Send metrics on request start

lib/metrics.py

@@ -45,6 +45,7 @@ class Metrics:
         self.model_metrics.workload_received += workload
         self.model_metrics.requests_recieved.add(reqnum)
         self.model_metrics.requests_working.add(reqnum)
+        self.update_pending = True
 
     def _request_end(self, workload: float, reqnum: int) -> None:
         """

start_server.sh

@@ -58,16 +58,10 @@ then
         source ~/.local/bin/env
     fi
 
-    if [[ ! -d $SERVER_DIR ]]; then
+    # Fork testing
-        git clone --depth=1 "${PYWORKER_REPO:-https://github.com/vast-ai/pyworker}" "$SERVER_DIR"
+    [[ ! -d $SERVER_DIR ]] && git clone "${PYWORKER_REPO:-https://github.com/vast-ai/pyworker}" "$SERVER_DIR"
-    fi
-
     if [[ -n ${PYWORKER_REF:-} ]]; then
-        (
+        (cd "$SERVER_DIR" && git checkout "$PYWORKER_REF")
-            cd "$SERVER_DIR"
-            git fetch --depth=1 origin "$PYWORKER_REF"
-            git checkout "$PYWORKER_REF"
-        )
     fi
 
     uv venv --python-preference only-managed "$ENV_PATH" -p 3.10