fix: specify test dependencies required for CI (#634)

Author: deependujha

requirements/test.txt

@@ -1,13 +1,19 @@
+av >=14.0.0
 coverage ==7.8.*
 cryptography==42.0.8
 mosaicml-streaming==0.11.0
+torchaudio >=2.7.0
 pytest ==8.4.*
 pytest-cov ==6.1.1
 pytest-timeout ==2.3.1
 pytest-rerunfailures ==14.0
 pytest-random-order ==1.1.1
 pandas
+pyarrow >=20.0.0
+polars >1.0.0
 lightning
 transformers <4.53.0
 zstd
 numpy < 2.0
+s5cmd >=0.2.0
+soundfile >=0.13.0 # required for torchaudio backend

tests/processing/test_data_processor.py

@@ -3,6 +3,7 @@
 import random
 import sys
 from functools import partial
+from io import BytesIO
 from queue import Empty
 from typing import Any, List
 from unittest import mock
@@ -1146,6 +1147,8 @@ def test_load_torch_audio(tmpdir, compression):
 
     import torchaudio
 
+    torchaudio.set_audio_backend("soundfile")
+
     optimize(
         fn=create_synthetic_audio_bytes,
         inputs=list(range(100)),
@@ -1157,9 +1160,11 @@ def test_load_torch_audio(tmpdir, compression):
 
     dataset = StreamingDataset(input_dir=str(tmpdir))
     sample = dataset[0]
-    tensor = torchaudio.load(sample["content"])
-    assert tensor[0].shape == torch.Size([1, 16000])
-    assert tensor[1] == 16000
+    buffer = BytesIO(sample["content"])
+    buffer.seek(0)
+    tensor, sample_rate = torchaudio.load(buffer, format="wav")
+    assert tensor.shape == torch.Size([1, 16000])
+    assert sample_rate == 16000
 
 
 def create_synthetic_audio_file(filepath) -> dict: