fairseq
fairseq copied to clipboard
How to reproduce the speech data2vec result?
I follow the the instructions in page (https://github.com/facebookresearch/fairseq/tree/main/examples/data2vec) to reproduce the speech data2vec result.
And i have got the pretrained model.
Then start to use "fairseq-hydra-train
distributed_training.distributed_port=$PORT
task.data=/path/to/data
model.w2v_path=/path/to/model.pt
--config-dir /path/to/fairseq-py/examples/wav2vec/config/finetuning
--config-name base_100h common.user_dir=examples/data2vec" to finetune.
First of all , fairseq-hydra-train: error: unrecognized arguments: --common.user_dir=examples/data2vec
Then , I use "fairseq-hydra-train \
task.data=/path/to/data \
model.w2v_path=/path/to/model.pt \
--config-dir /path/to/fairseq-py/examples/wav2vec/config/finetuning \
--config-name base_100h" to finetune.
However, the error occurs, "File "/home/research/jiayang/data2vector/fairseq/fairseq/tasks/fairseq_task.py", line 338, in build_model
model = models.build_model(cfg, self, from_checkpoint)
File "/home/research/jiayang/data2vector/fairseq/fairseq/models/init.py", line 102, in build_model "Available models: {}".format(MODEL_DATACLASS_REGISTRY.keys()) KeyError: "'_name'""
It's really confusing. The pretrain model config setting is data2vec. Then the finetune model config is wav2vec?
Is there some mistakes? Looking for help
i've fixed same issue by adding
common.user_dir=examples/data2vec
after
task.data=/path/to/data
and before --config
try this one
fairseq-hydra-train
distributed_training.distributed_port=$PORT
task.data=/path/to/data
common.user_dir=examples/data2vec
model.w2v_path=/path/to/model.pt
--config-dir /path/to/fairseq-py/examples/wav2vec/config/finetuning
--config-name base_100h