sdc paths missing

[PierreCKL]

Hello, i noticed that in any of the dataset (from the conf.py file like the : WOD_1_0_0_TRAINING ) the sdc_paths of the simulator state is always None, but it is used to compute to compute metrics like the off road metrics in the paper.

[samuela]

I'll add that using include_sdc_paths = True does not work either:

config = dataclasses.replace(_config.WOD_1_1_0_TRAINING, max_num_objects=32, include_sdc_paths=True, num_paths=45, num_points_per_path=800)
data_iter = dataloader.simulator_state_generator(config=config)
scenario = next(data_iter)

leads to

---------------------------------------------------------------------------
InvalidArgumentError                      Traceback (most recent call last)
Cell In[24], [line 3](vscode-notebook-cell:?execution_count=24&line=3)
      [1](vscode-notebook-cell:?execution_count=24&line=1) config = dataclasses.replace(_config.WOD_1_1_0_TRAINING, max_num_objects=32, include_sdc_paths=True, num_paths=45, num_points_per_path=800)
      [2](vscode-notebook-cell:?execution_count=24&line=2) data_iter = dataloader.simulator_state_generator(config=config)
----> [3](vscode-notebook-cell:?execution_count=24&line=3) scenario = next(data_iter)

File ~/dev/waymax/waymax/dataloader/dataloader_utils.py:229, in get_data_generator(config, preprocess_fn, postprocess_fn)
    [227](https://file+.vscode-resource.vscode-cdn.net/Users/samuelainsworth/dev/waymax/docs/notebooks/~/dev/waymax/waymax/dataloader/dataloader_utils.py:227)   postprocess_fn = jax.jit(postprocess_fn)
    [228](https://file+.vscode-resource.vscode-cdn.net/Users/samuelainsworth/dev/waymax/docs/notebooks/~/dev/waymax/waymax/dataloader/dataloader_utils.py:228) for example in dataset.as_numpy_iterator():
--> [229](https://file+.vscode-resource.vscode-cdn.net/Users/samuelainsworth/dev/waymax/docs/notebooks/~/dev/waymax/waymax/dataloader/dataloader_utils.py:229)   yield postprocess_fn(example)

File /nix/store/bpr4b8wpv9a9ab5rblj2zmjmifzawggv-python3-3.12.9-env/lib/python3.12/site-packages/tensorflow/python/data/ops/dataset_ops.py:4788, in NumpyIterator.__next__(self)
   [4785](https://file+.vscode-resource.vscode-cdn.net/nix/store/bpr4b8wpv9a9ab5rblj2zmjmifzawggv-python3-3.12.9-env/lib/python3.12/site-packages/tensorflow/python/data/ops/dataset_ops.py:4785)     numpy.setflags(write=False)
   [4786](https://file+.vscode-resource.vscode-cdn.net/nix/store/bpr4b8wpv9a9ab5rblj2zmjmifzawggv-python3-3.12.9-env/lib/python3.12/site-packages/tensorflow/python/data/ops/dataset_ops.py:4786)   return numpy
-> [4788](https://file+.vscode-resource.vscode-cdn.net/nix/store/bpr4b8wpv9a9ab5rblj2zmjmifzawggv-python3-3.12.9-env/lib/python3.12/site-packages/tensorflow/python/data/ops/dataset_ops.py:4788) return nest.map_structure(to_numpy, next(self._iterator))

File /nix/store/bpr4b8wpv9a9ab5rblj2zmjmifzawggv-python3-3.12.9-env/lib/python3.12/site-packages/tensorflow/python/data/ops/iterator_ops.py:826, in OwnedIterator.__next__(self)
    [824](https://file+.vscode-resource.vscode-cdn.net/nix/store/bpr4b8wpv9a9ab5rblj2zmjmifzawggv-python3-3.12.9-env/lib/python3.12/site-packages/tensorflow/python/data/ops/iterator_ops.py:824) def __next__(self):
    [825](https://file+.vscode-resource.vscode-cdn.net/nix/store/bpr4b8wpv9a9ab5rblj2zmjmifzawggv-python3-3.12.9-env/lib/python3.12/site-packages/tensorflow/python/data/ops/iterator_ops.py:825)   try:
--> [826](https://file+.vscode-resource.vscode-cdn.net/nix/store/bpr4b8wpv9a9ab5rblj2zmjmifzawggv-python3-3.12.9-env/lib/python3.12/site-packages/tensorflow/python/data/ops/iterator_ops.py:826)     return self._next_internal()
    [827](https://file+.vscode-resource.vscode-cdn.net/nix/store/bpr4b8wpv9a9ab5rblj2zmjmifzawggv-python3-3.12.9-env/lib/python3.12/site-packages/tensorflow/python/data/ops/iterator_ops.py:827)   except errors.OutOfRangeError:
    [828](https://file+.vscode-resource.vscode-cdn.net/nix/store/bpr4b8wpv9a9ab5rblj2zmjmifzawggv-python3-3.12.9-env/lib/python3.12/site-packages/tensorflow/python/data/ops/iterator_ops.py:828)     raise StopIteration

File /nix/store/bpr4b8wpv9a9ab5rblj2zmjmifzawggv-python3-3.12.9-env/lib/python3.12/site-packages/tensorflow/python/data/ops/iterator_ops.py:776, in OwnedIterator._next_internal(self)
...
   [6005](https://file+.vscode-resource.vscode-cdn.net/nix/store/bpr4b8wpv9a9ab5rblj2zmjmifzawggv-python3-3.12.9-env/lib/python3.12/site-packages/tensorflow/python/framework/ops.py:6005)   e.message += (" name: " + str(name if name is not None else ""))
-> [6006](https://file+.vscode-resource.vscode-cdn.net/nix/store/bpr4b8wpv9a9ab5rblj2zmjmifzawggv-python3-3.12.9-env/lib/python3.12/site-packages/tensorflow/python/framework/ops.py:6006)   raise core._status_to_exception(e) from None

InvalidArgumentError: {{function_node __wrapped__IteratorGetNext_output_types_36_device_/job:localhost/replica:0/task:0/device:CPU:0}} Error in user-defined function passed to ParallelMapDatasetV2:178 transformation with iterator: Iterator::Root::Prefetch::ParallelInterleaveV4[0]::ParallelMapV2: Feature: path_samples/arc_length (data type: float) is required but could not be found.
	 [[{{node ParseExample/ParseExampleV2}}]] [Op:IteratorGetNext] name: 

[wwwxLin]

I also encountered the same problem. May I ask if you have solved it

[samuela]

@wwwxLin sdc_paths is currently internal-only but we are looking towards releasing that dataset. Stay tuned!

[vcharraut]

Hello, upon the official release, we have a built an alternative dataset that includes SDC path -> https://huggingface.co/datasets/vcharraut/V-Max_Datasets/tree/main

[samuela]

I am happy to report that WOMD 1.3.1 has been released including sdc_paths! Check out the WOMD homepage for details, the tutorial notebook for code examples, and the updated tf.Example spec.

I'll close this now. Feel free to reach out if you have any questions!