Weights reload configuration

CHANGELOG.md

@@ -37,4 +37,5 @@ Documenting changes which affect configuration usage patterns (added/moved/remov
 - **`orchestrator.env.log`**: Removed. Use `orchestrator.log` for env worker logging instead (2026-01-15)
 - **`orchestrator.eval.retry.reraise`**: Changed default from `True` to `False`. When `False`, raises `tenacity.RetryError` after retries are exhausted instead of the original exception, allowing failed eval environments to be skipped with a warning (#1586, 2026-01-14)
 - **`model.ep`**: Expert parallelism now supported (with auto/custom impl only), changed from the old behaviour when `ep>1` was a no-op to a proper parallelization of the MoE layers. (#1595, 2026-01-15)
+- **`orchestrator.reload_weights_on_start`**: Added flag to control resetting inference weights to the base model when starting from scratch (default: True) (2026-01-21)
 - **`orchestrator.client.elastic`**: Added elastic inference pool with DNS-based service discovery. Supports dynamic server scaling via any DNS hostname with multiple A records (Kubernetes headless services, Consul, Route53, etc.). Automatically syncs LoRA adapters on new servers and only exposes ready servers to workers (#1617, 2026-01-19)

src/prime_rl/orchestrator/config.py

@@ -647,6 +647,14 @@ class OrchestratorConfig(BaseSettings):
     # The checkpoint configuration
     ckpt: CheckpointConfig | None = None
 
+    # Whether to reset inference weights to base model when starting from scratch
+    reload_weights_on_start: Annotated[
+        bool,
+        Field(
+            description="Whether to reset inference weights to the base model when starting from scratch."
+        ),
+    ] = True
+
     # The validation configuration
     val: ValConfig | None = None
 

src/prime_rl/orchestrator/orchestrator.py

@@ -234,9 +234,14 @@ async def orchestrate(config: OrchestratorConfig):
         lora_name = config.model.lora.name if config.model.lora else None
         await inference_pool.update_weights(weights_path, lora_name=lora_name, step=scheduler.ckpt_step)
     else:
-        logger.info("Training from scratch. Resetting weights to base model")
-        if config.model.lora is None:
-            await reload_weights(admin_clients)
+        if config.reload_weights_on_start:
+            if config.model.lora is None:
+                logger.info("Training from scratch. Resetting weights to base model")
+                await reload_weights(admin_clients)
+            else:
+                logger.info("Training from scratch. Skipping base weight reload because LoRA is enabled")
+        else:
+            logger.info("Training from scratch. Skipping base weight reload")
 
     # Iterate over dataset in batches
     max_steps = config.max_steps or int(1e9)