Closed pschubert closed 5 years ago
Regarding "imposed_patch_size" support in InferenceModel.predict_proba()
(splitting inputs to smaller chunks ("tiles") for inference):
To prevent confusion, I think we should call this "spatial chunking" feature "tiling", like in ELEKTRONN2. Proposed definitions/usage:
knossos_utils
's usage of this term.batch_size
in predict_proba
) to anything > 1
if the tile size (imposed_patch_size
) is already equal to the chunk size and the GPU memory is still only less than half utilized.Relevant ELEKTRONN2 code: https://github.com/ELEKTRONN/ELEKTRONN2/blob/c3da2027a895b7aa2425ac567b3c90785204c8e6/elektronn2/neuromancer/node_basic.py#L950-L996.
@pschubert do you have other thoughts on this issue?
I'm merging this PR and will work on input patch tiling later, as it's more complicated than I initially thought (taking overlapping tiles and interpolation of overlapping regions into consideration).