Merge branch 'dev0.8.0' of github.com:fastnlp/fastNLP into dev0.8.0

fastNLP/core/drivers/paddle_driver/fleet.py

@@ -19,7 +19,7 @@ from fastNLP.core.utils import (
     paddle_move_data_to_device,
     is_in_paddle_dist,
 )
-from fastNLP.core.samplers import ReproducibleIterator, RandomSampler, UnrepeatedDistributedSampler
+from fastNLP.core.samplers import ReproducibleIterator, RandomSampler, UnrepeatedSampler
 from fastNLP.envs.env import FASTNLP_DISTRIBUTED_CHECK, USER_CUDA_VISIBLE_DEVICES
 from fastNLP.core.log import logger
 
@@ -362,7 +362,7 @@ class PaddleFleetDriver(PaddleDriver):
                 return dataloader
         # evaluator
         elif dist == "unrepeatdist":
-            sampler = UnrepeatedDistributedSampler(
+            sampler = UnrepeatedSampler(
                 dataset=dataloader.dataset,
                 shuffle=shuffle,
                 seed=int(os.environ.get("FASTNLP_SEED", 0))

fastNLP/core/drivers/torch_driver/ddp.py

@@ -23,11 +23,12 @@ from fastNLP.core.drivers.torch_driver.utils import (
     ForwardState,
     _MODE_PARAMETER,
     reset_seed,
-    replace_sampler
+    replace_sampler,
+    replace_batch_sampler
 )
 from fastNLP.core.drivers.utils import distributed_open_proc
 from fastNLP.core.utils import auto_param_call, check_user_specific_params
-from fastNLP.core.samplers import ReproducibleIterator, RandomSampler, UnrepeatedDistributedSampler
+from fastNLP.core.samplers import ReproducibleIterator, RandomSampler, UnrepeatedSampler, ReproducibleBatchSampler
 from fastNLP.envs import FASTNLP_DISTRIBUTED_CHECK, FASTNLP_GLOBAL_RANK, FASTNLP_GLOBAL_SEED
 from fastNLP.core.log import logger
 from fastNLP.core.drivers.torch_driver.dist_utils import fastnlp_torch_all_gather, fastnlp_torch_broadcast_object
@@ -445,11 +446,25 @@ class TorchDDPDriver(TorchDriver):
         # return self.model(batch, **{_MODE_PARAMETER: ForwardState.TEST})
         return self._test_step(batch)
 
-    def set_dist_repro_dataloader(self, dataloader, dist: Optional[Union[str, ReproducibleIterator]],
-                                  reproducible: bool = False, sampler_or_batch_sampler=None):
+    def set_dist_repro_dataloader(self, dataloader, dist: Optional[Union[str, ReproducibleIterator, ReproducibleBatchSampler]]=None,
+                                  reproducible: bool = False):
+        if isinstance(dist, ReproducibleBatchSampler):
+            dist = re_instantiate_sampler(dist)
+            dist.set_distributed(
+                num_replicas=self.world_size,
+                rank=self.global_rank,
+                pad=True
+            )
+            return replace_batch_sampler(dataloader, dist)
+
         if isinstance(dist, ReproducibleIterator):
             # 注意这里不需要调用 dist_sampler.set_distributed；因为如果用户使用的是 TorchDDPDriver，那么其在 Trainer 初始化的时候就已经调用了该函数；
             dist = re_instantiate_sampler(dist)
+            dist.set_distributed(
+                num_replicas=self.world_size,
+                rank=self.global_rank,
+                pad=True
+            )
             return replace_sampler(dataloader, dist)
 
