Refactor WorkflowRuner and Scheduler to handling partially completed tasks

docs/sphinx_doc/source/tutorial/trinity_configs.md

@@ -412,15 +412,16 @@ explorer:
     engine_type: vllm
     engine_num: 1
     tensor_parallel_size: 1
-    enable_history: False
+    enable_history: false
   auxiliary_models:
   - model_path: Qwen/Qwen2.5-7B-Instruct
     tensor_parallel_size: 1
   eval_interval: 100
-  eval_on_startup: True
+  eval_on_startup: true
   over_rollout:
     ratio: 0.0
     wait_after_min: 30.0
+    return_partial_tasks: false
   dynamic_timeout:
     enable: false
     ratio: 3.0
@@ -443,13 +444,14 @@ explorer:
   - `external`: Use external API-based model engine.
 - `rollout_model.engine_num`: Number of inference engines.
 - `rollout_model.tensor_parallel_size`: Degree of tensor parallelism.
-- `rollout_model.enable_history`: Whether to enable model call history recording. If set to `True`, the model wrapper automatically records the return experiences of model calls. Please periodically extract the history via `extract_experience_from_history` to avoid out-of-memory issues. Default is `False`.
+- `rollout_model.enable_history`: Whether to enable model call history recording. If set to `true`, the model wrapper automatically records the return experiences of model calls. Please periodically extract the history via `extract_experience_from_history` to avoid out-of-memory issues. Default is `false`.
 - `auxiliary_models`: Additional models used for custom workflows.
 - `eval_interval`: Interval (in steps) for evaluating the model.
 - `eval_on_startup`: Whether to evaluate the model on startup. More precisely, at step 0 with the original model, so it will not be triggered when restarting.
 - `over_rollout`: [Experimental] Configurations for over-rollout mechanism, which allows the explorer to proceed with fewer tasks than the full batch size. It effectively increases throughput in scenarios where some tasks take significantly longer to complete than others. Only applicable when dynamic synchronization (`synchronizer.sync_style` is not `fixed`) is used.
   - `ratio`: Explorer will only wait for `(1 - ratio) * batch_size` of tasks at each step. Default is `0.0`, meaning waiting for all tasks.
   - `wait_after_min`: After reaching the minimum task threshold, wait for this many seconds before proceeding. Default is `30.0` seconds.
+  - `return_partial_tasks`: Whether to return the results of tasks that have only completed partially (e.g., only some runs in GRPO). Default is `false`, meaning only return results of tasks that have completed all runs.
 - `dynamic_timeout`: [Experimental] Configurations for dynamic timeout mechanism, which adjusts the timeout for each task based on the average time taken for successful tasks.
   - `enable`: Whether to enable dynamic timeout. Default is `false`.
   - `ratio`: The timeout for each task is dynamically set to `average_time_per_success_task * ratio`. Default is `3.0`.

docs/sphinx_doc/source_zh/tutorial/trinity_configs.md

@@ -409,15 +409,16 @@ explorer:
     engine_type: vllm
     engine_num: 1
     tensor_parallel_size: 1
-    enable_history: False
+    enable_history: false
   auxiliary_models:
   - model_path: Qwen/Qwen2.5-7B-Instruct
     tensor_parallel_size: 1
   eval_interval: 100
-  eval_on_startup: True
+  eval_on_startup: true
   over_rollout:
     ratio: 0.0
     wait_after_min: 30.0
+    return_partial_tasks: false
   dynamic_timeout:
     enable: false
     ratio: 3.0
@@ -447,6 +448,7 @@ explorer:
 - `over_rollout`: [实验性] 超量 rollout 机制的配置，允许 explorer 在每个步骤中使用少于完整批次大小的任务继续进行。这在某些任务显著耗时较长的场景中能有效地提高吞吐量。仅当使用动态同步（`synchronizer.sync_style` 不是 `fixed`）时适用。
   - `ratio`: explorer 在每个步骤中仅等待 `(1 - ratio) * batch_size` 的任务。默认为 `0.0`，表示等待所有任务。
   - `wait_after_min`: 达到最小任务阈值后，等待此秒数后再继续。
+  - `return_partial_tasks`: 是否返回仅部分完成的任务结果（例如，在 GRPO 中仅完成部分 run 的任务）。默认为 `false`，表示仅返回已完成组内所有 run 的任务结果。
 - `dynamic_timeout`: [实验性] 动态超时机制的配置，根据成功任务的平均耗时调整每个任务的超时时间。
   - `enable`: 是否启用动态超时。默认为 `false`。
   - `ratio`: 每个任务的超时时间动态设置为 `average_time_per_success_task * ratio`。默认为 `3.0`。

pyproject.toml

@@ -60,7 +60,7 @@ data = [
     "py-data-juicer>=1.4.3"
 ]
 agent = [
-    "agentscope[tuner]>=1.0.18"
+    "agentscope[tuner]>=1.0.19"
 ]
 openjudge = [
     "py-openjudge>=0.2.2"