Ty Todd
Ty Todd
[Bug]: TypeError: AsyncConnectionPool.__init__() got an unexpected keyword argument 'socket_options'
I'm having the same issue. Any updates?
@lhoestq Almost finished, but I'm having trouble understanding this test case. This is how it looks originally. The `map` function is called, and then `with_format` is called. According to the...
@lhoestq ya that makes sense, looks like this functionality lives in `src/datasets/formatting`, where an exception is made for VideoReader objects to remain as themselves when being formatted. I'll make the...
@lhoestq I'm assuming this was also the case for torchaudio objects?
@lhoestq (sorry for the spam btw) Looks like there's a # TODO to have these returned as np.arrays instead. I'm curious why the authors didn't do it initially. Maybe a...
Oh cool ya this is something that I could implement with torchcodec. I can add that to the PR as well.
Hey @lhoestq I ran into an error with this test case for the Audio feature ``` @require_sndfile @require_torchcodec def test_dataset_with_audio_feature_map_is_decoded(shared_datadir): audio_path = str(shared_datadir / "test_audio_44100.wav") data = {"audio": [audio_path], "text":...
Yes from what I understand torchcodec supports everything ffmpeg supports.
Okay just finished. However, I wasn't able to pass this test case: ```python @require_torchcodec @require_sndfile @pytest.mark.parametrize("streaming", [False, True]) def test_load_dataset_with_audio_feature(streaming, jsonl_audio_dataset_path, shared_datadir): from torchcodec.decoders import AudioDecoder audio_path = str(shared_datadir /...
@lhoestq just submitted it #7616