Training fails: sanitize_cli_args_for_public_logging missing + torchao import warning

Hi, I am unable to run training with SimpleTuner even after installing all required dependencies, including torchao. Environment Python: 3.12 CUDA: 13.0 SimpleTuner: latest release branch (Docker setup) # Issue 1: torchao warning Even after installing torchao explicitly: #### pip install torchao or #### pip install "torchao>=0.17.0,<0.18.0" I still get: ### "Unable to import `torchao` Tensor objects. This may affect loading checkpoints serialized with `torchao`" ( It looks like torchao is installed but not fully usable or not properly detected by SimpleTuner. ) # Issue 2: training crash ## Training fails immediately with: [RANK 0] 2026-06-18 15:50:50,679 [ERROR] Traceback (most recent call last): File "<string>", line 322, in <module> File "/app/SimpleTuner/simpletuner/helpers/training/trainer.py", line 7304, in run_trainer_job accelerate_result = _launch_with_accelerate() ^^^^^^^^^^^^^^^^^^^^^^^^^ File "/app/SimpleTuner/simpletuner/helpers/training/trainer.py", line 7183, in _launch_with_accelerate shlex.join(sanitize_cli_args_for_public_logging(cmd)), ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ NameError: name 'sanitize_cli_args_for_public_logging' is not defined ## Request Could you please: Confirm the correct torchao version required Fix or restore sanitize_cli_args_for_public_logging Ensure release branch is consistent and runnable Thanks