From c32513b99d29b3c6efee38df752a696bc2696c2f Mon Sep 17 00:00:00 2001
From: =?UTF-8?q?Eldar=20Kurti=C4=87?=
 <8884008+eldarkurtic@users.noreply.github.com>
Date: Wed, 30 Sep 2026 18:08:31 +0200
Subject: [PATCH] Add support for unquantized ngram in CT format (#59431)

---
 vllm/models/qwen4_exp/common/ngram_embedding.py | 8 ++++++++
 1 file changed, 8 insertions(+)

diff --git a/vllm/models/qwen4_exp/common/ngram_embedding.py b/vllm/models/qwen4_exp/common/ngram_embedding.py
index e7371770f8..9e6eef03df 100644
--- a/vllm/models/qwen4_exp/common/ngram_embedding.py
+++ b/vllm/models/qwen4_exp/common/ngram_embedding.py
@@ -23,6 +23,9 @@ from vllm.model_executor.layers.quantization.base_config import (
     QuantizationConfig,
     QuantizeMethodBase,
 )
+from vllm.model_executor.layers.quantization.compressed_tensors.compressed_tensors import (  # noqa: E501
+    CompressedTensorsConfig,
+)
 from vllm.model_executor.layers.quantization.fp8 import Fp8Config
 from vllm.model_executor.layers.quantization.modelopt import (
     ModelOptMixedPrecisionConfig,
@@ -182,6 +185,11 @@ class Qwen4ExpPLEEmbeddingMethod(QuantizeMethodBase):
             quant_config, ModelOptQuantConfigBase
         ) and quant_config.is_layer_excluded(prefix):
             return Qwen4ExpPLEUnquantizedEmbeddingMethod()
+        if (
+            isinstance(quant_config, CompressedTensorsConfig)
+            and quant_config.get_scheme_dict(None, layer_name=prefix) is None
+        ):
+            return Qwen4ExpPLEUnquantizedEmbeddingMethod()
         if not isinstance(quant_config, Fp8Config):
             raise NotImplementedError(
                 "Qwen4Exp PLE embedding does not support quantization config "