Update to burn 0.19.0

[SzczurekYT]

No description provided.

[SzczurekYT]

Yolox crashes with this error:

thread 'main' (74362) panicked at examples/inference.rs:106:10:
called `Result::unwrap()` on an `Err` value: "Failed to load pre-trained weights.\nError: DeserializeError(\"Store error: Pickle parsing error: Invalid data in pickle file: dict key must be a string. This may indicate an unsupported PyTorch file format or corrupted file.\")"

I don't think this is a direct consequence of changes here, maybe some regression upstream?

[laggui]

Thanks for keeping this up to date!

I'll have a look at the potential regression with the pytorch import recorder

[laggui]

Thank you 🙏

Changed to 0.19.1 with the patch release that fixes the pytorch import error on YOLOX and updated the TORCH_CUDA_VERSION