[Quant] Add SupportsQuant to phi3 and clip (#13104)
This commit is contained in:
@@ -20,6 +20,7 @@ class NeuronQuantConfig(QuantizationConfig):
|
||||
dequant_dtype: str = "f16",
|
||||
quantize_method: str = "vector_dynamic",
|
||||
) -> None:
|
||||
super().__init__()
|
||||
self.quant_dtype = os.getenv("NEURON_QUANT_DTYPE", "s8")
|
||||
if self.quant_dtype not in SUPPORTED_QUANT_DTYPE_LIST:
|
||||
raise ValueError(
|
||||
|
||||
Reference in New Issue
Block a user