Getting TypeError: RayTaskError.as_instanceof_cause.<locals>.cls.__init__() takes 2 positional arguments but 4 were given

[NikhilK-crypto]

After running below code

# Trainer
trainer = TorchTrainer(
    train_loop_per_worker=train_loop_per_worker,
    train_loop_config=train_loop_config,
    scaling_config=scaling_config,
    run_config=run_config,
    datasets={"train": train_ds, "val": val_ds},
    dataset_config=dataset_config,
    metadata={"class_to_index": preprocessor.class_to_index}
)

%%time
# Train
results = trainer.fit()

Setuptools version - 69.5.1
Ray version - 2.7.0 also tried 2.7.1 & 2.7.2

getting below error -

---

TypeError Traceback (most recent call last)
File :2

File /opt/anaconda3/envs/madewithml/lib/python3.10/site-packages/ray/train/base_trainer.py:653, in BaseTrainer.fit(self)
647 restore_msg = TrainingFailedError._RESTORE_MSG.format(
648 trainer_cls_name=self.class.name,
649 path=str(experiment_local_path),
650 )
652 try:
--> 653 result_grid = tuner.fit()
654 except TuneError as e:
655 # Catch any TuneErrors raised by the Tuner.fit call.
656 # Unwrap the TuneError if needed.
657 parent_error = e.cause or e

File /opt/anaconda3/envs/madewithml/lib/python3.10/site-packages/ray/tune/tuner.py:372, in Tuner.fit(self)
370 if not self._is_ray_client:
371 try:
--> 372 return self._local_tuner.fit()
373 except TuneError as e:
374 raise TuneError(
375 _TUNER_FAILED_MSG.format(
376 path=self._local_tuner.get_experiment_checkpoint_dir()
377 )
378 ) from e

File /opt/anaconda3/envs/madewithml/lib/python3.10/site-packages/ray/tune/impl/tuner_internal.py:585, in TunerInternal.fit(self)
581 analysis = self._fit_resume(trainable, param_space)
583 self._experiment_analysis = analysis
--> 585 return ResultGrid(self._experiment_analysis)

File /opt/anaconda3/envs/madewithml/lib/python3.10/site-packages/ray/tune/result_grid.py:83, in ResultGrid.init(self, experiment_analysis)
78 def init(
79 self,
80 experiment_analysis: ExperimentAnalysis,
81 ):
82 self._experiment_analysis = experiment_analysis
---> 83 self._results = [
84 self._trial_to_result(trial) for trial in self._experiment_analysis.trials
85 ]

File /opt/anaconda3/envs/madewithml/lib/python3.10/site-packages/ray/tune/result_grid.py:84, in (.0)
78 def init(
79 self,
80 experiment_analysis: ExperimentAnalysis,
81 ):
82 self._experiment_analysis = experiment_analysis
83 self._results = [
---> 84 self._trial_to_result(trial) for trial in self._experiment_analysis.trials
85 ]

File /opt/anaconda3/envs/madewithml/lib/python3.10/site-packages/ray/tune/result_grid.py:317, in ResultGrid._trial_to_result(self, trial)
309 else:
310 metrics_df = self._experiment_analysis.trial_dataframes.get(
311 trial.local_path
312 )
314 result = Result(
315 checkpoint=checkpoint,
316 metrics=trial.last_result.copy(),
--> 317 error=self._populate_exception(trial),
318 _local_path=trial.local_path,
319 _remote_path=trial.remote_path,
320 _storage_filesystem=(
321 self._experiment_analysis._fs
322 if isinstance(self._experiment_analysis, ExperimentAnalysis)
323 else None
324 ),
325 metrics_dataframe=metrics_df,
326 best_checkpoints=best_checkpoints,
327 )
328 return result

File /opt/anaconda3/envs/madewithml/lib/python3.10/site-packages/ray/tune/result_grid.py:262, in ResultGrid._populate_exception(trial)
260 if trial.pickled_error_file and os.path.exists(trial.pickled_error_file):
261 with open(trial.pickled_error_file, "rb") as f:
--> 262 e = cloudpickle.load(f)
263 return e
264 elif trial.error_file and os.path.exists(trial.error_file):

TypeError: RayTaskError.as_instanceof_cause..cls.init() takes 2 positional arguments but 4 were given

[anisharitakula]

I am getting the same error too @NikhilK-crypto . Were you able to find a fix/workaround?

[CtrlMj]

same here