refactor: rename recover count attribute (#301)

Since we will expand the world duplication count in a next PR, we need to rename recover_count to duplicate_count. Besides the worker failure when we need to increment this count, we need to also update that count for multiple scale_out / in / out scenarios. Second scale_out generated config will have the same world names as the first scale_out, therefore, there will be a world name conflict in multiworld.

Author: raresgaia123

infscale/configs/job.py

@@ -70,7 +70,7 @@ class WorldInfo:
     addr: str = "127.0.0.1"
     backend: Optional[str] = ""
     recover: bool = False
-    recover_count: int = 0
+    duplicate_count: int = 0
 
 
 @dataclass

infscale/controller/job_context.py

@@ -880,13 +880,13 @@ def _update_recovery_flow_graph(
         for world_info in recover_flow_graph:
             world_info.addr = ip
             world_info.recover = True
-            world_info.recover_count += 1
+            world_info.duplicate_count += 1
 
         for world_list in cfg.flow_graph.values():
             for world_info in world_list:
                 if recover_wid in world_info.peers:
                     world_info.recover = True
-                    world_info.recover_count += 1
+                    world_info.duplicate_count += 1
 
     def _update_recovery_worker_data(
         self, cfg: JobConfig, wrk_id: str, gpu_id: int

infscale/execution/pipeline.py

@@ -459,14 +459,14 @@ def _build_world_infos(self) -> dict[str, WorldInfo]:
                 else:
                     continue
 
-                name, backend, addr, data_port, ctrl_port, recover, recover_count = (
+                name, backend, addr, data_port, ctrl_port, recover, duplicate_count = (
                     cfg_world_info.name,
                     cfg_world_info.backend,
                     cfg_world_info.addr,
                     cfg_world_info.data_port,
                     cfg_world_info.ctrl_port,
                     cfg_world_info.recover,
-                    cfg_world_info.recover_count,
+                    cfg_world_info.duplicate_count,
                 )
 
                 world_size = len(cfg_world_info.peers) + 1
@@ -484,8 +484,8 @@ def _build_world_infos(self) -> dict[str, WorldInfo]:
                     "other_id": other_id,
                     "other": other_rank,
                     "recover": recover,
-                    "recover_count": recover_count,
-                    "multiworld_name": f"{name}-{recover_count}",
+                    "duplicate_count": duplicate_count,
+                    "multiworld_name": f"{name}-{duplicate_count}",
                 }
                 world_info = WorldInfo(**data)
                 world_infos[name] = world_info

infscale/execution/world.py

@@ -42,5 +42,5 @@ class WorldInfo:
     other: int  # other peer's rank
 
     recover: bool
-    recover_count: int
+    duplicate_count: int
     multiworld_name: str