Method dequantize

deepspeed/ops/fp_quantizer/quantize.py:39–40 · view source on GitHub ↗

(self, input_q, fp_out=None, q_bits=8, q_mantisa_bits=3, scale=None)

Source from the content-addressed store, hash-verified

37
38	@abc.abstractmethod
39	def dequantize(self, input_q, fp_out=None, q_bits=8, q_mantisa_bits=3, scale=None) -> torch.Tensor:
40	...
41
42
43	class FP_Quantize(Quantizer):

matmul_fp8_fallbackFunction · 0.45

dequantizeMethod · 0.45

dequantizedMethod · 0.45

no outgoing calls

no test coverage detected