         # trainer, evaluator
@@ -463,7 +478,15 @@ class TorchDDPDriver(TorchDriver):
         elif dist == "dist":
             args = self.get_dataloader_args(dataloader)
             # 如果用户的 trainer.use_dist_sampler 为 True，那么此时其是否进行断点重训，不影响这里的行为；
-            if isinstance(args.sampler, ReproducibleIterator):
+            if isinstance(args.batch_sampler, ReproducibleBatchSampler):
+                batch_sampler = re_instantiate_sampler(args.batch_sampler)
+                batch_sampler.set_distributed(
+                    num_replicas=self.world_size,
+                    rank=self.global_rank,
+                    pad=True
+                )
+                return replace_batch_sampler(dataloader, batch_sampler)
+            elif isinstance(args.sampler, ReproducibleIterator):
                 sampler = re_instantiate_sampler(args.sampler)
                 sampler.set_distributed(
                     num_replicas=self.world_size,
@@ -477,7 +500,6 @@ class TorchDDPDriver(TorchDriver):
                     shuffle=args.shuffle,
                     seed=int(os.environ.get(FASTNLP_GLOBAL_SEED, 0))
                 )
-                # todo 这个你写个todo吧，有两个角度；第一个是dataloader即使检测到sampler是我们reproducible，也不能直接set_distributeds; 第二个如果是单卡的，也需要替换sampler乃至切换sampler的状态，方式之前多卡，现在切换成单卡运行
                 sampler.set_distributed(
                     num_replicas=self.world_size,
                     rank=self.global_rank,
@@ -487,8 +509,11 @@ class TorchDDPDriver(TorchDriver):
 
         # evaluator
         elif dist == "unrepeatdist":
+            # todo @yh，补充 unrepeatdist 相关内容；
             args = self.get_dataloader_args(dataloader)
-            sampler = UnrepeatedDistributedSampler(
+
+            # todo 判断 batch_sampler；
+            sampler = UnrepeatedSampler(
                 dataset=args.dataset,
                 shuffle=args.shuffle,
             )

fastNLP/core/drivers/torch_driver/single_device.py

@@ -133,8 +133,10 @@ class TorchSingleDriver(TorchDriver):
     def set_dist_repro_dataloader(self, dataloader, dist: Union[str, ReproducibleBatchSampler, ReproducibleIterator]=None,
                                   reproducible: bool = False):
         if isinstance(dist, ReproducibleBatchSampler):
+            dist = re_instantiate_sampler(dist)
             return replace_batch_sampler(dataloader, dist)
         elif isinstance(dist, ReproducibleIterator):
+            dist = re_instantiate_sampler(dist)
             return replace_sampler(dataloader, dist)
 
         if reproducible:

fastNLP/core/drivers/torch_driver/torch_driver.py

@@ -244,8 +244,34 @@ class TorchDriver(Driver):
                 logger.debug("Load model.")
 
         # 3. 恢复 sampler 的状态；
+        """
+        使用场景：
+        
+        现在sampler/batch_sampler的替换情况：
+        1. 单卡多卡；
+        2. 是否断点重训；
+        
+        3. 用户通过 dist 传入；
+        4. 用户自己直接在外面替换dataloader的sampler或者 batchsampler；
+        
+        应当确定的规则：
+        batchsampler 优先级高于 sampler；
+        
+        单卡：
+            不是断点重训：
+                用户自己
+        
+        
+        用户不自己在外面直接替换 sampler 或者 batchsampler
+        1. 单卡：
+        
+        """
         dataloader_args = self.get_dataloader_args(dataloader)
 
+        # todo 先捋一下；
+        # batch_sampler = dataloader_args.batch_sampler
+        # if not (hasattr(batch_sampler, 'load_state_dict') and callable(batch_sampler.load_state_dict)):
+
         sampler = dataloader_args.sampler
         if not (hasattr(sampler, 'load_state_dict') and callable(sampler.load_state_dict)):
             # 说明这里需要使用 ReproduceSampler 来弄一下了

fastNLP/core/samplers/__init__.py

@@ -3,19 +3,24 @@ __all__ = [
     'SortedSampler',
     'ConstTokenNumSampler',
     'ConstantTokenNumSampler',
-    'UnrepeatedDistributedSampler',
+
     'MixSampler',
-    'InnerSampler',
     'DopedSampler',
     'MixSequentialSampler',
     'PollingSampler',
+
     'ReproducibleIterator',
     'RandomSampler',
-    're_instantiate_sampler'
+
+    're_instantiate_sampler',
+
+    'UnrepeatedSampler',
+    "UnrepeatedSortedSampler"
 ]
 
-from .sampler import BucketSampler, SortedSampler, ConstTokenNumSampler, ConstantTokenNumSampler, UnrepeatedDistributedSampler
-from .mix_sampler import MixSampler, InnerSampler, DopedSampler, MixSequentialSampler, PollingSampler
+from .sampler import BucketSampler, SortedSampler, ConstTokenNumSampler, ConstantTokenNumSampler
+from .unrepeated_sampler import UnrepeatedSampler, UnrepeatedSortedSampler
+from .mix_sampler import MixSampler, DopedSampler, MixSequentialSampler, PollingSampler
 from .reproducible_sampler import ReproducibleIterator, RandomSampler, re_instantiate_sampler
 from .reproducible_batch_sampler import ReproducibleBatchSampler, BucketedBatchSampler
 

fastNLP/core/samplers/mix_sampler.py

@@ -4,7 +4,6 @@ from typing import Union, List, Iterable, Dict
 
 __all__ = [
     'MixSampler',
-    'InnerSampler',
     'DopedSampler',
     'MixSequentialSampler',
     'PollingSampler'

fastNLP/core/samplers/reproducible_sampler.py

@@ -16,7 +16,6 @@ def re_instantiate_sampler(sampler):
     return type(sampler)(**all_attributes)
 
 
-
 class ReproducibleIterator:
     """
     注意所有继承 `ReproducibleIterator` 的类的  `__init__` 方法中都需要加入参数 `**kwargs`，用来使我们再断点重训时重新实例化这个 sampler

fastNLP/core/samplers/sampler.py

@@ -7,7 +7,6 @@ __all__ = [
     "SortedSampler",
     'ConstTokenNumSampler',
     "ConstantTokenNumSampler",
-    "UnrepeatedDistributedSampler",
 ]
 
 from itertools import chain
@@ -18,7 +17,7 @@ import numpy as np
 from fastNLP.envs.imports import _NEED_IMPORT_TORCH
 
 if _NEED_IMPORT_TORCH:
-    from torch.utils.data import SequentialSampler, Sampler, RandomSampler
+    from torch.utils.data import Sampler
 else:
     from fastNLP.core.utils.dummy_class import DummyClass as Sampler
 
@@ -727,87 +726,3 @@ def k_means_bucketing(lengths, buckets):
         if buckets[bucket_id] is None or lengths[idx] <= buckets[bucket_id]:
             bucket_data[bucket_id].append(idx)
     return bucket_data
-
-
-class UnrepeatedDistributedSampler:
-    def __init__(self, dataset, shuffle: bool = False, seed: int = 0):
-        """
-        考虑在多卡evaluate的场景下，不能重复sample。
-
-        :param dataset:
-        :param shuffle:
-        :param seed:
-        """
-        self.dataset = dataset
-        self.shuffle = shuffle
-        self.seed = seed
-
-        # 多卡的相关的参数
-        self.num_replicas = 1
-        self.rank = 0
-        self.epoch = -1
-
-    def __len__(self):
-        """
-        返回 sampler 一次完整的迭代过程会产生多少个index。多卡的情况下，只考虑当前rank；
-        :return:
-        """
-        num_common = len(self.dataset)//self.num_replicas
-        self.num_samples = num_common + int(self.rank < (len(self.dataset)-num_common*self.num_replicas))
-        return self.num_samples
-
-    def __iter__(self):
-        r"""
-        当前使用num_consumed_samples做法会在交替使用的时候遇到问题；
-        Example:
-            >>> sampler = RandomSampler()
-            >>> iter1 = iter(sampler)
-            >>> iter2 = iter(sampler)
-            >>> next(iter1)
-            >>> next(iter2)  # 当前num_consumed_samples的数量会发生变化
-        """
-
-        indices = self.generate_indices()
-
-        # subsample
-        indices = indices[self.rank:len(indices):self.num_replicas]
-        assert len(indices) == len(self)
-
-        for index in indices:
-            yield index
-
-    def generate_indices(self) -> List[int]:
-        """
-        生成随机序列
-
-        :return:
-        """
-        if self.shuffle:
-            indices = list(range(len(self.dataset)))
-            seed = self.seed + self.epoch
-            rng = np.random.default_rng(abs(seed))
-            rng.shuffle(indices)
-            if self.epoch < 0:  # 防止用户忘记调用 set_epoch，至少这样可以保证每次epoch出来的index顺序不同。
-                self.epoch -= 1
-        else:
-            indices = list(range(len(self.dataset)))
-        return indices
-
-    def set_epoch(self, epoch: int) -> None:
-        self.epoch = epoch
-
-    def set_distributed(self, num_replicas, rank):
-        """
-        该方法本质上等同于 ddp 情形下的没有完成的初始化，应当在初始化该 sampler 本身后立即被调用；
-
-        :param num_replicas:
-        :param rank:
-        :return:
-        """
-        assert num_replicas>0 and isinstance(num_replicas, int)
-        assert isinstance(rank, int) and 0<=rank<num_replicas
-        # 注意初始化该函数时，所有的状态都应当默认是一个 epoch 刚开始训练的状态；
-        self.num_replicas = num_replicas
-        self.rank = rank
-
-        return self

