fncokg
commented
3 weeks ago Generally speaking, this error is caused by a error in the linking libcuda.so.1 shared library during kaldi run, which may due to many reasons, for example, your CUDA Toolkit or GPU driver are not well installed.
However, I believe, in the context of MFA usage, this error can be solved by installing a CPU-built version of kaldi via conda install kaldi=*=*cpu*.
BTW, you can first have a look if you installed a GPU version of kaldi by conda list kaldi.
For Reproducing your issue Please fill out the following:
Corpus structure What language is the corpus in? Mandarin How many files/speakers? 4 Are you using lab files or TextGrid files for input? Textgrid files
Dictionary Are you using a dictionary from MFA? If so, which one?
From MFA mandarin_china_mfa.dict If it's a custom dictionary, what is the phoneset?
Acoustic model If you're using an acoustic model, is it one download through MFA? If so, which one? From MFA mandarin_mfa.zip If it's a model you've trained, what data was it trained on?
Log file Please attach the log file for the run that encountered an error (by default these will be stored in ).~/Documents/MFA from the log,it tells me: 2024-06-03 21:32:14,764 - mfa - DEBUG - Beginning run for 语料库 2024-06-03 21:32:14,764 - mfa - DEBUG - Using "global" profile 2024-06-03 21:32:14,764 - mfa - DEBUG - Using multiprocessing with 3 2024-06-03 21:32:14,764 - mfa - DEBUG - Set up logger for MFA version: 2.2.17 2024-06-03 21:32:14,764 - mfa - WARNING - Skipped the following configuration keys: min_language_model_weight, max_language_model_weight, and word_insertion_penalties 2024-06-03 21:32:16,228 - mfa - DEBUG - Using UNKNOWN 2024-06-03 21:32:25,266 - mfa - DEBUG - Loaded dictionary in 9.038 seconds 2024-06-03 21:32:25,274 - mfa - INFO - Setting up corpus information... 2024-06-03 21:32:25,275 - mfa - DEBUG - Could not load from temp 2024-06-03 21:32:25,275 - mfa - INFO - Loading corpus from source files... 2024-06-03 21:32:26,316 - mfa - DEBUG - Processing queue: 0.0568211019999989 2024-06-03 21:32:26,336 - mfa - DEBUG - Parsed corpus directory with 3 jobs in 0.07415009099999992 seconds 2024-06-03 21:32:26,347 - mfa - INFO - Found 1 speaker across 4 file, average number of utterances per speaker: 1.0 2024-06-03 21:32:26,348 - mfa - DEBUG - Loaded corpus in 1.082 seconds 2024-06-03 21:32:26,351 - mfa - INFO - Initializing multiprocessing jobs... 2024-06-03 21:32:26,352 - mfa - WARNING - Number of jobs was specified as 3, but due to only having 1 speakers, MFA will only use 1 jobs. Use the --single_speaker flag if you would like to split utterances across jobs regardless of their speaker. 2024-06-03 21:32:26,366 - mfa - DEBUG - Initialized jobs in 0.017 seconds 2024-06-03 21:32:26,368 - mfa - INFO - Normalizing text... 2024-06-03 21:32:27,997 - mfa - DEBUG - Wrote lexicon information in 0.120 seconds 2024-06-03 21:32:28,000 - mfa - INFO - Creating corpus split for feature generation... 2024-06-03 21:32:29,047 - mfa - DEBUG - Created corpus split directory in 1.050 seconds 2024-06-03 21:32:29,051 - mfa - INFO - Generating MFCCs... 2024-06-03 21:32:32,383 - mfa - ERROR - There was an error in the run, please see the log. and I see the make_mfcc.1.log,it says: /root/anaconda3/envs/aligner/bin/compute-mfcc-feats: error while loading shared libraries: libcuda.so.1: cannot open shared object file: No such file or directory /root/anaconda3/envs/aligner/bin/copy-feats: error while loading shared libraries: libcuda.so.1: cannot open shared object file: No such file or directory /root/anaconda3/envs/aligner/bin/compute-and-process-kaldi-pitch-feats: error while loading shared libraries: libcuda.so.1: cannot open shared object file: No such file or directory /root/anaconda3/envs/aligner/bin/copy-feats: error while loading shared libraries: libcuda.so.1: cannot open shared object file: No such file or directory Using: sqlite:////root/Documents/MFA/语料库/语料库.db
Desktop (please complete the following information):
OS: Linux Version centos7 Any other details about the setup (Cloud, Docker, etc): installed with mambda rather than conda Additional context Add any other context about the problem here. I use virtual machines for forced alignment on VMvare, and I use MFA with conda. Conda verion: 24.5.0 and MFA version: 2.2.17. Since my wav audio has no text, I decided to use the 'transcribe' command to transcribe, but with the above error, I don't know what to do. I hope someone can give me some advice, thank you! (Is it because the virtual machine on my VMvare cannot use the gpu?)