diff --git a/packaging/Makefile b/packaging/Makefile
index a601f1c9a..eaea196bb 100644
--- a/packaging/Makefile
+++ b/packaging/Makefile
@@ -48,7 +48,7 @@ download-ipfs-kubo: target-dir build-dir
 	curl -fsSL https://github.com/ipfs/kubo/releases/download/v0.23.0/kubo_v0.23.0_linux-amd64.tar.gz | tar -xz --directory ./target/kubo
 
 target/bin/sevctl:
-	cargo install --git https://github.com/virtee/sevctl.git --rev v0.6.0 --target x86_64-unknown-linux-gnu --root ./target
+	cargo install --locked --git https://github.com/virtee/sevctl.git --rev v0.6.0 --target x86_64-unknown-linux-gnu --root ./target
 	./target/bin/sevctl -V
 
 version:
diff --git a/src/aleph/vm/pool.py b/src/aleph/vm/pool.py
index 90cd697a1..edcccd43a 100644
--- a/src/aleph/vm/pool.py
+++ b/src/aleph/vm/pool.py
@@ -249,7 +249,7 @@ async def load_persistent_executions(self):
             if execution.is_running:
                 # TODO: Improve the way that we re-create running execution
                 # Load existing GPUs assigned to VMs
-                execution.gpus = parse_raw_as(List[HostGPU], saved_execution.gpus)
+                execution.gpus = parse_raw_as(List[HostGPU], saved_execution.gpus) if saved_execution.gpus else []
                 # Load and instantiate the rest of resources and already assigned GPUs
                 await execution.prepare()
                 if self.network: