Add warnings for Windows timeouts and clarify parallel backend options; update get_results method to conditionally save outputs

Author: recursix

src/agentlab/experiments/exp_utils.py

@@ -50,6 +50,7 @@ def timeout_manager(seconds: int = None):
     """Context manager to handle timeouts."""
     if seconds is None or sys.platform == "win32":
         try:
+            logger.warning("Timeouts are not supported on Windows.")
             yield
         finally:
             pass

src/agentlab/experiments/launch_exp.py

@@ -27,7 +27,8 @@ def run_experiments(
         exp_dir: Path
             Directory where the experiments will be saved.
         parallel_backend: str
-            Parallel backend to use. Either "joblib", "dask" or "sequential".
+            Parallel backend to use. Either "joblib", "ray" or "sequential".
+            The only backend that supports webarena graph dependencies correctly is ray or sequential.
         avg_step_timeout: int
             Will raise a TimeoutError if the episode is not finished after env_args.max_steps * avg_step_timeout seconds.
     """
@@ -51,7 +52,8 @@ def run_experiments(
         if parallel_backend == "joblib":
             from joblib import Parallel, delayed
 
-            # split sequential
+
+            # split sequential (should be no longer needed with dependencies)
             sequential_exp_args, exp_args_list = _split_sequential_exp(exp_args_list)
 
             logging.info(

src/agentlab/experiments/study.py

@@ -138,7 +138,6 @@ def run(
 
             suffix = f"trial_{i + 1}_of_{n_relaunch}"
             _, summary_df, error_report = self.get_results(suffix=suffix)
-            logger.info("\n" + error_report)
             logger.info("\n" + str(summary_df))
 
             n_incomplete, n_error = self.find_incomplete(include_errors=relaunch_errors)
@@ -202,15 +201,16 @@ def append_to_journal(self, strict_reproducibility=True):
             strict_reproducibility=strict_reproducibility,
         )
 
-    def get_results(self, suffix=""):
+    def get_results(self, suffix="", also_save=True):
         result_df = inspect_results.load_result_df(self.dir)
         error_report = inspect_results.error_report(result_df, max_stack_trace=3, use_log=True)
         summary_df = inspect_results.summarize_study(result_df)
 
-        suffix = f"_{suffix}" if suffix else ""
-        result_df.to_csv(self.dir / f"result_df{suffix}.csv")
-        summary_df.to_csv(self.dir / f"summary_df{suffix}.csv")
-        (self.dir / f"error_report{suffix}.md").write_text(error_report)
+        if also_save:
+            suffix = f"_{suffix}" if suffix else ""
+            result_df.to_csv(self.dir / f"result_df{suffix}.csv")
+            summary_df.to_csv(self.dir / f"summary_df{suffix}.csv")
+            (self.dir / f"error_report{suffix}.md").write_text(error_report)
 
         return result_df, summary_df, error_report