Joe Fioti

All of the AsBindGroup macro code is located here: https://github.com/bevyengine/bevy/blob/main/crates/bevy_render/macros/src/as_bind_group.rs

@alice-i-cecile Is there any way to get a vector of data to a shader currently without this? My manual implementation of the shader worked in 0.7 but broke in 0.8,...

@wrapperup Thanks for the response! ShaderType is not implemented for ArrayVec. I did however just use a normal rust array. This isn't optimal of course since a variable amount of...

How would the gather operation work? I was thinking turning indexes into one hots and doing matmuls is too inefficient, all but one of the embeds would be multiplied by...

@coreylowman Do you think that at some point it will make more sense to use flat arrays for tensor storage with strides and shapes or keep it in actual arrays?

I'd love to just do flat and call it a day right now, but one issue will be that we are forced to go full nightly. The size of the...

@coreylowman Have you looked into how to do this? I'm going to need to transpose for my MultiHeadAttention implementation, so I could either try to do this (though I would...

Ok, I ended up not needing transpose for MultiHeadAttention, but I implemented a forward pass of it anyway for Tensor3D: ```rust impl Transpose for Tensor3D { type Output = Tensor3D;...

@coreylowman Oh ... wow. Yeah an optimal setup would be to have a single generic permute function to take in 4 const arguments, and based on that return different shapes,...

I think this is a good idea, but I do thing the operations are moving pretty fast (like the potential cutting down on matmul functions) so I'm not sure if...