Mitchell Mosure

 I am able to reproduce when using VSCode `ritwickdey.LiveServer` extension to serve the above SVG.

> @mosure use fly.io? currently have new_media working remotely on a fly.io gpu. wondering if you'll be able to test a similar PR here. I haven't worked with fly.io, it...

see: https://lightgaussian.github.io/

@MiniMinerX I appreciate the help! the gaussian splat forward rasterization requires per-view (camera) depth sort. in the current implementation, the sort buffers are instantiated once per gaussian cloud entity. this...

related issue: https://github.com/mosure/bevy_gaussian_splatting/issues/47

this is encountered on main. here is a more minimal reproducible example which clarifies the `/Concat_output_0` initializer/output ambiguity:  python ```python #!/usr/bin/env python3 # used to generate model: onnx-tests/tests/resize/resize.onnx import...

the same scope import error happens with constant input tensors too, e.g.  python ```python #!/usr/bin/env python3 # used to generate model: onnx-tests/tests/resize/resize.onnx import onnx from onnx import helper, TensorProto,...

a minimal `gather` import with `Initializer` indices also has a scope panic. `gather` imports correctly with `Constant` indices. I don't think `gather` has the same empty input names as `resize`:...

adding this line to `consume` seems to fix the initializer import: `self.inputs.extend(self.initializers.values().cloned());` https://github.com/tracel-ai/burn/blob/71bd5efbfaf220b5058dbf4f963326e0b0d38527/crates/burn-import/src/onnx/from_onnx.rs#L163 --- [`OnnxGraph::inputs`](https://github.com/tracel-ai/burn/blob/71bd5efbfaf220b5058dbf4f963326e0b0d38527/crates/burn-import/src/onnx/ir.rs#L181) doesn't capture the initializers. this needs to contain the initializer tensor names when passed to...

if I convert the initializer to a constant and disable constant lifting, I run into an error where an `Int` constant tensor is added to the imported model, however, all...