search

MiguelMonteiro / permutohedral_lattice

Permutohedral Lattice C++/CUDA implementation + TensorFlow Op (CPU/GPU)
83 stars 18 forks source link

about Batch_size #16

Closed KingZZW closed 5 years ago

KingZZW commented 5 years ago

Hi, Miguel,

When batch_size>1， the memory that needs to be allocated is batch_size=1 or batch_size>1.

Thanks,

MiguelMonteiro commented 5 years ago

It's batch_size = 1, it is a loop in the batch dimension and it is not vectorized