search

m-bain / whisperX

WhisperX: Automatic Speech Recognition with Word-level Timestamps (& Diarization)
BSD 2-Clause "Simplified" License
11.73k stars 1.24k forks source link

Error While Using Machine With Only CPU (EC2 Instance) #800

Open anappdev opened 4 months ago

anappdev commented 4 months ago

Trying to install on EC2 Amazon Linux, Number of vCPUs= 4, instance type = t2.xlarge

It gives error Traceback (most recent call last): File "/home/ec2-user/xwhisper.py", line 6, in import whisperx File "/mnt/efs/fs1/ml/whisperx/whisperx/init.py", line 1, in from .transcribe import load_model File "/mnt/efs/fs1/ml/whisperx/whisperx/transcribe.py", line 9, in from .alignment import align, load_align_model File "/mnt/efs/fs1/ml/whisperx/whisperx/alignment.py", line 11, in import torchaudio File "/home/ec2-user/.local/lib/python3.9/site-packages/torchaudio/init.py", line 1, in from torchaudio import ( # noqa: F401 File "/home/ec2-user/.local/lib/python3.9/site-packages/torchaudio/_extension/init.py", line 43, in _load_lib("libtorchaudio") File "/home/ec2-user/.local/lib/python3.9/site-packages/torchaudio/_extension/utils.py", line 61, in _load_lib torch.ops.load_library(path) File "/home/ec2-user/.local/lib/python3.9/site-packages/torch/_ops.py", line 643, in load_library ctypes.CDLL(path) File "/usr/lib64/python3.9/ctypes/init.py", line 374, in init self._handle = _dlopen(self._name, mode) OSError: libtorch_cuda.so: cannot open shared object file: No such file or directory

Any support for CPU only environment?

SeeknnDestroy commented 4 months ago

Hey @anappdev,

Can you please try to use int8 as compute type?

import whisperx

model = whisperx.load_model(compute_type="int8")
anappdev commented 4 months ago

Hey @anappdev,

Can you please try to use int8 as compute type?

import whisperx

model = whisperx.load_model(compute_type="int8")

I used compute_type="int8", when I got that error

kepstin commented 4 months ago

This error looks like you have installed the "cpu only" version of torch (pytorch) but your torchaudio doesn't match, and is the gpu version. What method did you use to install torch and torchaudio? It seems to be pip from the file paths, but it would be good to confirm.

If you use pip, run

pip3 install torch torchaudio --index-url https://download.pytorch.org/whl/cpu

(add --force-reinstall if needed to replace the other version)