MCPcopy Create free account
hub / github.com/deepspeedai/DeepSpeed / FloatTensor

Method FloatTensor

accelerator/hpu_accelerator.py:219–220  ·  view source on GitHub ↗
(self)

Source from the content-addressed store, hash-verified

217
218 @property
219 def FloatTensor(self):
220 return functools.partial(torch.tensor, dtype=torch.float, device='hpu')
221
222 @property
223 def HalfTensor(self):

Callers 9

check_using_normMethod · 0.45
clip_grad_norm_Function · 0.45
get_flattened_grad_normFunction · 0.45
get_grad_zerosFunction · 0.45
get_weight_normFunction · 0.45
to_tensorFunction · 0.45
complete_grad_norm_calculation_for_cpu_offloadMethod · 0.45
get_grad_norm_directMethod · 0.45
complete_grad_norm_calculation_for_cpu_offloadMethod · 0.45

Calls

no outgoing calls

Tested by

no test coverage detected