fastNLP/core/samplers/unrepeated_sampler.py

@@ -0,0 +1,114 @@
+__all__ = [
+    'UnrepeatedSortedSampler',
+    'UnrepeatedSampler'
+]
+
+from typing import List, Union
+from fastNLP.core.dataset import DataSet
+
+import numpy as np
+
+
+class UnrepeatedSampler:
+    def __init__(self, dataset, shuffle: bool = False, seed: int = 0, **kwargs):
+        """
+        考虑在多卡evaluate的场景下，不能重复sample。
+
+        :param dataset:
+        :param shuffle:
+        :param seed:
+        """
+        self.dataset = dataset
+        self.shuffle = shuffle
+        self.seed = seed
+
+        # 多卡的相关的参数
+        self.num_replicas = kwargs.get('num_replicas', 1)
+        self.rank = kwargs.get('rank', 0)
+        self.epoch = kwargs.get('epoch', -1)
+
+    def __len__(self):
+        """
+        返回 sampler 一次完整的迭代过程会产生多少个index。多卡的情况下，只考虑当前rank；
+        :return:
+        """
+        num_common = len(self.dataset)//self.num_replicas
+        self.num_samples = num_common + int(self.rank < (len(self.dataset)-num_common*self.num_replicas))
+        return self.num_samples
+
+    def __iter__(self):
+        indices = self.generate_indices()
+
+        # subsample
+        indices = indices[self.rank:len(indices):self.num_replicas]
+        assert len(indices) == len(self)
+
+        for index in indices:
+            yield index
+
+    def generate_indices(self) -> List[int]:
+        """
+        生成随机序列
+
+        :return:
+        """
+        if self.shuffle:
+            indices = list(range(len(self.dataset)))
+            seed = self.seed + self.epoch
+            rng = np.random.default_rng(abs(seed))
+            rng.shuffle(indices)
+            if self.epoch < 0:  # 防止用户忘记调用 set_epoch，至少这样可以保证每次epoch出来的index顺序不同。
+                self.epoch -= 1
+        else:
+            indices = list(range(len(self.dataset)))
+        return indices
+
+    def set_epoch(self, epoch: int) -> None:
+        self.epoch = epoch
+
+    def set_distributed(self, num_replicas, rank):
+        """
+        该方法本质上等同于 ddp 情形下的没有完成的初始化，应当在初始化该 sampler 本身后立即被调用；
+
+        :param num_replicas:
+        :param rank:
+        :return:
+        """
+        assert num_replicas>0 and isinstance(num_replicas, int)
+        assert isinstance(rank, int) and 0<=rank<num_replicas
+        # 注意初始化该函数时，所有的状态都应当默认是一个 epoch 刚开始训练的状态；
+        self.num_replicas = num_replicas
+        self.rank = rank
+
+        return self
+
+
+class UnrepeatedSortedSampler(UnrepeatedSampler):
+    def __init__(self, dataset, length:Union[str, List], seed: int = 0):
+        """
+        将 dataset 中的数据根据 length 从长到短进行迭代，并且保证在多卡场景下数据不重复。本 sampler 可能导致各个机器上的
+            batch 数量不完全一致。
+
+        :param dataset: 实现了 __len__ 方法的数据容器。
+        :param length: 如果为 List，应当与 dataset 有一样的长度，表示 dataset 中每个元素的数量；仅当传入的 dataset 为 fastNLP 的
+            DataSet 时支持传入 str，会将该str理解为 dataset 的 field 名称，若 field 中的元素为 int，则认为该值是 sample 的长度。
+        :param shuffle: 如果为 True，将不进行 shuffle，实际上数据会以从长到短的方式输出。
+        :param seed: 设置的随机数种子
+        :param kwargs: fastNLP 保留使用
+        """
+        super().__init__(dataset=dataset, shuffle=False, seed=seed)
+        if isinstance(dataset, DataSet):
+            length = dataset.get_field(length)
+            if not isinstance(length[0], int):
+                length = list(map(len, length))
+        else:
+            assert len(length) == len(dataset), "When the dataset is not fastNLP.DataSet, " \
+                                                "the length parameter can only be List[int]"
+
+        assert len(length) == len(dataset), "The length of `data` and `length` should be equal."
+
+        self.length = np.array(length, dtype=int)  # 按照长到短排列的序号。
+        self.sorted_indices = np.argsort(self.length)[::-1].tolist()  # 按长度从高到低排序的
+
+    def generate_indices(self) -> List[int]:
+        return self.sorted_indices

