test/datapipes/core/test_readers.py, line 200 (link)

syntax: Debug print statement should be removed

test/datapipes/core/transforms/test_subsample.py, line 126-127 (link)

syntax: Comment has a line break in the middle of a word

v2.0-MIGRATION-GUIDE.md, line 58 (link)

syntax: Extra asterisk in 'DataPipes**' should be 'DataPipes'

physicsnemo/datapipes/core/__init__.py, line 37 (link)

syntax: Inconsistent naming in example - uses dp.Downsample but exports SubsamplePoints

examples/minimal/datapipes/tutorial_04_hydra_config.py, line 54 (link)

style: torch is imported but not used in this tutorial

_{Note: If this suggestion doesn't match your team's coding style, reply to this and let me know. I'll remember it for next time!}

test/datapipes/core/test_dataset.py, line 238 (link)

style: Inconsistent device specification - uses cuda:0 in line 138 but cuda here without device index

examples/minimal/datapipes/generate_variable_points_data.py, line 298 (link)

syntax: Backend choice inconsistency: example shows '-b npy' but valid choices are only 'npz' and 'zarr'

test/datapipes/core/test_collate.py, line 227-228 (link)

logic: Test function returns first sample without proper tuple unpacking - should return (data, metadata) tuple for consistency

examples/minimal/datapipes/README.md, line 25 (link)

syntax: Typo: 'ascynchronous' should be 'asynchronous'

examples/minimal/datapipes/README.md, line 30 (link)

syntax: Typo: 'reproducability' should be 'reproducibility'

examples/minimal/datapipes/README.md, line 81 (link)

syntax: Grammar: 'for understand' should be 'for understanding'

physicsnemo/datapipes/core/dataloader.py, line 56 (link)

syntax: Import statement uses incorrect module name - should be from physicsnemo.datapipes import ... instead of from datapipe import ...

examples/minimal/datapipes/tutorial_02_transforms.py, line 292 (link)

syntax: Grammar: 'it's' should be 'its' (possessive, not contraction)

.github/CODEOWNERS, line 62 (link)

style: Trailing whitespace after @pzharrington should be removed

examples/minimal/datapipes/tutorial_03_custom_gnn_datapipe.py, line 241-242 (link)

logic: Variable n_edges is misleading - this represents edges per node (k value), not total edges in the graph.

examples/minimal/datapipes/tutorial_03_custom_gnn_datapipe.py, line 246-247 (link)

logic: The calculation n_edges / n_nodes is incorrect for average degree - should be n_edges since n_edges already represents k neighbors per node.

physicsnemo/datapipes/core/collate.py, line 169 (link)

logic: Using torch.stack() on TensorDict objects directly - should verify this is the intended API as TensorDict.stack() might be the correct method

Should this be TensorDict.stack(data_list, dim=self.stack_dim) instead of torch.stack()?

physicsnemo/datapipes/core/collate.py, line 353 (link)

logic: The function signature indicates it returns a tuple but uses a DefaultCollator instance that may not collate metadata by default

Should the _default_collator be initialized with collate_metadata=True to match the return type annotation?

test/datapipes/core/test_transforms.py, line 517-612 (link)

style: Large blocks of commented test code should be removed or converted to proper TODO issues rather than left in the codebase

_{Note: If this suggestion doesn't match your team's coding style, reply to this and let me know. I'll remember it for next time!}

test/datapipes/core/test_transforms.py, line 619-659 (link)

style: Another large block of commented test code that should be cleaned up

_{Note: If this suggestion doesn't match your team's coding style, reply to this and let me know. I'll remember it for next time!}

examples/minimal/datapipes/tutorial_01_getting_started.py, line 300 (link)

logic: This line will cause a TypeError. batch_data is a TensorDict, not a tuple/list, so batch_data[1] is invalid. Remove this debug print statement or access a valid key.

test/datapipes/core/test_transforms.py, line 764-767 (link)

style: Remove commented test code to keep the file clean

_{Note: If this suggestion doesn't match your team's coding style, reply to this and let me know. I'll remember it for next time!}

physicsnemo/datapipes/core/transforms/compose.py, line 93-95 (link)

style: Missing return type annotation for __iter__ method. Should be Iterator[Transform].

physicsnemo/datapipes/core/readers/numpy.py, line 186 (link)

style: This random sampling could cause reproducibility issues if no seed is set. Consider using a seeded random state or documenting the need for external seed management. Should this method accept a random state parameter or use a class-level random state to ensure reproducible subsampling?

_{Note: If this suggestion doesn't match your team's coding style, reply to this and let me know. I'll remember it for next time!}

physicsnemo/datapipes/core/readers/numpy.py, line 252 (link)

style: The np.array(arr) wrapping may create an unnecessary copy. Since arr is already a numpy array from npz, torch.from_numpy(arr) should work directly.

