Add torchcodec in docstrings/tests for datasets 4.0 (#39156)

* fix dataset run_object_detection

* bump version

* keep same dataset actually

* torchcodec in docstrings and testing utils

* torchcodec in dockerfiles and requirements

* remove duplicate

* add torchocodec to all the remaining docker files

* fix tests

* support torchcodec in audio classification and ASR

* [commit to revert] build ci-dev images

* [commit to revert] trigger circleci

* [commit to revert] build ci-dev images

* fix

* fix modeling_hubert

* backward compatible run_object_detection

* revert ci trigger commits

* fix mono conversion and support torch tensor as input

* revert map_to_array docs + fix it

* revert mono

* nit in docstring

* style

* fix modular

---------

Co-authored-by: ydshieh <ydshieh@users.noreply.github.com>

examples/pytorch/_tests_requirements.txt

@@ -22,6 +22,7 @@ protobuf
 torch
 torchvision
 torchaudio
+torchcodec
 jiwer
 librosa
 evaluate >= 0.2.0

examples/pytorch/object-detection/requirements.txt

@@ -1,5 +1,5 @@
 albumentations >= 1.4.16
 timm
-datasets
+datasets>=4.0
 torchmetrics
 pycocotools

examples/pytorch/object-detection/run_object_detection.py

@@ -399,7 +399,10 @@ def main():
         dataset["validation"] = split["test"]
 
     # Get dataset categories and prepare mappings for label_name <-> label_id
-    categories = dataset["train"].features["objects"].feature["category"].names
+    if isinstance(dataset["train"].features["objects"], dict):
+        categories = dataset["train"].features["objects"]["category"].feature.names
+    else:  # (for old versions of `datasets` that used Sequence({...}) of the objects)
+        categories = dataset["train"].features["objects"].feature["category"].names
     id2label = dict(enumerate(categories))
     label2id = {v: k for k, v in id2label.items()}
 

examples/pytorch/object-detection/run_object_detection_no_trainer.py

@@ -460,7 +460,10 @@ def main():
         dataset["validation"] = split["test"]
 
     # Get dataset categories and prepare mappings for label_name <-> label_id
-    categories = dataset["train"].features["objects"].feature["category"].names
+    if isinstance(dataset["train"].features["objects"], dict):
+        categories = dataset["train"].features["objects"]["category"].feature.names
+    else:  # (for old versions of `datasets` that used Sequence({...}) of the objects)
+        categories = dataset["train"].features["objects"].feature["category"].names
     id2label = dict(enumerate(categories))
     label2id = {v: k for k, v in id2label.items()}