improving efficiency of mlflow metric logging (#14232)

Signed-off-by: Walter Martin <wamartin@microsoft.com>

src/transformers/integrations.py

@@ -726,9 +726,10 @@ class MLflowCallback(TrainerCallback):
         if not self._initialized:
             self.setup(args, state, model)
         if state.is_world_process_zero:
+            metrics = {}
             for k, v in logs.items():
                 if isinstance(v, (int, float)):
-                    self._ml_flow.log_metric(k, v, step=state.global_step)
+                    metrics[k] = v
                 else:
                     logger.warning(
                         f"Trainer is attempting to log a value of "
@@ -736,6 +737,7 @@ class MLflowCallback(TrainerCallback):
                         f"MLflow's log_metric() only accepts float and "
                         f"int types so we dropped this attribute."
                     )
+            self._ml_flow.log_metrics(metrics=metrics, step=state.global_step)
 
     def on_train_end(self, args, state, control, **kwargs):
         if self._initialized and state.is_world_process_zero: