diff --git a/pyproject.toml b/pyproject.toml
index ef3833a..eda6e94 100644
--- a/pyproject.toml
+++ b/pyproject.toml
@@ -11,6 +11,7 @@ dependencies = [
     "numpy==1.26.4",
     "psutil>=7.1.0",
     "regex>=2025.9.1",
+    "setuptools>=68",
     "tiktoken>=0.11.0",
     "tokenizers>=0.22.0",
     "torch>=2.8.0",
@@ -22,15 +23,38 @@ dependencies = [
 requires = ["maturin>=1.7,<2.0"]
 build-backend = "maturin"
 
-# target torch to cuda 12.8
-[tool.uv.sources]
-torch = [
-    { index = "pytorch-cu128" },
-]
+[project.optional-dependencies]  
+cpu = [  
+    "torch>=2.8.0",  
+]  
+gpu = [  
+    "torch>=2.8.0",  
+]  
+  
+[tool.uv]  
+conflicts = [  
+    [  
+        { extra = "cpu" },  
+        { extra = "gpu" },  
+    ],  
+]  
+  
 
-[[tool.uv.index]]
-name = "pytorch-cu128"
-url = "https://download.pytorch.org/whl/cu128"
+# target torch to cuda 12.8 or CPU
+[tool.uv.sources]
+torch = [  
+    { index = "pytorch-cpu", extra = "cpu" },  
+    { index = "pytorch-cu128", extra = "gpu" },  
+]
+  
+[[tool.uv.index]]  
+name = "pytorch-cpu"  
+url = "https://download.pytorch.org/whl/cpu"  
+explicit = true  
+  
+[[tool.uv.index]]  
+name = "pytorch-cu128"  
+url = "https://download.pytorch.org/whl/cu128"  
 explicit = true
 
 [tool.maturin]
diff --git a/speedrun.sh b/speedrun.sh
index d2498ee..d6d812d 100644
--- a/speedrun.sh
+++ b/speedrun.sh
@@ -23,7 +23,9 @@ command -v uv &> /dev/null || curl -LsSf https://astral.sh/uv/install.sh | sh
 # create a .venv local virtual environment (if it doesn't exist)
 [ -d ".venv" ] || uv venv
 # install the repo dependencies
-uv sync
+# picking the NVIDIA GPU if available, otherwise CPU
+choose_extra() { nvidia-smi -L >/dev/null 2>&1 && printf gpu || printf cpu; }
+uv sync --extra "$(choose_extra)"
 # activate venv so that `python` uses the project's venv instead of system python
 source .venv/bin/activate