[Refactor] refactor humming linear and moe backends to use explicit layer configs (#49610)

Signed-off-by: Jinzhen Lin <jinzhen.ljz@antgroup.com>
Co-authored-by: OpenAI Codex <noreply@openai.com>
This commit is contained in:
Jinzhen Lin
2026-08-07 10:03:19 -07:00
committed by GitHub
co-authored by OpenAI Codex
parent e229fdbc87
commit fcde8e1460
37 changed files with 448 additions and 341 deletions
+3 -2
View File
@@ -161,7 +161,6 @@ def test_moe_wna16_setup_forwards_selected_backend(monkeypatch):
assert method.moe_kernel is kernel
assert captured["backend"] == WNA16MoEBackend.HUMMING
assert captured["layer"] is layer
def test_moe_wna16_humming_adapter_repacks_uint8_tensors():
@@ -198,7 +197,9 @@ def test_moe_wna16_uses_humming_quant_config(monkeypatch):
monkeypatch.setattr(
humming_utils,
"get_humming_moe_quant_config",
lambda actual_layer: quant_config if actual_layer is layer else None,
lambda actual_layer, *args, **kwargs: (
quant_config if actual_layer is layer else None
),
)
assert method.get_fused_moe_quant_config(layer) is quant_config