training with images issues with "a fps < 0"

[soshette7]

Hello , I succeed to train LTXV with a video , but with images I get errors:

• My configuration A6000 48G VRAM 60G RAM
• The train yaml I used is the same as for the video (see in attachment
  train.yaml.txt
  I had to rename train.yaml to train.yaml.txt to attach)
• my steps for training:
  • scripts/preprocess_dataset.py resultat.json --caption-column caption --video-column media_path --resolution-buckets 768x1024x1
  • python scripts/train.py train.yaml

I know that the error is because the dataset are images , because with videos it is working.

The error below:

  File "/workspace/LTX-Video-Trainer/scripts/train.py", line 55, in <module>
    app()
  File "/workspace/LTX-Video-Trainer/.venv/lib/python3.11/site-packages/typer/main.py", line 340, in __call__
    raise e
  File "/workspace/LTX-Video-Trainer/.venv/lib/python3.11/site-packages/typer/main.py", line 323, in __call__
    return get_command(self)(*args, **kwargs)
           ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/workspace/LTX-Video-Trainer/.venv/lib/python3.11/site-packages/click/core.py", line 1161, in __call__
    return self.main(*args, **kwargs)
           ^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/workspace/LTX-Video-Trainer/.venv/lib/python3.11/site-packages/typer/core.py", line 680, in main
    return _main(
           ^^^^^^
  File "/workspace/LTX-Video-Trainer/.venv/lib/python3.11/site-packages/typer/core.py", line 198, in _main
    rv = self.invoke(ctx)
         ^^^^^^^^^^^^^^^^
  File "/workspace/LTX-Video-Trainer/.venv/lib/python3.11/site-packages/click/core.py", line 1443, in invoke
    return ctx.invoke(self.callback, **ctx.params)
           ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/workspace/LTX-Video-Trainer/.venv/lib/python3.11/site-packages/click/core.py", line 788, in invoke
    return __callback(*args, **kwargs)
           ^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/workspace/LTX-Video-Trainer/.venv/lib/python3.11/site-packages/typer/main.py", line 698, in wrapper
    return callback(**use_params)
           ^^^^^^^^^^^^^^^^^^^^^^
  File "/workspace/LTX-Video-Trainer/scripts/train.py", line 51, in main
    trainer.train()
  File "/workspace/LTX-Video-Trainer/src/ltxv_trainer/trainer.py", line 223, in train
    loss = self._training_step(batch)
           ^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/workspace/LTX-Video-Trainer/src/ltxv_trainer/trainer.py", line 348, in _training_step
    training_batch = self._training_strategy.prepare_batch(batch, self._timestep_sampler)
                     ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/workspace/LTX-Video-Trainer/src/ltxv_trainer/training_strategies.py", line 259, in prepare_batch
    rope_interpolation_scale_factors = get_rope_scale_factors(fps)
                                       ^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/workspace/LTX-Video-Trainer/src/ltxv_trainer/ltxv_utils.py", line 203, in get_rope_scale_factors
    raise ValueError("FPS must be a positive number.")
ValueError: FPS must be a positive number.
terminate called without an active exception

[matanby]

Thanks for reporting this issue. We've pushed a small fix to address it (see this commit).

If you pull the latest revision and reprocess your dataset, everything should work properly.