ImportError: cannot import name 'metrics' from 'fairseq' (unknown location)

[ouwen18]

When I am training the decoder with audio, I encounter the following issue：

[2023-11-29 22:38:41,125][opensr.hubert_dataset][INFO] - pad_audio=True, random_crop=False, normalize=True, max_sample_size=500, seqs2seq data=True, [2023-11-29 22:38:41,125][opensr.hubert_dataset][INFO] - Noise wav: /usr/zzs/data/mvlrs_v1/noise/babble/train.tsv->1 wav, Prob: 0.25, SNR: 0, Number of mixture: 1 [2023-11-29 22:38:42,169][fairseq.logging.progress_bar][WARNING] - tensorboard not found, please install with: pip install tensorboard [2023-11-29 22:38:42,170][fairseq.trainer][INFO] - begin training epoch 1 [2023-11-29 22:38:42,171][fairseq_cli.train][INFO] - Start iterating over samples 0%| | 0/201 [00:00<?, ?it/s]Traceback (most recent call last): File "", line 1, in File "/home/aa/anaconda3/envs/openSR/lib/python3.8/multiprocessing/spawn.py", line 116, in spawn_main exitcode = _main(fd, parent_sentinel) File "/home/aa/anaconda3/envs/openSR/lib/python3.8/multiprocessing/spawn.py", line 125, in _main prepare(preparation_data) File "/home/aa/anaconda3/envs/openSR/lib/python3.8/multiprocessing/spawn.py", line 236, in prepare _fixup_main_from_path(data['init_main_from_path']) File "/home/aa/anaconda3/envs/openSR/lib/python3.8/multiprocessing/spawn.py", line 287, in _fixup_main_from_path main_content = runpy.run_path(main_path, File "/home/aa/anaconda3/envs/openSR/lib/python3.8/runpy.py", line 265, in run_path return _run_module_code(code, init_globals, run_name, File "/home/aa/anaconda3/envs/openSR/lib/python3.8/runpy.py", line 97, in _run_module_code _run_code(code, mod_globals, init_globals, File "/home/aa/anaconda3/envs/openSR/lib/python3.8/runpy.py", line 87, in _run_code exec(code, run_globals) File "/home/aa/anaconda3/envs/openSR/bin/fairseq-hydra-train", line 5, in from fairseq_cli.hydra_train import cli_main File "/home/aa/OpenSR/fairseq/fairseq_cli/hydra_train.py", line 11, in from fairseq_cli.train import main as pre_main File "/home/aa/OpenSR/fairseq/fairseq_cli/train.py", line 30, in from fairseq import ( File "/home/aa/OpenSR/fairseq/fairseq/checkpoint_utils.py", line 26, in from fairseq.models import FairseqDecoder, FairseqEncoder File "/home/aa/OpenSR/fairseq/fairseq/models/init.py", line 225, in import_models(models_dir, "fairseq.models") File "/home/aa/OpenSR/fairseq/fairseq/models/init.py", line 207, in import_models importlib.import_module(namespace + "." + model_name) File "/home/aa/anaconda3/envs/openSR/lib/python3.8/importlib/init.py", line 127, in import_module return _bootstrap._gcd_import(name[level:], package, level) File "/home/aa/OpenSR/fairseq/fairseq/models/wav2vec/init.py", line 6, in from .wav2vec import # noqa File "/home/aa/OpenSR/fairseq/fairseq/models/wav2vec/wav2vec.py", line 25, in from fairseq.tasks import FairseqTask File "/home/aa/OpenSR/fairseq/fairseq/tasks/init.py", line 15, in from .fairseq_task import FairseqTask, LegacyFairseqTask # noqa File "/home/aa/OpenSR/fairseq/fairseq/tasks/fairseq_task.py", line 13, in from fairseq import metrics, search, tokenizer, utils ImportError: cannot import name 'metrics' from 'fairseq' (unknown location) Traceback (most recent call last): File "", line 1, in File "/home/aa/anaconda3/envs/openSR/lib/python3.8/multiprocessing/spawn.py", line 116, in spawn_main exitcode = _main(fd, parent_sentinel) File "/home/aa/anaconda3/envs/openSR/lib/python3.8/multiprocessing/spawn.py", line 125, in _main prepare(preparation_data) File "/home/aa/anaconda3/envs/openSR/lib/python3.8/multiprocessing/spawn.py", line 236, in prepare _fixup_main_from_path(data['init_main_from_path']) File "/home/aa/anaconda3/envs/openSR/lib/python3.8/multiprocessing/spawn.py", line 287, in _fixup_main_from_path main_content = runpy.run_path(main_path, File "/home/aa/anaconda3/envs/openSR/lib/python3.8/runpy.py", line 265, in run_path return _run_module_code(code, init_globals, run_name, File "/home/aa/anaconda3/envs/openSR/lib/python3.8/runpy.py", line 97, in _run_module_code _run_code(code, mod_globals, init_globals, File "/home/aa/anaconda3/envs/openSR/lib/python3.8/runpy.py", line 87, in _run_code exec(code, run_globals) File "/home/aa/anaconda3/envs/openSR/bin/fairseq-hydra-train", line 5, in from fairseq_cli.hydra_train import cli_main File "/home/aa/OpenSR/fairseq/fairseq_cli/hydra_train.py", line 11, in from fairseq_cli.train import main as pre_main File "/home/aa/OpenSR/fairseq/fairseq_cli/train.py", line 30, in from fairseq import ( File "/home/aa/OpenSR/fairseq/fairseq/checkpoint_utils.py", line 26, in from fairseq.models import FairseqDecoder, FairseqEncoder File "/home/aa/OpenSR/fairseq/fairseq/models/init.py", line 225, in import_models(models_dir, "fairseq.models") File "/home/aa/OpenSR/fairseq/fairseq/models/init.py", line 207, in import_models importlib.import_module(namespace + "." + model_name) File "/home/aa/anaconda3/envs/openSR/lib/python3.8/importlib/init.py", line 127, in import_module return _bootstrap._gcd_import(name[level:], package, level) File "/home/aa/OpenSR/fairseq/fairseq/models/wav2vec/init.py", line 6, in from .wav2vec import # noqa File "/home/aa/OpenSR/fairseq/fairseq/models/wav2vec/wav2vec.py", line 25, in from fairseq.tasks import FairseqTask File "/home/aa/OpenSR/fairseq/fairseq/tasks/init.py", line 15, in from .fairseq_task import FairseqTask, LegacyFairseqTask # noqa File "/home/aa/OpenSR/fairseq/fairseq/tasks/fairseq_task.py", line 13, in from fairseq import metrics, search, tokenizer, utils ImportError: cannot import name 'metrics' from 'fairseq' (unknown location)

[Exgc]

The package fairseq should have been installed with the command:

cd fairseq pip install --editable ./

[ouwen18]

I confirm that I have followed your instructions to install, but the issue still persists.

[ouwen18]

Meanwhile, I noticed that in opensr/opensr_large_vox_audio.yaml, it indicates the need to input noise_wav, and it seems that this is not mentioned in the command line.

[Exgc]

Refer to: https://github.com/Exgc/OpenSR/tree/main/opensr/preparation#audio-noise-preparation-optional

[Exgc]

The problem with fairseq is supposed to be an issue when reproducing AV-Hubert, and I suggest you refer to its repository.

[ouwen18]

Thank you for ur suggestion. I will try to solve these problems