run Voxceleb Recipe [Speaker Recognition]

[TianyuCao]

When running Voxceleb Recipe [Speaker Recognition], I met the error as shown below. I am not sure where the codes in "runSnowdarXvector-extended-spec-am.py" wrong to make this type error. Thank you for your help!

(xmuspeech) tcao7@c06:~/kaldi/egs/xmuspeech/voxceleb1$ subtools/runPytorchLauncher.sh runSnowdarXvector-extended-spec-am.py --stage=0
Traceback (most recent call last):
File "runSnowdarXvector-extended-spec-am.py", line 282, in
utils.init_multi_gpu_training(args.gpu_id, args.multi_gpu_solution, args.port)
TypeError: init_multi_gpu_training() takes from 0 to 2 positional arguments but 3 were given

[sssyousen]

Sorry, it's a bug we didn't fixed. In the new version of ASV-Subtools, if using multi_gpu_training, it don't need the parameter "args.port" anymore, so you can delete the " args.port" and we will fixed this bug soon.

[TianyuCao]

Sorry to bother you again. I recently would like to running Voxceleb Recipe [Speaker Recognition]. After delete the " args.port", I ran subtools/runPytorchLauncher.sh runSnowdarXvector-extended-spec-am.py --stage=0, by using the mfcc23&pitch features in baidu pan. However, I found no folder called voxceleb1_train_aug_nosil in the path data/mfcc_23_pitch/voxceleb1_train_aug_nosil after unzipping the files and therefore, it failed to open script file feats.scp since the file data/mfcc_23_pitch/voxceleb1_train_aug_nosil/feats.scp does not exist. Any chance if you can upload the complete files again to baidu pan? Many thanks for your help in advance!!!

subtools/pytorch/pipeline/preprocess_to_egs.sh: stage 1
feat-to-len scp:data/mfcc_23_pitch/voxceleb1_train_aug_nosil/feats.scp ark,t:data/mfcc_23_pitch/voxceleb1_train_aug_nosil/utt2num_frames
WARNING (feat-to-len[5.5.10351-3dd90]:Open():util/kaldi-table-inl.h:106) Failed to open script file data/mfcc_23_pitch/voxceleb1_train_aug_nosil/feats.scp
ERROR (feat-to-len[5.5.10351-3dd90]:SequentialTableReader():util/kaldi-table-inl.h:860) Error constructing TableReader: rspecifier is scp:data/mfcc_23_pitch/voxceleb1_train_aug_nosil/feats.scp

Traceback (most recent call last):
File "subtools/pytorch/pipeline/onestep/get_chunk_egs.py", line 163, in main
get_chunk_egs(args)
File "subtools/pytorch/pipeline/onestep/get_chunk_egs.py", line 103, in get_chunk_egs
dataset = KaldiDataset.load_data_dir(args.data_dir)
File "subtools/pytorch/libs/egs/kaldi_dataset.py", line 83, in load_data_dir
return self(data_dir, expected_files)
File "subtools/pytorch/libs/egs/kaldi_dataset.py", line 77, in init
self.load_data_()
File "subtools/pytorch/libs/egs/kaldi_dataset.py", line 108, in load_data_
raise ValueError("The file {0} is not exist.".format(file_path))
ValueError: The file data/mfcc_23_pitch/voxceleb1_train_aug_nosil/feats.scp is not exist.
Traceback (most recent call last):
File "runSnowdarXvector-extended-spec-am.py", line 311, in
egs_dir=egs_dir))
File "subtools/pytorch/libs/support/kaldi_common.py", line 156, in execute_command
p.returncode, command))
Exception: Command exited with status 1: sh subtools/pytorch/pipeline/preprocess_to_egs.sh --stage 0 --endstage 3 --valid-split-type --total-spk --nj 20 --cmn true --limit-utts 8 --min-chunk 200 --overlap 0.1 --sample-type speaker_balance --chunk-num -1 --scale 1.5 --force-clear false --valid-num-utts 1024 --valid-chunk-num 2 data/mfcc_23_pitch/voxceleb1_train_aug exp/egs/mfcc_23_pitch_voxceleb1_train_aug_speaker_balance

[sssyousen]

The "_nosil" dir is generated in stage 0 by using feats.scp and vad.scp. You can check the stage 0 and make sure the content of feats.scp in data/mfcc_23_pitch/voxceleb1_train_aug are correct, expecially the absolute path of ark file.