Remove transforms from public decoder API

docs/source/index.rst

@@ -133,4 +133,3 @@ Encoding
    api_ref_decoders
    api_ref_encoders
    api_ref_samplers
-   api_ref_transforms

src/torchcodec/__init__.py

@@ -9,7 +9,7 @@
 # Note: usort wants to put Frame and FrameBatch after decoders and samplers,
 # but that results in circular import.
 from ._frame import AudioSamples, Frame, FrameBatch  # usort:skip # noqa
-from . import decoders, encoders, samplers, transforms  # noqa
+from . import decoders, encoders, samplers  # noqa
 
 try:
     # Note that version.py is generated during install.

src/torchcodec/decoders/_video_decoder.py

@@ -8,18 +8,17 @@
 import json
 import numbers
 from pathlib import Path
-from typing import List, Literal, Optional, Sequence, Tuple, Union
+from typing import Literal, Optional, Tuple, Union
 
 import torch
-from torch import device as torch_device, nn, Tensor
+from torch import device as torch_device, Tensor
 
 from torchcodec import _core as core, Frame, FrameBatch
 from torchcodec.decoders._decoder_utils import (
     _get_cuda_backend,
     create_decoder,
     ERROR_REPORTING_INSTRUCTIONS,
 )
-from torchcodec.transforms import DecoderTransform, Resize
 
 
 class VideoDecoder:
@@ -68,11 +67,6 @@ class VideoDecoder:
             probably is. Default: "exact".
             Read more about this parameter in:
             :ref:`sphx_glr_generated_examples_decoding_approximate_mode.py`
-        transforms (sequence of transform objects, optional): Sequence of transforms to be
-            applied to the decoded frames by the decoder itself, in order. Accepts both
-            :class:`~torchcodec.transforms.DecoderTransform` and
-            :class:`~torchvision.transforms.v2.Transform`
-            objects. Read more about this parameter in: TODO_DECODER_TRANSFORMS_TUTORIAL.
         custom_frame_mappings (str, bytes, or file-like object, optional):
             Mapping of frames to their metadata, typically generated via ffprobe.
             This enables accurate frame seeking without requiring a full video scan.
@@ -111,7 +105,6 @@ def __init__(
         num_ffmpeg_threads: int = 1,
         device: Optional[Union[str, torch_device]] = None,
         seek_mode: Literal["exact", "approximate"] = "exact",
-        transforms: Optional[Sequence[Union[DecoderTransform, nn.Module]]] = None,
         custom_frame_mappings: Optional[
             Union[str, bytes, io.RawIOBase, io.BufferedReader]
         ] = None,
@@ -167,7 +160,6 @@ def __init__(
             device = str(device)
 
         device_variant = _get_cuda_backend()
-        transform_specs = _make_transform_specs(transforms)
 
         core.add_video_stream(
             self._decoder,
@@ -176,7 +168,6 @@ def __init__(
             num_threads=num_ffmpeg_threads,
             device=device,
             device_variant=device_variant,
-            transform_specs=transform_specs,
             custom_frame_mappings=custom_frame_mappings_data,
         )
 
@@ -448,78 +439,6 @@ def _get_and_validate_stream_metadata(
     )
 
 
-def _convert_to_decoder_transforms(
-    transforms: Sequence[Union[DecoderTransform, nn.Module]],
-) -> List[DecoderTransform]:
-    """Convert a sequence of transforms that may contain TorchVision transform
-    objects into a list of only TorchCodec transform objects.
-
-    Args:
-        transforms: Squence of transform objects. The objects can be one of two
-        types:
-                1. torchcodec.transforms.DecoderTransform
-                2. torchvision.transforms.v2.Transform, but our type annotation
-                   only mentions its base, nn.Module. We don't want to take a
-                   hard dependency on TorchVision.
-
-    Returns:
-        List of DecoderTransform objects.
-    """
-    try:
-        from torchvision.transforms import v2
-
-        tv_available = True
-    except ImportError:
-        tv_available = False
-
-    converted_transforms: list[DecoderTransform] = []
-    for transform in transforms:
-        if not isinstance(transform, DecoderTransform):
-            if not tv_available:
-                raise ValueError(
-                    f"The supplied transform, {transform}, is not a TorchCodec "
-                    " DecoderTransform. TorchCodec also accept TorchVision "
-                    "v2 transforms, but TorchVision is not installed."
-                )
-            elif isinstance(transform, v2.Resize):
-                converted_transforms.append(Resize._from_torchvision(transform))
-            else:
-                raise ValueError(
-                    f"Unsupported transform: {transform}. Transforms must be "
-                    "either a TorchCodec DecoderTransform or a TorchVision "
-                    "v2 transform."
-                )
-        else:
-            converted_transforms.append(transform)
-
-    return converted_transforms
-
-
-def _make_transform_specs(
-    transforms: Optional[Sequence[Union[DecoderTransform, nn.Module]]],
-) -> str:
-    """Given a sequence of transforms, turn those into the specification string
-       the core API expects.
-
-    Args:
-        transforms: Optional sequence of transform objects. The objects can be
-            one of two types:
-                1. torchcodec.transforms.DecoderTransform
-                2. torchvision.transforms.v2.Transform, but our type annotation
-                   only mentions its base, nn.Module. We don't want to take a
-                   hard dependency on TorchVision.
-
-    Returns:
-        String of transforms in the format the core API expects: transform
-        specifications separate by semicolons.
-    """
-    if transforms is None:
-        return ""
-
-    transforms = _convert_to_decoder_transforms(transforms)
-    return ";".join([t._make_transform_spec() for t in transforms])
-
-
 def _read_custom_frame_mappings(
     custom_frame_mappings: Union[str, bytes, io.RawIOBase, io.BufferedReader]
 ) -> tuple[Tensor, Tensor, Tensor]: