From c5ecc7a4de05e51c8a3c119e93d53a1151edce9a Mon Sep 17 00:00:00 2001
From: =?UTF-8?q?Dani=C3=ABl=20de=20Kok?= <me@danieldk.eu>
Date: Mon, 10 Mar 2025 15:08:23 +0100
Subject: [PATCH] Small test and typing fixes (#3078)

* test_weights: add modules_to_not_convert

* More typing fixes
---
 server/tests/utils/test_weights.py                    | 2 ++
 server/text_generation_server/layers/gptq/__init__.py | 2 --
 server/text_generation_server/utils/quantization.py   | 4 ++--
 3 files changed, 4 insertions(+), 4 deletions(-)

diff --git a/server/tests/utils/test_weights.py b/server/tests/utils/test_weights.py
index 556fcea1e..b21749d11 100644
--- a/server/tests/utils/test_weights.py
+++ b/server/tests/utils/test_weights.py
@@ -25,6 +25,7 @@ def gptq_weights_loader():
         quant_method="gptq",
         quantize="gptq",
         sym=True,
+        modules_to_not_convert=[],
     )
 
 
@@ -37,6 +38,7 @@ def gptq_weights_loader_awq():
         quant_method="awq",
         quantize="awq",
         sym=True,
+        modules_to_not_convert=[],
     )
 
 
diff --git a/server/text_generation_server/layers/gptq/__init__.py b/server/text_generation_server/layers/gptq/__init__.py
index f8a62cf53..b5549916b 100644
--- a/server/text_generation_server/layers/gptq/__init__.py
+++ b/server/text_generation_server/layers/gptq/__init__.py
@@ -188,8 +188,6 @@ class GPTQWeightsLoader(WeightsLoader):
     def is_layer_skipped_quantization(
         self, prefix: str, modules_to_not_convert: List[str]
     ):
-        if modules_to_not_convert is None:
-            return False
         return any(module_name in prefix for module_name in modules_to_not_convert)
 
     def get_weights_col_packed(
diff --git a/server/text_generation_server/utils/quantization.py b/server/text_generation_server/utils/quantization.py
index 7324b33ff..e460361af 100644
--- a/server/text_generation_server/utils/quantization.py
+++ b/server/text_generation_server/utils/quantization.py
@@ -21,7 +21,7 @@ class _QuantizerConfig:
     quant_method: str
     sym: bool
     weight_block_size: Optional[List[int]]
-    modules_to_not_convert: Optional[List[str]]
+    modules_to_not_convert: List[str]
 
 
 @dataclass
@@ -52,7 +52,7 @@ def _get_quantizer_config(model_id, revision):
     sym = False
     desc_act = False
     weight_block_size = None
-    modules_to_not_convert = None
+    modules_to_not_convert = []
 
     filename = "config.json"
     try: