Albert Zeyer

@vieting I pushed another small change. Can you try again?

Is that the original stdout + stderr, or just the log? It looks a bit like maybe RASR does not correctly starts at all? You should e.g. see this then...

What about this? ``` configuration error: failed to open file "neural-network-trainer.config" for reading. (No such file or directory) ```

And in your stdout, you see the actual error: ``` Fatal Python error: Segmentation fault Current thread 0x00007f69453ea380 (most recent call first): File "/work/asr4/vieting/tmp/20231108_tf213_sprint_op/returnn/returnn/sprint/control.py", line 499 in _handle_cmd_export_allophone_state_fsa_by_segment_name File "/work/asr4/vieting/tmp/20231108_tf213_sprint_op/returnn/returnn/sprint/control.py",...

There it seems that RASR does not start at all. I see: ``` /work/asr4/vieting/programs/rasr/20230707/rasr/arch/linux-x86_64-standard/nn-trainer.linux-x86_64-standard: error while loading shared libraries: libtensorflow_cc.so.2: cannot open shared object file: No such file or directory...

Btw, the RASR segmentation fault looks actually like a bug in RASR. RASR should never segfault.

Whenever RASR gives a segfault, that's a bug in RASR. It should never segfault. Can you link corresponding RASR issues here? Or if this is not reported yet, can you...

Why do you put "chore:" into the description? Where do you get this from? This is inconsistent to our normal commit messages, so we should not use this.

Does this really need to be a new option? I personally prefer to have fewer options if possible, esp if not really needed. Can't we just always do those integrity...

Please check failing tests. ``` returnn/datasets/hdf.py:548: WEAK WARNING PyIncorrectDocstringInspection: Missing parameter data in docstring returnn/datasets/hdf.py:548: WEAK WARNING PyIncorrectDocstringInspection: Missing parameter seq_name in docstring returnn/datasets/hdf.py:550: WARNING PyUnresolvedReferencesInspection: Function 'check_data_integrity' does not...