diff --git a/server/text_generation_server/models/custom_modeling/flash_llama_modeling.py b/server/text_generation_server/models/custom_modeling/flash_llama_modeling.py
index 55b1aae9..cb0c1e85 100644
--- a/server/text_generation_server/models/custom_modeling/flash_llama_modeling.py
+++ b/server/text_generation_server/models/custom_modeling/flash_llama_modeling.py
@@ -179,7 +179,7 @@ def _load_gqa(config, prefix: str, weights):
         dim=0,
     )
 
-    if config.quantize != "gptq":
+    if config.quantize not in ["gptq", "awq"]:
         weight = weight.to(dtype=weights.dtype).to(device=weights.device)
 
         head_size = config.hidden_size // config.num_attention_heads