[MoE Refactor] Standardize Humming MoE experts + utilities (#43373)

Signed-off-by: Bill Nell <bnell@redhat.com>
Co-authored-by: Robert Shaw <114415538+robertgshaw2-redhat@users.noreply.github.com>
This commit is contained in:
bnellnm
2026-06-29 06:19:29 -07:00
committed by GitHub
co-authored by Robert Shaw
parent 59575da46d
commit bc8481af09
8 changed files with 1176 additions and 390 deletions
+13
View File
@@ -15,6 +15,7 @@ _EXPORTS: dict[str, str] = {
"dtypes": "humming.dtypes",
"DataType": "humming.dtypes:DataType",
"GemmType": "humming.config:GemmType",
"WeightScaleType": "humming.config:WeightScaleType",
"HummingMethod": "humming.layer:HummingMethod",
"HummingLayerMeta": "humming.layer:HummingLayerMeta",
"BaseInputSchema": "humming.schema:BaseInputSchema",
@@ -22,6 +23,18 @@ _EXPORTS: dict[str, str] = {
"HummingInputSchema": "humming.schema:HummingInputSchema",
"HummingWeightSchema": "humming.schema:HummingWeightSchema",
"quantize_weight": "humming.utils.weight:quantize_weight",
"AWQWeightSchema": "humming.schema:AWQWeightSchema",
"BitnetWeightSchema": "humming.schema:BitnetWeightSchema",
"ModeloptMxfp8WeightSchema": "humming.schema.modelopt:ModeloptMxfp8WeightSchema",
"ModeloptNvfp4InputSchema": "humming.schema.modelopt:ModeloptNvfp4InputSchema",
"ModeloptNvfp4WeightSchema": "humming.schema.modelopt:ModeloptNvfp4WeightSchema",
"CompressedTensorsInputSchema": "humming.schema:CompressedTensorsInputSchema",
"CompressedTensorsWeightSchema": "humming.schema:CompressedTensorsWeightSchema",
"Fp8InputSchema": "humming.schema:Fp8InputSchema",
"Fp8WeightSchema": "humming.schema.fp8:Fp8WeightSchema",
"Mxfp4WeightSchema": "humming.schema:Mxfp4WeightSchema",
"GptOssMxfp4WeightSchema": "humming.schema:GptOssMxfp4WeightSchema",
"GPTQWeightSchema": "humming.schema:GPTQWeightSchema",
}