Thiago Crepaldi

@liqunfu FYI

https://github.com/onnx/onnx/issues/6071 is a duplicate - and more specific case for Conv-11 which has float16 but not bfloat16 support

Hi @amyeroberts, * Indeed the number of models might be an issue for such conversion, but if that is something that huggingface servers can handle, Backward Compatibility wouldn't be a...

> @thiagocrepaldi > > > Indeed the number of models might be an issue for such conversion, but if that is something that huggingface servers can handle, Backward Compatibility wouldn't...

Thank you, I will try proposing individual PRs. It would be nice to establish minimal Pytorch versions for the newer models to guarantee performance and compatibility with newer features provided...

> Hi @thiagocrepaldi, thanks for raising this issue! > > I'm going to cc in @Narsil, the king of safetensors here. > > If you want to be able to...

This is also tracked by https://github.com/pytorch/pytorch/issues/106732 on PyTorch PyTorch's `torch.load` fixes `FakeTensorMode` support with #119990, but the issue above will still repro. Note that the fix for PyTorch side also...

`get_tensor` is a [Rust implementation](https://github.com/huggingface/safetensors/blob/08db34094e9e59e2f9218f2df133b7b4aaff5a99/bindings/python/src/lib.rs#L510) which instantiates a `Storage` of their own ```rust pub fn get_tensor(&self, name: &str) -> PyResult { let info = self.metadata.info(name).ok_or_else(|| { SafetensorError::new_err(format!("File does not contain...

> Hi @thiagocrepaldi, thanks for raising this issue! > > I'm going to cc in @Narsil, the king of safetensors here. > > If you want to be able to...

> Hi @thiagocrepaldi, thanks for raising this issue! > > I'm going to cc in @Narsil, the king of safetensors here. > > If you want to be able to...