tests/core/samplers/test_unrepeated_sampler.py

@@ -0,0 +1,64 @@
+from itertools import chain
+
+import pytest
+
+from fastNLP.core.samplers import UnrepeatedSampler, UnrepeatedSortedSampler
+
+
+class DatasetWithVaryLength:
+    def __init__(self, num_of_data=100):
+        self.data = list(range(num_of_data))
+
+    def __getitem__(self, item):
+        return self.data[item]
+
+    def __len__(self):
+        return len(self.data)
+
+
+class TestUnrepeatedSampler:
+    @pytest.mark.parametrize('shuffle', [True, False])
+    def test_single(self, shuffle):
+        num_of_data = 100
+        data = DatasetWithVaryLength(num_of_data)
+        sampler = UnrepeatedSampler(data, shuffle)
+        indexes = set(sampler)
+        assert indexes==set(range(num_of_data))
+
+    @pytest.mark.parametrize('num_replica', [2, 3])
+    @pytest.mark.parametrize('num_of_data', [2, 3, 4, 100])
+    @pytest.mark.parametrize('shuffle', [False, True])
+    def test_multi(self, num_replica, num_of_data, shuffle):
+        data = DatasetWithVaryLength(num_of_data=num_of_data)
+        samplers = []
+        for i in range(num_replica):
+            sampler = UnrepeatedSampler(dataset=data, shuffle=shuffle)
+            sampler.set_distributed(num_replica, rank=i)
+            samplers.append(sampler)
+
+        indexes = set(chain(*samplers))
+        assert indexes==set(range(num_of_data))
+
+
+class TestUnrepeatedSortedSampler:
+    @pytest.mark.parametrize('shuffle', [True, False])
+    def test_single(self, shuffle):
+        num_of_data = 100
+        data = DatasetWithVaryLength(num_of_data)
+        sampler = UnrepeatedSortedSampler(data, length=data.data)
+        indexes = list(sampler)
+        assert indexes==list(range(num_of_data-1, -1, -1))
+
+    @pytest.mark.parametrize('num_replica', [2, 3])
+    @pytest.mark.parametrize('num_of_data', [2, 3, 4, 100])
+    @pytest.mark.parametrize('shuffle', [False, True])
+    def test_multi(self, num_replica, num_of_data, shuffle):
+        data = DatasetWithVaryLength(num_of_data=num_of_data)
+        samplers = []
+        for i in range(num_replica):
+            sampler = UnrepeatedSortedSampler(dataset=data, length=data.data)
+            sampler.set_distributed(num_replica, rank=i)
+            samplers.append(sampler)
+
+        indexes = set(chain(*samplers))
+        assert indexes==set(range(num_of_data))