OCANNL: OCaml Compiles Algorithms for Neural Networks Learning
BSD 2-Clause "Simplified" License
69
stars
2
forks
source link
The use of `shared_merge_buffer` should depend on the tensor node's `memory_mode` (on all backends) #290
Open
lukstafi opened 3 weeks ago
Currently this is very broken and confused, e.g. the cuda backend always uses it when merging by-copy.