fast_convert module

Full Documentation for hippynn.custom_kernels.fast_convert module. Click here for a summary page.

This module implements a version of converting from pytorch tensors to numba DeviceNDArrays that skips much of the indirection that takes place in the numba implementation.

Note: This is not entirely API safe as numba has not exposed all of these functions directly.

batch_convert_torch_to_numba(*tensors, typedict={torch.complex128: dtype('complex128'), torch.complex64: dtype('complex64'), torch.float16: dtype('float16'), torch.float32: dtype('float32'), torch.float64: dtype('float64'), torch.int16: dtype('int16'), torch.int32: dtype('int32'), torch.int64: dtype('int64'), torch.int8: dtype('int8'), torch.uint8: dtype('uint8')})