diff --git a/lmlib/serve/lm_inference.py b/lmlib/serve/lm_inference.py
index bef6cee39..8b2995826 100644
--- a/lmlib/serve/lm_inference.py
+++ b/lmlib/serve/lm_inference.py
@@ -56,7 +56,7 @@ def get_gpu_memory(max_gpus: Union[int, None] = None) -> list[float]:
         with torch.cuda.device(gpu_id):
             device = torch.cuda.current_device()
             gpu_properties = torch.cuda.get_device_properties(device)
-            total_memory = gpu_properties.total_memory / (1024**3)
+            total_memory = gpu_properties.total_memory / (1024**3)  # type: ignore
             allocated_memory = torch.cuda.memory_allocated() / (1024**3)
             available_memory = total_memory - allocated_memory
             gpu_memory.append(available_memory)
diff --git a/poetry.lock b/poetry.lock
index b4bbadee8..1c532fede 100644
--- a/poetry.lock
+++ b/poetry.lock
@@ -4025,4 +4025,4 @@ multidict = ">=4.0"
 [metadata]
 lock-version = "2.0"
 python-versions = ">=3.11, <3.12"
-content-hash = "7986e728c7268aeefcee4be7c60effb32ee44bcf4f5c4d47d9d7493d0ac91760"
+content-hash = "810e8727345c0f11ba1582344c6a81555a38ae6c57f9809c7f022f7946c6c91d"
diff --git a/pyproject.toml b/pyproject.toml
index 2dfa1e6a0..c9c2b57a6 100644
--- a/pyproject.toml
+++ b/pyproject.toml
@@ -28,7 +28,7 @@ absl-py = "^2.0.0"
 names = "^0.3.0"
 together = "^0.2.4"
 pydantic = "1.10.12"
-mypy = "^1.6.0"
+mypy = "^1.8.0"
 beartype = "^0.14.0"
 torch = "^2.1"
 transformers = "^4.34.0"