Bjorn Beishline

@antimora I've been playing around with safetensors and burn-store. And from the current API, this seems to be the most memory-efficient manner of loading weights for use. Unless the underlying...

@antimora just following up on this since I can assign a developer to begin working on implementing the onnx nodes to burn safetensor converter. Would this be something you're interested...

Added support for other boards in my new fork called [elf2flash](https://github.com/BjornTheProgrammer/elf2flash) alongside some other improvements. Instructions on how to run with another board are in the README.

This would substantially help with https://github.com/tracel-ai/burn/issues/2871. I've been doing some investigations into memory usage in the past few days, and I've found that nearly all of the memory comes from...

Although not streaming, my new PR (#3615) creates the groundwork for loading in a single layer at a time, and might help contribute to solving this in the long run....

I've preordered the Raspberry Pi Pico 2 and would be willing to test it on actual hardware and write/verify some examples.

@ionspin [`elf2uf2-rs`](https://github.com/JoNil/elf2uf2-rs) has been updated to fix the HARD-FLOAT error. You can install the new version with `cargo install elf2uf2-rs`

I'm not 100% sure because I no longer have a Mac, but my fork, [elf2flash](https://github.com/BjornTheProgrammer/elf2flash), might fix this issue because it doesn't need the UF2 device to be mounted by...

I'm maintaining a new fork of this project called [`elf2flash`](https://github.com/BjornTheProgrammer/elf2flash). It fixes this issue and some others. Let me know if you encounter any issues with my project.

I'm more than willing to be a maintainer. I've already worked on a new fork called [`elf2flash`](https://github.com/BjornTheProgrammer/elf2flash) and would be willing to merge some of the key improvements I've made...