[BUG] TypeError: Can't instantiate abstract class AlfworldWorkflow with abstract method run

[qishenghu]

Bug Description

TypeError: Can't instantiate abstract class AlfworldWorkflow with abstract method run

I tried to run trinity run --config examples/grpo_alfworld/alfworld.yaml, it raised this error.

Environment Information

• Operating System: [e.g., Ubuntu 20.04]
• Python Version: 3.10.18
• GPU: 4xH20
• CUDA Version: 12.6
• Installation Method: git clone then pip
• Trinity-RFT Version: 0.3.1.dev0
• Other relevant dependencies or configurations you think might be helpful

Steps to Reproduce

Please provide a minimal, self-contained, and reproducible example.

trinity run --config examples/grpo_alfworld/alfworld.yaml

Expected Behavior

Should be running the training normally.

Actual Behavior

Error when trying to instantiate AlfworldWorkflow.

Log Information

(WorkflowRunner pid=30655) ERROR 09-23 13:06:48 [workflow_runner.py:149] Traceback: [repeated 21x across cluster]
(WorkflowRunner pid=30655) ERROR 09-23 13:06:48 [workflow_runner.py:149] Traceback (most recent call last): [repeated 21x across cluster]
(WorkflowRunner pid=30655) ERROR 09-23 13:06:48 [workflow_runner.py:149]   File "/root/autodl-tmp/Trinity-RFT/trinity/explorer/workflow_runner.py", line 115, in run_task [repeated 21x across cluster]
(WorkflowRunner pid=30655) ERROR 09-23 13:06:48 [workflow_runner.py:149]     exps = await self._run_task(task, repeat_times, run_id_base) [repeated 21x across cluster]
(WorkflowRunner pid=30655) ERROR 09-23 13:06:48 [workflow_runner.py:149]   File "/root/miniconda3/envs/trinity/lib/python3.10/site-packages/ray/util/tracing/tracing_helper.py", line 461, in _resume_span [repeated 42x across cluster]
(WorkflowRunner pid=30655) ERROR 09-23 13:06:48 [workflow_runner.py:149]     return await method(self, *_args, **_kwargs) [repeated 21x across cluster]
(WorkflowRunner pid=30655) ERROR 09-23 13:06:48 [workflow_runner.py:149]   File "/root/autodl-tmp/Trinity-RFT/trinity/explorer/workflow_runner.py", line 90, in _run_task [repeated 21x across cluster]
(WorkflowRunner pid=30655) ERROR 09-23 13:06:48 [workflow_runner.py:149]     self._create_workflow_instance(task) [repeated 21x across cluster]
(WorkflowRunner pid=30655) ERROR 09-23 13:06:48 [workflow_runner.py:149]     return method(self, *_args, **_kwargs) [repeated 21x across cluster]
(WorkflowRunner pid=30655) ERROR 09-23 13:06:48 [workflow_runner.py:149]   File "/root/autodl-tmp/Trinity-RFT/trinity/explorer/workflow_runner.py", line 75, in _create_workflow_instance [repeated 21x across cluster]
(WorkflowRunner pid=30655) ERROR 09-23 13:06:48 [workflow_runner.py:149]     self.workflow_instance = task.to_workflow( [repeated 21x across cluster]
(WorkflowRunner pid=30655) ERROR 09-23 13:06:48 [workflow_runner.py:149]   File "/root/autodl-tmp/Trinity-RFT/trinity/common/workflows/workflow.py", line 56, in to_workflow [repeated 21x across cluster]
(WorkflowRunner pid=30655) ERROR 09-23 13:06:48 [workflow_runner.py:149]     return self.workflow( [repeated 21x across cluster]
(WorkflowRunner pid=30655) ERROR 09-23 13:06:48 [workflow_runner.py:149] TypeError: Can't instantiate abstract class AlfworldWorkflow with abstract method run [repeated 21x across cluster]
(WorkflowRunner pid=30655) ERROR 09-23 13:06:48 [workflow_runner.py:149]  [repeated 21x across cluster]

[pan-x-c]

Thanks for reporting.
Remove the abstract run method in trinity.common.workflows.workflow.MultiTurnWorkflow can fix this issue.

[qishenghu]

Thanks. The problem has been resolved.