diff --git a/requirements/test/cuda.in b/requirements/test/cuda.in index 5cf3a69e1f..30076ff707 100644 --- a/requirements/test/cuda.in +++ b/requirements/test/cuda.in @@ -21,7 +21,6 @@ vocos # required for minicpmo_26 test peft>=0.18.1 # required for phi-4-mm test pqdm ray[cgraph,default]>=2.48.0 # Ray Compiled Graph, required by pipeline parallelism tests -resampy # required for audio tests sentence-transformers>=5.2.0 # required for embedding tests soundfile # required for audio tests jiwer # required for audio tests diff --git a/requirements/test/cuda.txt b/requirements/test/cuda.txt index ed67685e6e..449939b965 100644 --- a/requirements/test/cuda.txt +++ b/requirements/test/cuda.txt @@ -555,7 +555,6 @@ numba==0.61.2 # -c requirements/cuda.txt # -r requirements/test/cuda.in # librosa - # resampy numpy==2.2.6 # via # -r requirements/test/cuda.in @@ -596,7 +595,6 @@ numpy==2.2.6 # pyogrio # pywavelets # rasterio - # resampy # rioxarray # rouge-score # runai-model-streamer @@ -1015,8 +1013,6 @@ requests==2.32.3 # tacoreader # tiktoken # wandb -resampy==0.4.3 - # via -r requirements/test/cuda.in responses==0.25.3 # via genai-perf rfc3339-validator==0.1.4 diff --git a/requirements/test/rocm.in b/requirements/test/rocm.in index dbb1500edc..0c9335240f 100644 --- a/requirements/test/rocm.in +++ b/requirements/test/rocm.in @@ -23,7 +23,6 @@ vocos # required for minicpmo_26 test peft>=0.15.0 # required for phi-4-mm test pqdm ray[cgraph,default]>=2.48.0 # Ray Compiled Graph, required by pipeline parallelism tests -resampy # required for audio tests sentence-transformers>=5.2.0 # required for embedding tests soundfile # required for audio tests jiwer # required for audio tests diff --git a/requirements/test/rocm.txt b/requirements/test/rocm.txt index ba9cd3dfdc..2e03df1bdc 100644 --- a/requirements/test/rocm.txt +++ b/requirements/test/rocm.txt @@ -662,7 +662,6 @@ numba==0.61.2 # -c requirements/rocm.txt # -r requirements/test/rocm.in # librosa - # resampy numkong==7.1.1 # via albucore numpy==2.2.6 @@ -708,7 +707,6 @@ numpy==2.2.6 # pytrec-eval-terrier # pywavelets # rasterio - # resampy # rioxarray # rouge-score # runai-model-streamer @@ -1193,8 +1191,6 @@ requests==2.32.5 # tacoreader # tiktoken # wandb -resampy==0.4.3 - # via -r requirements/test/rocm.in responses==0.26.0 # via genai-perf rfc3339-validator==0.1.4 diff --git a/setup.py b/setup.py index af52cb58df..37782062ae 100644 --- a/setup.py +++ b/setup.py @@ -1093,7 +1093,6 @@ setup( "runai": ["runai-model-streamer[s3,gcs,azure] >= 0.15.7"], "audio": [ "av", - "resampy", "scipy", "soundfile", "mistral_common[audio]", diff --git a/vllm/multimodal/audio.py b/vllm/multimodal/audio.py index 0a748a6d15..c232bb9ea6 100644 --- a/vllm/multimodal/audio.py +++ b/vllm/multimodal/audio.py @@ -16,11 +16,6 @@ try: except ImportError: av = PlaceholderModule("av") # type: ignore[assignment] -try: - import resampy -except ImportError: - resampy = PlaceholderModule("resampy") # type: ignore[assignment] - try: import scipy.signal as scipy_signal except ImportError: @@ -229,15 +224,6 @@ def resample_audio_pyav( return result[:expected_len] -def resample_audio_resampy( - audio: npt.NDArray[np.floating], - *, - orig_sr: float, - target_sr: float, -) -> npt.NDArray[np.floating]: - return resampy.resample(audio, sr_orig=orig_sr, sr_new=target_sr) - - def resample_audio_scipy( audio: npt.NDArray[np.floating], *, @@ -257,7 +243,7 @@ class AudioResampler: def __init__( self, target_sr: float | None = None, - method: Literal["pyav", "resampy", "scipy"] = "resampy", + method: Literal["pyav", "scipy"] = "pyav", ): self.target_sr = target_sr self.method = method @@ -281,10 +267,6 @@ class AudioResampler: return audio if self.method == "pyav": return resample_audio_pyav(audio, orig_sr=orig_sr, target_sr=self.target_sr) - if self.method == "resampy": - return resample_audio_resampy( - audio, orig_sr=orig_sr, target_sr=self.target_sr - ) elif self.method == "scipy": return resample_audio_scipy( audio, orig_sr=orig_sr, target_sr=self.target_sr diff --git a/vllm/multimodal/media/audio.py b/vllm/multimodal/media/audio.py index 4d41248f14..26d58d13a7 100644 --- a/vllm/multimodal/media/audio.py +++ b/vllm/multimodal/media/audio.py @@ -10,6 +10,7 @@ import pybase64 import torch from vllm.logger import init_logger +from vllm.multimodal.audio import resample_audio_pyav from vllm.utils.import_utils import PlaceholderModule from vllm.utils.serial_utils import tensor2base64 @@ -28,12 +29,6 @@ except ImportError: soundfile = PlaceholderModule("soundfile") # type: ignore[assignment] -try: - import resampy -except ImportError: - resampy = PlaceholderModule("resampy") # type: ignore[assignment] - - # Public libsndfile error codes exposed via `soundfile.LibsndfileError.code`, soundfile # being librosa's main backend. Used to validate if an audio loading error is due to a # server error vs a client error (invalid audio file). @@ -129,7 +124,7 @@ def load_audio_soundfile( y = np.mean(y, axis=tuple(range(y.ndim - 1))) if sr is not None and sr != native_sr: - y = resampy.resample(y, sr_orig=native_sr, sr_new=sr) + y = resample_audio_pyav(y, orig_sr=native_sr, target_sr=sr) return y, int(sr) return y, native_sr