved to data/model/starcoder2
✓ Checkpoint imported to data/model/starcoder2

[ashors1]

Hi, are you able to run with the imported starcoder2 model as expected? That warning message is coming from PyTorch Lightning, and I think you can safely ignore it in this case. If you observe any other issues with the model, please feel free to share. Thanks!