Torchvsion API documentation page update

[mdabek-nvidia]

Category:

Other (Documentation)

Description:

Update of the Torchvision API's documentation:

• Added tables with supported operators for both APIs

• Updated cells that list supported operators

Additional information:

Affected modules and functionalities:

Key points relevant for the review:

Tests:

• Existing tests apply
• New tests added
  • Python tests
  • GTests
  • Benchmark
  • Other
• N/A

Checklist

Documentation

• Existing documentation applies
• Documentation updated
  • Docstring
  • Doxygen
  • RST
  • Jupyter
  • Other
• N/A

DALI team only

Requirements

• Implements new requirements
• Affects existing requirements
• N/A

REQ IDs: N/A

JIRA TASK: N/A

[review-notebook-app]

Check out this pull request on 

See visual diffs & provide feedback on Jupyter Notebooks.

---

Powered by ReviewNB

[mdabek-nvidia]

@greptileai please review

[greptile-apps]

Greptile Summary

This PR adds two operator reference tables to the Torchvision API getting-started notebook: one for the object-oriented API and one for the functional API. It also updates the execution-output cells to reflect the expanded operator list now available in both modules.

• Two new tables added (cells 3–5): every operator currently exported by dtv.__all__ and dtv_fn.__all__ is listed with its torchvision.transforms.v2 counterpart (linked to official Torchvision docs) and a short description; the operator counts in the tables match the updated output cells exactly.
• Addressed prior review feedback: to_tensor now carries a deprecation note ("Deprecated in Torchvision."), and PILToTensor now has a detailed description that clearly distinguishes its marker-only role from the actual-conversion role of the functional pil_to_tensor.

Confidence Score: 5/5

Documentation-only change with no executable code modifications; safe to merge.

The change adds two Markdown tables to a Jupyter notebook and refreshes output cells. The operator names in the tables match exactly what the all output cells print, previously raised review concerns about to_tensor deprecation and PILToTensor marker semantics have both been addressed, and no source code is modified.

No files require special attention.

Important Files Changed

Filename	Overview
docs/examples/getting_started/torchvision_api.ipynb	Added two operator reference tables (OO and functional) and updated output cells; operator lists in tables match the actual all output cells exactly.

Class Diagram

%%{init: {'theme': 'neutral'}}%%
classDiagram
    direction LR
    class OO_API["nvidia.dali.experimental.torchvision (OO)"] {
        CenterCrop
        ColorJitter
        Compose
        GaussianBlur
        Grayscale
        Normalize
        Pad
        PILToTensor
        RandomApply
        RandomCrop
        RandomGrayscale
        RandomHorizontalFlip
        RandomResizedCrop
        RandomVerticalFlip
        Resize
        ToPILImage
        ToPureTensor
    }
    class Functional_API["nvidia.dali.experimental.torchvision.v2.functional"] {
        center_crop
        crop
        gaussian_blur
        get_dimensions
        get_image_size
        get_size
        horizontal_flip
        normalize
        pad
        pil_to_tensor
        resize
        resized_crop
        rgb_to_grayscale
        to_grayscale
        to_pil_image
        to_tensor
        vertical_flip
    }
    class TV2["torchvision.transforms.v2 (upstream)"]
    TV2 <|.. OO_API : mirrors
    TV2 <|.. Functional_API : mirrors

Loading

_{Reviews (4): Last reviewed commit: "Available operators update" | Re-trigger Greptile}

[review-notebook-app]

View / edit / reply to this conversation on ReviewNB

JanuszL commented on 2026-04-24T12:36:49Z
----------------------------------------------------------------

I wonder if we remember ot update it, do you think we can have it generated outside of jupyter on a separated docs page?

mdabek-nvidia commented on 2026-05-07T16:36:38Z
----------------------------------------------------------------

I don't expect to have an influx of new operators, but I've created JIRA ticket to automatically generate this table: https://jirasw.nvidia.com/browse/DALI-4703

JanuszL commented on 2026-05-13T19:04:48Z
----------------------------------------------------------------

👍

[review-notebook-app]

View / edit / reply to this conversation on ReviewNB

JanuszL commented on 2026-04-24T12:36:50Z
----------------------------------------------------------------

Is it possible to make below go?

/home/mdabek/.local/lib/python3.10/site-packages/matplotlib/projections/__init__.py:63: UserWarning: Unable to import Axes3D. This may be due to multiple versions of Matplotlib being installed (e.g. as a system package and as a pip package). As a result, the 3D projection is not available. 
  warnings.warn("Unable to import Axes3D. This may be due to multiple versions of " 

mdabek-nvidia commented on 2026-05-07T16:36:53Z
----------------------------------------------------------------

Fixed

[jantonguirao]

Comments added

[mdabek-nvidia]

I don't expect to have an influx of new operators, but I've created JIRA ticket to automatically generate this table: https://jirasw.nvidia.com/browse/DALI-4703

---

View entire conversation on ReviewNB

[mdabek-nvidia]

Fixed

---

View entire conversation on ReviewNB

[jantonguirao]

Thanks for the fixes — all four prior comments are addressed.

One very minor leftover from comment #2 (the consistency suggestion): the ToPILImage row (~L108) still uses the original "Acts as a return-type hint for Compose" phrasing, while PILToTensor (~L104) and ToPureTensor (~L109) now both use the "Marker placed in a Compose op list — signals…" phrasing. Consider aligning ToPILImage for consistency, e.g.:

ToPILImage | v2.ToPILImage | Marker placed at the end of a Compose op list — signals the CHW pipeline to return a PIL.Image instead of a torch.Tensor. The conversion is performed by Compose; the operator does not modify the data. |

Not a blocker — feel free to address in a follow-up or leave as is.

[JanuszL]

👍

---

View entire conversation on ReviewNB

[dali-automaton]

CI MESSAGE: [51397152]: BUILD FAILED

[mdabek-nvidia]

!build

[dali-automaton]

CI MESSAGE: [53399787]: BUILD STARTED

[dali-automaton]

CI MESSAGE: [53399787]: BUILD PASSED