[Refactor] Remove dead code (#43234)
Signed-off-by: yewentao256 <zhyanwentao@126.com> Co-authored-by: mergify[bot] <37929162+mergify[bot]@users.noreply.github.com>
This commit is contained in:
co-authored by
mergify[bot] <37929162+mergify[bot]@users.noreply.github.com>
parent
d63108fb18
commit
710f077617
@@ -1518,7 +1518,6 @@ def fused_experts(
|
||||
def _get_config_quant_dtype(
|
||||
use_fp8_w8a8: bool,
|
||||
use_int8_w8a8: bool,
|
||||
ocp_mx_scheme: str | None,
|
||||
) -> None | torch.dtype | str:
|
||||
"""
|
||||
Get the quantization type based on the quantization strategy flags.
|
||||
@@ -1529,18 +1528,8 @@ def _get_config_quant_dtype(
|
||||
"""
|
||||
if use_fp8_w8a8:
|
||||
return current_platform.fp8_dtype()
|
||||
elif use_int8_w8a8:
|
||||
if use_int8_w8a8:
|
||||
return torch.int8
|
||||
elif ocp_mx_scheme == "w_mxfp4_a_mxfp4":
|
||||
return "mxfp4"
|
||||
elif ocp_mx_scheme in {"w_mxfp4_a_mxfp6_e3m2", "w_mxfp6_e3m2_a_mxfp6_e3m2"}:
|
||||
return "mxfp6_e3m2"
|
||||
elif ocp_mx_scheme in {"w_mxfp4_a_mxfp6_e2m3", "w_mxfp6_e2m3_a_mxfp6_e2m3"}:
|
||||
return "mxfp6_e2m3"
|
||||
elif ocp_mx_scheme in {"w_mxfp4", "w_mxfp6_e3m2", "w_mxfp6_e2m3"}:
|
||||
return torch.bfloat16
|
||||
elif ocp_mx_scheme in {"w_mxfp4_a_fp8", "w_mxfp6_e3m2_a_fp8", "w_mxfp6_e2m3_a_fp8"}:
|
||||
return torch.float8_e4m3fn
|
||||
|
||||
return None
|
||||
|
||||
@@ -1615,7 +1604,6 @@ def fused_experts_impl(
|
||||
quant_dtype = _get_config_quant_dtype(
|
||||
use_fp8_w8a8=use_fp8_w8a8,
|
||||
use_int8_w8a8=use_int8_w8a8,
|
||||
ocp_mx_scheme=None,
|
||||
)
|
||||
|
||||
get_config_func = functools.partial(
|
||||
|
||||
@@ -50,9 +50,6 @@ WEIGHT_LOADER_V2_SUPPORTED = [
|
||||
"AWQLinearMethod",
|
||||
"AutoGPTQLinearMethod",
|
||||
"Fp8LinearMethod",
|
||||
"MarlinLinearMethod",
|
||||
"GPTQMarlin24LinearMethod",
|
||||
"TPUInt8LinearMethod",
|
||||
"FBGEMMFp8LinearMethod",
|
||||
"ModelOptFp8LinearMethod",
|
||||
"ModelOptFp8PcPtLinearMethod",
|
||||
|
||||
@@ -46,7 +46,6 @@ QuantizationMethods = Literal[
|
||||
QUANTIZATION_METHODS: list[str] = list(get_args(QuantizationMethods))
|
||||
|
||||
DEPRECATED_QUANTIZATION_METHODS = [
|
||||
"tpu_int8",
|
||||
"fbgemm_fp8",
|
||||
"fp_quant",
|
||||
]
|
||||
|
||||
Reference in New Issue
Block a user