_{Note: If this suggestion doesn't match your team's coding style, reply to this and let me know. I'll remember it for next time!}

physicsnemo/datapipes/core/transforms/geometric.py, line 53-57 (link)

syntax: Example uses undefined Sample class - should use TensorDict

physicsnemo/datapipes/core/transforms/geometric.py, line 225-228 (link)

syntax: Example references TranslationInvariance but class is named Translate

physicsnemo/datapipes/core/transforms/geometric.py, line 282-287 (link)

logic: Logic error: device assignment after type error could be unreachable

physicsnemo/datapipes/core/transforms/geometric.py, line 305-308 (link)

syntax: Example references ScaleInvariance but class is named ReScale

physicsnemo/datapipes/core/transforms/subsample.py, line 147-150 (link)

style: The example uses Sample but the actual parameter type is TensorDict. This inconsistency could confuse users.

_{Note: If this suggestion doesn't match your team's coding style, reply to this and let me know. I'll remember it for next time!}

physicsnemo/datapipes/core/transforms/subsample.py, line 244 (link)

logic: Missing validation that weights tensor has the same length as the data tensors being sampled.

physicsnemo/datapipes/core/readers/tensorstore_zarr.py, line 34 (link)

style: Using check_version_spec instead of the recommended check_min_version from MOD-011. The coding standards specify using check_min_version(package, version, hard_fail=False) for optional dependency handling.

Context Used: File from greptile.json - CODING_STANDARDS/MODELS_IMPLEMENTATION.md (source)

physicsnemo/datapipes/core/readers/tensorstore_zarr.py, line 179-185 (link)

style: The _spec_template is defined but never used in the implementation. Consider removing it or documenting its intended purpose.

_{Note: If this suggestion doesn't match your team's coding style, reply to this and let me know. I'll remember it for next time!}

physicsnemo/datapipes/core/readers/tensorstore_zarr.py, line 272 (link)

style: Field discovery is called twice (line 167 and here) for the same group, which is inefficient. Consider caching the result from the first discovery.

_{Note: If this suggestion doesn't match your team's coding style, reply to this and let me know. I'll remember it for next time!}

physicsnemo/datapipes/core/transforms/spatial.py, line 53-58 (link)

logic: Example uses undefined Sample class instead of TensorDict. Should be consistent with actual data structure used in the transforms.

physicsnemo/datapipes/core/transforms/normalize.py, line 285-291 (link)

style: Device transfer modifies internal state during forward pass which could cause issues in multi-threaded environments or when the same transform is used across different devices simultaneously

_{Note: If this suggestion doesn't match your team's coding style, reply to this and let me know. I'll remember it for next time!}

physicsnemo/datapipes/core/transforms/normalize.py, line 197 (link)

syntax: Return type annotation is imprecise - should be dict[str, dict[str, torch.Tensor]] to match the actual structure returned

physicsnemo/datapipes/core/transforms/normalize.py, line 357-371 (link)

logic: Inconsistent with forward pass - inverse method doesn't update internal state when transferring to device, which could cause device mismatch issues

physicsnemo/datapipes/core/transforms/spatial.py, line 285-286 (link)

logic: Potential indexing error when k=1. Slicing [:, 1:] on a tensor with shape [M, 1, ...] results in empty tensor [M, 0, ...]. Should there be validation that k > 1, or should the slicing logic handle the k=1 case differently?

physicsnemo/datapipes/core/dataset.py, line 77 (link)

syntax: Import path in docstring example uses from datapipe import but should be from physicsnemo.datapipes.core import to match the actual module structure

physicsnemo/datapipes/core/readers/hdf5.py, line 128 (link)

style: Opening HDF5 file handle in constructor without explicit mode could lead to resource leaks if constructor fails after this point.

_{Note: If this suggestion doesn't match your team's coding style, reply to this and let me know. I'll remember it for next time!}

physicsnemo/datapipes/core/readers/vtk.py, line 32 (link)

logic: Using deprecated function name. Should use check_min_version instead of check_version_spec according to MOD-011.

Context Used: File from greptile.json - CODING_STANDARDS/MODELS_IMPLEMENTATION.md (source)

physicsnemo/datapipes/core/readers/vtk.py, line 165 (link)

logic: Duplicate key surface_normals exists in both _stl_keys (line 161) and _vtp_keys (line165). This could cause confusion when filtering keys.

physicsnemo/datapipes/core/readers/vtk.py, line 247-252 (link)

logic: Inconsistent logic: need_stl defaults to True when keys_to_read is None, but need_vtp and need_vtu require explicit keys. This asymmetry may cause unexpected behavior.

physicsnemo/datapipes/core/transforms/field_processing.py, line 111-112 (link)

logic: Scalar expansion logic may not handle multi-dimensional features correctly. A 1D feature with shape (5,) would be stacked incorrectly with scalars that get expanded to (1,).

physicsnemo/datapipes/core/transforms/field_processing.py, line 120 (link)

syntax: broadcast_to(n_points, -1) syntax is invalid. PyTorch requires explicit target shape like (n_points, fx.shape[-1]).

physicsnemo/datapipes/core/transforms/field_processing.py, line 124-128 (link)

style: __repr__ is missing n_points_key parameter in the representation string

_{Note: If this suggestion doesn't match your team's coding style, reply to this and let me know. I'll remember it for next time!}