|
|
@@ -48,7 +48,7 @@ pub(crate) async fn start_cluster(
|
|
|
// "stalls". Refuse instead.
|
|
|
if nnodes > 1 && recipe.distributed_backend.is_none() {
|
|
|
bail!(
|
|
|
- "recipe '{}' runs on {} nodes but sets no distributed_backend — add `distributed_backend = "mp"` (or "ray") so the launcher emits the rendezvous flags; otherwise every node would launch an independent single-node engine and try to load the full checkpoint",
|
|
|
+ "recipe '{}' runs on {} nodes but sets no distributed_backend — add distributed_backend = \"mp\" (or \"ray\") so the launcher emits the rendezvous flags; otherwise every node launches an independent single-node engine and tries to load the full checkpoint",
|
|
|
recipe_name,
|
|
|
nnodes
|
|
|
);
|