System Info

Currently optimum-neuron pins protobuf to <4:

However, pipdeptree shows there's no longer such limits on protobuf version for torch/torch-neuronx/neuronx-cc:

├── torch-neuronx [required: Any, installed: 2.9.0.2.11.15156+43ca749.dev]
│   ├── torch [required: ==2.9.*, installed: 2.9.0+cpu]
...
│   ├── libneuronxla [required: >=2.2,<2.3, installed: 2.2.13706.0+d67bc329]
│   │   ├── neuronx-cc [required: ~=2.0, installed: 2.0.204784.0a0+39e2a6e7]
...
│   │   │   ├── protobuf [required: >=3, installed: 6.32.1]
...
│   ├── protobuf [required: >3.19, installed: 6.32.1]



torch-xla 2.9 likely will need protobuf >3.20. So let's start removing the pin.

Who can help?

@dacorvo @JingyaHuang @michaelbenayoun

Information

• The official example scripts
• My own modified scripts

Tasks

• An officially supported task in the examples folder (such as GLUE/SQuAD, ...)
• My own task or dataset (give details below)

Reproduction (minimal, reproducible, runnable)

Run https://awsdocs-neuron.readthedocs-hosted.com/en/latest/src/examples/pytorch/torch-neuronx/t5-inference-tutorial.html with torch-xla 2.9.

Expected behavior

No crash due to protobuf conflict.