hub / github.com/deepspeedai/DeepSpeed / quantize

Method quantize

deepspeed/ops/fp_quantizer/quantize.py:30–36 · view source on GitHub ↗

(self,
                 input,
                 q_bits=8,
                 q_mantisa_bits=3,
                 stochastic_mode=False,
                 return_meta_tensor=False)

Source from the content-addressed store, hash-verified

28
29	@abc.abstractmethod
30	def quantize(self,
31	input,
32	q_bits=8,
33	q_mantisa_bits=3,
34	stochastic_mode=False,
35	return_meta_tensor=False) -> torch.Tensor:
36	...
37
38	@abc.abstractmethod
39	def dequantize(self, input_q, fp_out=None, q_bits=8, q_mantisa_bits=3, scale=None) -> torch.Tensor:

Callers 2

__init__Method · 0.45

quantizeMethod · 0.45

Calls

no outgoing calls

Tested by

no test coverage detected