transformers memory tool

[winglian]

⚠️ Please check that this feature request hasn't been suggested before.

• [X] I searched previous Ideas in Discussions didn't find any similar feature requests.
• [X] I searched previous Issues didn't find any similar feature requests.

🔖 Feature description

https://tinkerd.net/blog/machine-learning/distributed-training/#memory-requirements-of-transformers

we should add this as a cli tool that calculates a memory estimate from the YML file.

✔️ Solution

see above

❓ Alternatives

No response

📝 Additional Context

No response

Acknowledgements

• [X] My issue title is concise, descriptive, and in title casing.
• [X] I have searched the existing issues to make sure this feature has not been requested yet.
• [X] I have provided enough information for the maintainers to understand and evaluate this request.

[KCaverly]

This would be pretty useful for me, I’d be happy to give this a go this week.

[NanoCode012]

Just fyi, HF has a space for this: https://huggingface.co/spaces/hf-accelerate/model-memory-usage