[Quant] Enable humming w[2-7]a[4,8] inference with compressed-tensors (#46390)

Signed-off-by: HDCharles <charlesdavidhernandez@gmail.com>
This commit is contained in:
HDCharles
2026-07-14 20:22:31 -06:00
committed by GitHub
parent 3ad85e0de4
commit 9b2be4e9a5
8 changed files with 544 additions and 3 deletions
@@ -0,0 +1,8 @@
model_name: "nm-testing/Qwen3-4B-mixed-quant-RTN-wNaM"
accuracy_threshold: 0.85
num_questions: 1319
num_fewshot: 5
server_args: >-
--enforce-eager
--max-model-len 4096
--linear-backend humming
@@ -2,3 +2,4 @@ Qwen3-30B-A3B-Instruct-2507-quantized.w8a8-humming-act-int8.yaml
Qwen3-30B-A3B-GPTQ-Int4-humming-act-int8.yaml
Qwen3-30B-A3B-AWQ-humming.yaml
Qwen3.5-35B-A3B-experts-int8-humming-act-int8.yaml
Qwen3-4B-mixed-quant-RTN-humming.yaml