trainer: converge to VM2 patches + re-apply _portable export (recovered from transcript)

scripts/meta_learner_trainer.py

@@ -1552,10 +1552,23 @@ def _export_deployment(export_row, results_df, out_dir, ce_configs, kl_configs,
                   clean_name(model_name), e))
             return declared_cols
 
+    def _portable(path):
+        """Deployment configs must survive machine moves (VM1 <-> VM2 <-> HF).
+        Store artifact dirs relative to the project root (cwd at export time);
+        the inference script's resolver finds them via its cwd rule on any
+        machine. Absolute paths are kept only if outside the project tree."""
+        if not path:
+            return None
+        rp = Path(path).resolve()
+        try:
+            return str(rp.relative_to(Path.cwd()))
+        except ValueError:
+            return str(rp)
+
     model_entries = []
     for cname in exp_ce:
         c = _model_info(cname, "CE")
-        ce_art = str(Path(args.ce_artifacts_dir).resolve()) if args.ce_artifacts_dir else None
+        ce_art = _portable(args.ce_artifacts_dir)
         verified_cols = _verify_other_cols(
             c.get("model_name", cname), ce_art,
             c.get("other_cols", []), c.get("split_unknown_stage", False))
@@ -1574,7 +1587,7 @@ def _export_deployment(export_row, results_df, out_dir, ce_configs, kl_configs,
         })
     for cname in exp_kl:
         c = _model_info(cname, "KL")
-        kl_art = str(Path(args.kl_artifacts_dir).resolve()) if args.kl_artifacts_dir else None
+        kl_art = _portable(args.kl_artifacts_dir)
         verified_cols = _verify_other_cols(
             c.get("model_name", cname), kl_art,
             c.get("other_cols", []), c.get("split_unknown_stage", False))