Fred Reiss

Nice proposal! It's good to see the general approach from my blog posts getting traction in this community. A shared-memory store for PyTorch would be helpful for inference use cases...

Following up on my [previous comment](https://github.com/pytorch/pytorch/issues/64932#issuecomment-1234489550): Now that I think about it, it is actually possible to implement zero-copy loading of entire models for inference directly from a read-only memory-mapped...

I'm not seeing this on my machine as of today. Is this still a problem for you @ptitzler ?

This problem appears to be broader than just specifying `GPUOptions` for TensorFlow sessions. There is no native Go API for passing *any* options when creating a session. The user must...

I am looking into this.

Created PR #26682 with a Go API to create ConfigOptions protocol buffer messages. I made the changes as narrow in scope as I could.

+1 to implementing this feature in Arrow. In the meantime, I appreciate the workaround. Here's a version that optionally retains information about offsets of values in the original lists: ```python...

Thanks for reaching out @NicholasMcElroy! Might you have a self-contained piece of Python code that reproduces the problem you are seeing?

Sorry, I'm still having trouble reproducing this. Could you provide a stack trace so I can see which of the calls from `get_forward_walk_ops()` to `get_unique_graph()` is triggering this error?

Sorry for the delay in getting back to this. The most recent stack trace seems to indicate that there's a problem in the conversion from protocol buffers to `Node` and...