Merge pull request #687 from alibaba-damo-academy/dev_lhn

hnluo

2023-06-29 c2dee5e3c29eba79e591d9e9caebaef15ea4e56b

 funasr/tasks/abs_task.py

@@ -30,6 +30,7 @@
import torch.nn
import torch.optim
import yaml
from funasr.models.base_model import FunASRModel
from torch.utils.data import DataLoader
from typeguard import check_argument_types
from typeguard import check_return_type
@@ -44,19 +45,18 @@
from funasr.iterators.multiple_iter_factory import MultipleIterFactory
from funasr.iterators.sequence_iter_factory import SequenceIterFactory
from funasr.main_funcs.collect_stats import collect_stats
from funasr.optimizers.sgd import SGD
from funasr.optimizers.fairseq_adam import FairseqAdam
from funasr.optimizers.sgd import SGD
from funasr.samplers.build_batch_sampler import BATCH_TYPES
from funasr.samplers.build_batch_sampler import build_batch_sampler
from funasr.samplers.unsorted_batch_sampler import UnsortedBatchSampler
from funasr.schedulers.noam_lr import NoamLR
from funasr.schedulers.warmup_lr import WarmupLR
from funasr.schedulers.tri_stage_scheduler import TriStageLR
from funasr.schedulers.warmup_lr import WarmupLR
from funasr.torch_utils.load_pretrained_model import load_pretrained_model
from funasr.torch_utils.model_summary import model_summary
from funasr.torch_utils.pytorch_version import pytorch_cudnn_version
from funasr.torch_utils.set_all_random_seed import set_all_random_seed
from funasr.train.abs_espnet_model import AbsESPnetModel
from funasr.train.class_choices import ClassChoices
from funasr.train.distributed_utils import DistributedOption
from funasr.train.trainer import Trainer
@@ -230,8 +230,8 @@
        >>> cls.check_task_requirements()
        If your model is defined as following,

        >>> from funasr.train.abs_espnet_model import AbsESPnetModel
        >>> class Model(AbsESPnetModel):
        >>> from funasr.models.base_model import FunASRModel
        >>> class Model(FunASRModel):
        ...     def forward(self, input, output, opt=None):  pass

        then "required_data_names" should be as
@@ -251,8 +251,8 @@
        >>> cls.check_task_requirements()
        If your model is defined as follows,

        >>> from funasr.train.abs_espnet_model import AbsESPnetModel
        >>> class Model(AbsESPnetModel):
        >>> from funasr.models.base_model import FunASRModel
        >>> class Model(FunASRModel):
        ...     def forward(self, input, output, opt=None):  pass

        then "optional_data_names" should be as
@@ -263,8 +263,9 @@

    @classmethod
    @abstractmethod
    def build_model(cls, args: argparse.Namespace) -> AbsESPnetModel:
    def build_model(cls, args: argparse.Namespace) -> FunASRModel:
        raise NotImplementedError


    @classmethod
    def get_parser(cls) -> config_argparse.ArgumentParser:
@@ -445,6 +446,12 @@
            help='Perform on "collect stats" mode',
        )
        group.add_argument(
            "--mc",
            type=bool,
            default=False,
            help="MultiChannel input",
        )
        group.add_argument(
            "--write_collected_feats",
            type=str2bool,
            default=False,
@@ -463,6 +470,12 @@
            type=int,
            default=sys.maxsize,
            help="The maximum number update step to train",
        )
        parser.add_argument(
            "--batch_interval",
            type=int,
            default=-1,
            help="The batch interval for saving model.",
        )
        group.add_argument(
            "--patience",
@@ -541,6 +554,12 @@
            type=int,
            default=1,
            help="The number of gradient accumulation",
        )
        group.add_argument(
            "--bias_grad_times",
            type=float,
            default=1.0,
            help="To scale the gradient of contextual related params",
        )
        group.add_argument(
            "--no_forward_run",
@@ -629,8 +648,8 @@
        group.add_argument(
            "--init_param",
            type=str,
            action="append",
            default=[],
            nargs="*",
            help="Specify the file path used for initialization of parameters. "
                 "The format is '<file_path>:<src_key>:<dst_key>:<exclude_keys>', "
                 "where file_path is the model file path, "
@@ -656,7 +675,7 @@
            "--freeze_param",
            type=str,
            default=[],
            nargs="*",
            action="append",
            help="Freeze parameters",
        )

@@ -1147,14 +1166,15 @@
        elif args.distributed and args.simple_ddp:
            distributed_option.init_torch_distributed_pai(args)
            args.ngpu = dist.get_world_size()
            if args.dataset_type == "small":
            if args.dataset_type == "small" and args.ngpu > 0:
                if args.batch_size is not None:
                    args.batch_size = args.batch_size * args.ngpu
                if args.batch_bins is not None:
                if args.batch_bins is not None and args.ngpu > 0:
                    args.batch_bins = args.batch_bins * args.ngpu

        # filter samples if wav.scp and text are mismatch
        if (args.train_shape_file is None and args.dataset_type == "small") or args.train_data_file is None and args.dataset_type == "large":
        if (
                args.train_shape_file is None and args.dataset_type == "small") or args.train_data_file is None and args.dataset_type == "large":
            if not args.simple_ddp or distributed_option.dist_rank == 0:
                filter_wav_text(args.data_dir, args.train_set)
                filter_wav_text(args.data_dir, args.dev_set)
@@ -1163,8 +1183,10 @@

        if args.train_shape_file is None and args.dataset_type == "small":
            if not args.simple_ddp or distributed_option.dist_rank == 0:
                calc_shape(args.data_dir, args.train_set, args.frontend_conf, args.speech_length_min, args.speech_length_max)
                calc_shape(args.data_dir, args.dev_set, args.frontend_conf, args.speech_length_min, args.speech_length_max)
                calc_shape(args.data_dir, args.train_set, args.frontend_conf, args.speech_length_min,
                           args.speech_length_max)
                calc_shape(args.data_dir, args.dev_set, args.frontend_conf, args.speech_length_min,
                           args.speech_length_max)
            if args.simple_ddp:
                dist.barrier()
            args.train_shape_file = [os.path.join(args.data_dir, args.train_set, "speech_shape")]
@@ -1193,12 +1215,18 @@
            # logging.basicConfig() is invoked in main_worker() instead of main()
            # because it can be invoked only once in a process.
            # FIXME(kamo): Should we use logging.getLogger()?
            # BUGFIX: Remove previous handlers and reset log level
            for handler in logging.root.handlers[:]:
                logging.root.removeHandler(handler)
            logging.basicConfig(
                level=args.log_level,
                format=f"[{os.uname()[1].split('.')[0]}]"
                       f" %(asctime)s (%(module)s:%(lineno)d) %(levelname)s: %(message)s",
            )
        else:
            # BUGFIX: Remove previous handlers and reset log level
            for handler in logging.root.handlers[:]:
                logging.root.removeHandler(handler)
            # Suppress logging if RANK != 0
            logging.basicConfig(
                level="ERROR",
@@ -1220,9 +1248,9 @@

        # 2. Build model
        model = cls.build_model(args=args)
        if not isinstance(model, AbsESPnetModel):
        if not isinstance(model, FunASRModel):
            raise RuntimeError(
                f"model must inherit {AbsESPnetModel.__name__}, but got {type(model)}"
                f"model must inherit {FunASRModel.__name__}, but got {type(model)}"
            )
        model = model.to(
            dtype=getattr(torch, args.train_dtype),
@@ -1304,6 +1332,7 @@
                    data_path_and_name_and_type=args.train_data_path_and_name_and_type,
                    key_file=train_key_file,
                    batch_size=args.batch_size,
                    mc=args.mc,
                    dtype=args.train_dtype,
                    num_workers=args.num_workers,
                    allow_variable_data_keys=args.allow_variable_data_keys,
@@ -1315,6 +1344,7 @@
                    data_path_and_name_and_type=args.valid_data_path_and_name_and_type,
                    key_file=valid_key_file,
                    batch_size=args.valid_batch_size,
                    mc=args.mc,
                    dtype=args.train_dtype,
                    num_workers=args.num_workers,
                    allow_variable_data_keys=args.allow_variable_data_keys,
@@ -1346,19 +1376,10 @@

            # 7. Build iterator factories
            if args.dataset_type == "large":
                from funasr.datasets.large_datasets.build_dataloader import ArkDataLoader
                train_iter_factory = ArkDataLoader(args.train_data_file, args.token_list, args.dataset_conf,
                                                   frontend_conf=args.frontend_conf if hasattr(args, "frontend_conf") else None,
                                                   seg_dict_file=args.seg_dict_file if hasattr(args,
                                                                                               "seg_dict_file") else None,
                                                   punc_dict_file=args.punc_list if hasattr(args, "punc_list") else None,
                                                   mode="train")
                valid_iter_factory = ArkDataLoader(args.valid_data_file, args.token_list, args.dataset_conf, 
                                                   frontend_conf=args.frontend_conf if hasattr(args, "frontend_conf") else None,
                                                   seg_dict_file=args.seg_dict_file if hasattr(args,
                                                                                               "seg_dict_file") else None,
                                                   punc_dict_file=args.punc_list if hasattr(args, "punc_list") else None,
                                                   mode="eval")
                from funasr.datasets.large_datasets.build_dataloader import LargeDataLoader
                train_iter_factory = LargeDataLoader(args, mode="train")
                valid_iter_factory = LargeDataLoader(args, mode="eval")

            elif args.dataset_type == "small":
                train_iter_factory = cls.build_iter_factory(
                    args=args,
@@ -1570,13 +1591,21 @@
    ) -> AbsIterFactory:
        assert check_argument_types()

        if hasattr(args, "frontend_conf"):
            if args.frontend_conf is not None and "fs" in args.frontend_conf:
                dest_sample_rate = args.frontend_conf["fs"]
            else:
                dest_sample_rate = 16000
        else:
            dest_sample_rate = 16000

        dataset = ESPnetDataset(
            iter_options.data_path_and_name_and_type,
            float_dtype=args.train_dtype,
            preprocess=iter_options.preprocess_fn,
            max_cache_size=iter_options.max_cache_size,
            max_cache_fd=iter_options.max_cache_fd,
            dest_sample_rate=args.frontend_conf["fs"],
            dest_sample_rate=dest_sample_rate,
        )
        cls.check_task_requirements(
            dataset, args.allow_variable_data_keys, train=iter_options.train
@@ -1895,7 +1924,7 @@
            model_file: Union[Path, str] = None,
            cmvn_file: Union[Path, str] = None,
            device: str = "cpu",
    ) -> Tuple[AbsESPnetModel, argparse.Namespace]:
    ) -> Tuple[FunASRModel, argparse.Namespace]:
        """Build model from the files.

        This method is used for inference or fine-tuning.
@@ -1922,9 +1951,9 @@
            args["cmvn_file"] = cmvn_file
        args = argparse.Namespace(**args)
        model = cls.build_model(args)
        if not isinstance(model, AbsESPnetModel):
        if not isinstance(model, FunASRModel):
            raise RuntimeError(
                f"model must inherit {AbsESPnetModel.__name__}, but got {type(model)}"
                f"model must inherit {FunASRModel.__name__}, but got {type(model)}"
            )
        model.to(device)
        if model_file is not None: