From 40bec21684ed293ecc113e11f3ad080ee37bbe83 Mon Sep 17 00:00:00 2001 From: yh_cc Date: Sat, 19 Sep 2020 11:10:36 +0800 Subject: [PATCH] =?UTF-8?q?=E4=BF=AE=E5=A4=8Dsave=5Fpath=E5=9C=A8dev?= =?UTF-8?q?=E4=B8=BA=E7=A9=BA=E7=9A=84=E6=97=B6=E5=80=99=E7=9A=84bug?= MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit --- fastNLP/core/losses.py | 6 +++--- fastNLP/core/trainer.py | 5 +++++ fastNLP/embeddings/bert_embedding.py | 4 ++-- fastNLP/embeddings/gpt2_embedding.py | 4 ++-- fastNLP/embeddings/roberta_embedding.py | 6 +++--- fastNLP/embeddings/static_embedding.py | 4 ++-- fastNLP/modules/generator/seq2seq_generator.py | 2 +- test/core/test_trainer.py | 13 ++++++++++++- 8 files changed, 30 insertions(+), 14 deletions(-) diff --git a/fastNLP/core/losses.py b/fastNLP/core/losses.py index afd2d083..a821d85a 100644 --- a/fastNLP/core/losses.py +++ b/fastNLP/core/losses.py @@ -178,7 +178,7 @@ class LossFunc(LossBase): r""" 提供给用户使用自定义损失函数的类 - :param func: 用户自行定义的损失函数，应当为一个函数或者callable(func)为True的ojbect + :param func: 用户自行定义的损失函数，应当为一个函数。 :param dict key_map: 参数映射表。键为Model/DataSet参数名，值为损失函数参数名。 fastNLP的trainer将在训练时从模型返回值或者训练数据DataSet的target=True的field中 找到相对应的参数名为value的参数，并传入func中作为参数名为key的参数 @@ -186,8 +186,8 @@ class LossFunc(LossBase): 使用方法:: - func = torch.nn.CrossEntropyLoss() - loss_func = LossFunc(func, input="pred", target="label") + import torch.nn.functional as F + loss_func = LossFunc(F.cross_entropy, input="pred", target="label") # 这表示构建了一个损失函数类，由func计算损失函数，其中将从模型返回值或者DataSet的target=True的field # 当中找到一个参数名为`pred`的参数传入func一个参数名为`input`的参数；找到一个参数名为`label`的参数 # 传入func作为一个名为`target`的参数 diff --git a/fastNLP/core/trainer.py b/fastNLP/core/trainer.py index e183632e..a562c58a 100644 --- a/fastNLP/core/trainer.py +++ b/fastNLP/core/trainer.py @@ -630,6 +630,11 @@ class Trainer(object): self.logger.info("Reloaded the best model.") else: self.logger.info("Fail to reload best model.") + + if self.dev_data is None and self.save_path is not None: + model_name = "_".join([self.model.__class__.__name__, self.start_time]) + self._save_model(self.model, model_name) + finally: if self.dev_data is not None and self.best_dev_perf is not None: self.logger.info( diff --git a/fastNLP/embeddings/bert_embedding.py b/fastNLP/embeddings/bert_embedding.py index 6ec5df99..e3b91934 100644 --- a/fastNLP/embeddings/bert_embedding.py +++ b/fastNLP/embeddings/bert_embedding.py @@ -89,7 +89,7 @@ class BertEmbedding(ContextualEmbedding): word pieces后的内容，并将第512个word piece置为[SEP]。超过长度的部分的encode结果直接全部置零。一般仅有只使用[CLS] 来进行分类的任务将auto_truncate置为True。 :param kwargs: - int min_freq: 小于该次数的词会被unk代替 + int min_freq: 小于该次数的词会被unk代替, 默认为1 """ super(BertEmbedding, self).__init__(vocab, word_dropout=word_dropout, dropout=dropout) @@ -110,7 +110,7 @@ class BertEmbedding(ContextualEmbedding): if '[CLS]' in vocab: self._word_cls_index = vocab['CLS'] - min_freq = kwargs.get('min_freq', 2) + min_freq = kwargs.get('min_freq', 1) self._min_freq = min_freq self.model = _BertWordModel(model_dir_or_name=model_dir_or_name, vocab=vocab, layers=layers, pool_method=pool_method, include_cls_sep=include_cls_sep, diff --git a/fastNLP/embeddings/gpt2_embedding.py b/fastNLP/embeddings/gpt2_embedding.py index 371d9e22..dc45b26b 100644 --- a/fastNLP/embeddings/gpt2_embedding.py +++ b/fastNLP/embeddings/gpt2_embedding.py @@ -83,7 +83,7 @@ class GPT2Embedding(ContextualEmbedding): only_use_pretrain_bpe = kwargs.get('only_use_pretrain_bpe', False) truncate_embed = kwargs.get('truncate_embed', True) - min_freq = kwargs.get('min_freq', 2) + min_freq = kwargs.get('min_freq', 1) self.lm_loss =language_model self.model = _GPT2Model(model_dir_or_name=model_dir_or_name, vocab=vocab, layers=layers, @@ -315,7 +315,7 @@ class GPT2WordPieceEncoder(nn.Module): class _GPT2Model(nn.Module): def __init__(self, model_dir_or_name, vocab, layers, pool_method='first', auto_truncate=True, language_model=False, - only_use_pretrain_bpe=False, min_freq=2, truncate_embed=False): + only_use_pretrain_bpe=False, min_freq=1, truncate_embed=False): super().__init__() self.tokenzier = GPT2Tokenizer.from_pretrained(model_dir_or_name) diff --git a/fastNLP/embeddings/roberta_embedding.py b/fastNLP/embeddings/roberta_embedding.py index 8e16c055..d81abaf5 100644 --- a/fastNLP/embeddings/roberta_embedding.py +++ b/fastNLP/embeddings/roberta_embedding.py @@ -78,7 +78,7 @@ class RobertaEmbedding(ContextualEmbedding): word pieces后的内容，并将第512个word piece置为。超过长度的部分的encode结果直接全部置零。一般仅有只使用 来进行分类的任务将auto_truncate置为True。 :param kwargs: - int min_freq: 小于该次数的词会被unk代替 + int min_freq: 小于该次数的词会被unk代替， 默认为1 """ super().__init__(vocab, word_dropout=word_dropout, dropout=dropout) @@ -93,7 +93,7 @@ class RobertaEmbedding(ContextualEmbedding): if '' in vocab: self._word_cls_index = vocab[''] - min_freq = kwargs.get('min_freq', 2) + min_freq = kwargs.get('min_freq', 1) self._min_freq = min_freq self.model = _RobertaWordModel(model_dir_or_name=model_dir_or_name, vocab=vocab, layers=layers, @@ -464,7 +464,7 @@ class RobertaWordPieceEncoder(nn.Module): os.makedirs(os.path.join(folder, ROBERTA_ENCODER_FOLDER), exist_ok=True) self.model.save(os.path.join(folder, ROBERTA_ENCODER_FOLDER)) - logger.debug(f"BertWordPieceEncoder has been saved in {folder}") + logger.debug(f"RobertaWordPieceEncoder has been saved in {folder}") @classmethod def load(cls, folder): diff --git a/fastNLP/embeddings/static_embedding.py b/fastNLP/embeddings/static_embedding.py index ef40f30d..a8139807 100644 --- a/fastNLP/embeddings/static_embedding.py +++ b/fastNLP/embeddings/static_embedding.py @@ -97,8 +97,8 @@ class StaticEmbedding(TokenEmbedding): :param int min_freq: Vocabulary词频数小于这个数量的word将被指向unk。 :param dict kwargs: bool only_train_min_freq: 仅对train中的词语使用min_freq筛选; - bool only_norm_found_vector: 是否仅对在预训练中找到的词语使用normalize; - bool only_use_pretrain_word: 仅使用出现在pretrain词表中的词，如果该词没有在预训练的词表中出现则为unk。如果embedding不需要更新建议设置为True。 + bool only_norm_found_vector: 默认为False, 是否仅对在预训练中找到的词语使用normalize; + bool only_use_pretrain_word: 默认为False, 仅使用出现在pretrain词表中的词，如果该词没有在预训练的词表中出现则为unk。如果embedding不需要更新建议设置为True。 """ super(StaticEmbedding, self).__init__(vocab, word_dropout=word_dropout, dropout=dropout) if embedding_dim > 0: diff --git a/fastNLP/modules/generator/seq2seq_generator.py b/fastNLP/modules/generator/seq2seq_generator.py index 8aa9eddc..cf6e0978 100644 --- a/fastNLP/modules/generator/seq2seq_generator.py +++ b/fastNLP/modules/generator/seq2seq_generator.py @@ -308,7 +308,7 @@ def _beam_search_generate(decoder: Seq2SeqDecoder, tokens=None, state=None, max_ max_len_eos_mask = max_lengths.eq(cur_len+1) eos_scores = scores[:, _eos_token_id] # 如果已经达到最大长度，就把eos的分数加大 - scores[:, _eos_token_id] = torch.where(max_len_eos_mask, eos_scores+100, eos_scores) + scores[:, _eos_token_id] = torch.where(max_len_eos_mask, eos_scores+1e12, eos_scores) if do_sample: if temperature > 0 and temperature != 1: diff --git a/test/core/test_trainer.py b/test/core/test_trainer.py index 37688bc2..01767738 100644 --- a/test/core/test_trainer.py +++ b/test/core/test_trainer.py @@ -76,8 +76,19 @@ class TrainerTestGround(unittest.TestCase): use_tqdm=True, check_code_level=2) trainer.train() import os + import shutil + self.assertTrue(os.path.exists(save_path)) + if os.path.exists(save_path): + shutil.rmtree(save_path) + + # 无dev_data的训练 + trainer = Trainer(train_set, model, optimizer=SGD(lr=0.1), loss=BCELoss(pred="predict", target="y"), + batch_size=32, n_epochs=10, print_every=50, dev_data=None, + metrics=None, validate_every=-1, save_path=save_path, + use_tqdm=True, check_code_level=2) + trainer.train() + self.assertTrue(os.path.exists(save_path)) if os.path.exists(save_path): - import shutil shutil.rmtree(save_path) def test_trainer_suggestion1(self):