diff --git a/docs/source/figures/fitlogChart.png b/docs/source/figures/fitlogChart.png new file mode 100644 index 00000000..57ae1683 Binary files /dev/null and b/docs/source/figures/fitlogChart.png differ diff --git a/docs/source/figures/fitlogTable.png b/docs/source/figures/fitlogTable.png new file mode 100644 index 00000000..37551634 Binary files /dev/null and b/docs/source/figures/fitlogTable.png differ diff --git a/docs/source/index.rst b/docs/source/index.rst index 219e32f9..03a192dc 100644 --- a/docs/source/index.rst +++ b/docs/source/index.rst @@ -55,6 +55,7 @@ fastNLP 在 :mod:`~fastNLP.models` 模块中内置了如 :class:`~fastNLP.models 安装指南 快速入门 详细指南 + 科研指南 API 文档 ------------- diff --git a/docs/source/user/with_fitlog.rst b/docs/source/user/with_fitlog.rst index 97c3ea71..14cf7d91 100644 --- a/docs/source/user/with_fitlog.rst +++ b/docs/source/user/with_fitlog.rst @@ -2,4 +2,121 @@ 科研向导 ================= -本文介绍使用 fastNLP 和 fitlog 进行科学研究的方法 \ No newline at end of file +本文介绍使用 fastNLP 和 fitlog 结合进行科研的方法。 + +首先，我们需要安装 `fitlog `_ 。你需要确认你的电脑中没有其它名为为 `fitlog` 的命令。 + +我们从命令行中进入到一个文件夹，现在我们要在文件夹中创建我们的 fastNLP 项目。你可以在命令行输入 `fitlog init test1` ， +然后你会看到如下提示:: + + Initialized empty Git repository in /Users/fdujyn/workspaces/test1/.git/ + Auto commit by fitlog + Initialized empty Git repository in /Users/fdujyn/workspaces/test1/.git/ + Fitlog project test1 is initialized. + +这表明你已经创建成功了项目文件夹，并且在项目文件夹中已经初始化了 Git。如果你不想初始化 Git， +可以参考文档 `命令行工具 `_ + +现在我们进入你创建的项目文件夹 test1 中，可以看到有一个名为 logs 的文件夹，后面我们将会在里面存放你的实验记录。 +同时也有一个名为 main.py 的文件，是我们推荐你使用的入口文件。文件的内容如下:: + + import fitlog + + fitlog.commit(__file__) # auto commit your codes + fitlog.add_hyper_in_file (__file__) # record your hyperparameters + + """ + Your training code here, you may use these functions to log your result: + fitlog.add_hyper() + fitlog.add_loss() + fitlog.add_metric() + fitlog.add_best_metric() + ...... + """ + + fitlog.finish() # finish the logging + +我们推荐你保留除注释外的四行代码，它们有助于你的实验， +他们的具体用处参见文档 `用户 API `_ + +我们假定你要进行前两个教程中的实验，并已经把数据复制到了项目根目录下的 tutorial_sample_dataset.csv 文件中。 +现在我们编写如下的训练代码，使用 :class:`~fastNLP.core.callback.FitlogCallback` 进行实验记录保存:: + + import fitlog + from fastNLP import Vocabulary, Trainer, CrossEntropyLoss, AccuracyMetric + from fastNLP.io import CSVLoader + from fastNLP.models import CNNText + from fastNLP.core.callback import FitlogCallback + + fitlog.commit(__file__) # auto commit your codes + fitlog.add_hyper_in_file (__file__) # record your hyperparameters + + ############hyper + word_embed = 50 + dropout = 0.1 + ############hyper + + loader = CSVLoader(headers=('raw_sentence', 'label'), sep='\t') + dataset = loader.load("tutorial_sample_dataset.csv") + + dataset.apply(lambda x: x['raw_sentence'].lower(), new_field_name='sentence') + dataset.apply(lambda x: x['sentence'].split(), new_field_name='words', is_input=True) + dataset.apply(lambda x: int(x['label']), new_field_name='target', is_target=True) + vocab = Vocabulary(min_freq=2).from_dataset(dataset, field_name='words') + vocab.index_dataset(dataset, field_name='words',new_field_name='words') + + model = CNNText((len(vocab),word_embed), num_classes=5, padding=2, dropout=dropout) + + train_dev_data, test_data = dataset.split(0.1) + train_data, dev_data = train_dev_data.split(0.1) + + trainer = Trainer(model=model, train_data=train_data, dev_data=dev_data, + loss=CrossEntropyLoss(), metrics=AccuracyMetric(), + callbacks=[FitlogCallback(test_data)]) + trainer.train() + + fitlog.finish() # finish the logging + +用命令行在项目目录下执行 `python main.py` 之后，输出结果如下:: + + Auto commit by fitlog + input fields after batch(if batch size is 2): + words: (1)type:torch.Tensor (2)dtype:torch.int64, (3)shape:torch.Size([2, 11]) + target fields after batch(if batch size is 2): + target: (1)type:torch.Tensor (2)dtype:torch.int64, (3)shape:torch.Size([2]) + + training epochs started 2019-05-23-21-11-51 + Evaluation at Epoch 1/10. Step:2/20. AccuracyMetric: acc=0.285714 + + Evaluation at Epoch 2/10. Step:4/20. AccuracyMetric: acc=0.285714 + + Evaluation at Epoch 3/10. Step:6/20. AccuracyMetric: acc=0.285714 + + Evaluation at Epoch 4/10. Step:8/20. AccuracyMetric: acc=0.428571 + + Evaluation at Epoch 5/10. Step:10/20. AccuracyMetric: acc=0.571429 + + Evaluation at Epoch 6/10. Step:12/20. AccuracyMetric: acc=0.571429 + + Evaluation at Epoch 7/10. Step:14/20. AccuracyMetric: acc=0.285714 + + Evaluation at Epoch 8/10. Step:16/20. AccuracyMetric: acc=0.142857 + + Evaluation at Epoch 9/10. Step:18/20. AccuracyMetric: acc=0.285714 + + Evaluation at Epoch 10/10. Step:20/20. AccuracyMetric: acc=0.571429 + + + In Epoch:5/Step:10, got best dev performance:AccuracyMetric: acc=0.571429 + Reloaded the best model. + +现在，我们在项目目录下输入 `fitlog log logs` ，命令行会启动一个网页，默认 url 为 ``0.0.0.0:5000`` 。 +我们在浏览器中打开网页，可以看到如下的统计表格: + +.. image:: ../figures/fitlogTable.png + +如果我们点击action中的最后一个键钮，可以看到详细的 loss 图: + +.. image:: ../figures/fitlogChart.png + +更多的教程还在编写中，敬请期待~ \ No newline at end of file diff --git a/fastNLP/core/callback.py b/fastNLP/core/callback.py index 7fad2d0b..07654cc8 100644 --- a/fastNLP/core/callback.py +++ b/fastNLP/core/callback.py @@ -54,6 +54,7 @@ __all__ = [ "GradientClipCallback", "EarlyStopCallback", "TensorboardCallback", + "FitlogCallback", "LRScheduler", "ControlC", @@ -65,6 +66,7 @@ import os import torch from copy import deepcopy + try: from tensorboardX import SummaryWriter @@ -81,6 +83,7 @@ try: except: pass + class Callback(object): """ 别名：:class:`fastNLP.Callback` :class:`fastNLP.core.callback.Callback` @@ -431,14 +434,13 @@ class EarlyStopCallback(Callback): else: raise exception # 抛出陌生Error + class FitlogCallback(Callback): """ - 别名: :class:`fastNLP.FitlogCallback` :class:`fastNLP.core.callback.FitlogCallback` - 该callback将loss和progress自动写入到fitlog中; 如果Trainer有dev的数据，将自动把dev的结果写入到log中; 同时还支持传入 - 一个(或多个)test数据集进行测试(只有在trainer具有dev时才能使用)，每次在dev上evaluate之后会在这些数据集上验证一下。 - 并将验证结果写入到fitlog中。这些数据集的结果是根据dev上最好的结果报道的，即如果dev在第3个epoch取得了最佳，则 - fitlog中记录的关于这些数据集的结果就是来自第三个epoch的结果。 + 一个(或多个)test数据集进行测试(只有在trainer具有dev时才能使用)，每次在dev上evaluate之后会在这些数据集上验证一下。 + 并将验证结果写入到fitlog中。这些数据集的结果是根据dev上最好的结果报道的，即如果dev在第3个epoch取得了最佳，则 + fitlog中记录的关于这些数据集的结果就是来自第三个epoch的结果。 :param DataSet,dict(DataSet) data: 传入DataSet对象，会使用多个Trainer中的metric对数据进行验证。如果需要传入多个 DataSet请通过dict的方式传入，dict的key将作为对应dataset的name传递给fitlog。若tester不为None时，data需要通过 @@ -447,7 +449,9 @@ class FitlogCallback(Callback): :param int verbose: 是否在终端打印内容，0不打印 :param bool log_exception: fitlog是否记录发生的exception信息 """ - + # 还没有被导出到 fastNLP 层 + # 别名: :class:`fastNLP.FitlogCallback` :class:`fastNLP.core.callback.FitlogCallback` + def __init__(self, data=None, tester=None, verbose=0, log_exception=False): super().__init__() self.datasets = {} @@ -460,7 +464,7 @@ class FitlogCallback(Callback): assert 'test' not in data, "Cannot use `test` as DataSet key, when tester is passed." setattr(tester, 'verbose', 0) self.testers['test'] = tester - + if isinstance(data, dict): for key, value in data.items(): assert isinstance(value, DataSet), f"Only DataSet object is allowed, not {type(value)}." @@ -470,23 +474,23 @@ class FitlogCallback(Callback): self.datasets['test'] = data else: raise TypeError("data receives dict[DataSet] or DataSet object.") - + self.verbose = verbose - + def on_train_begin(self): - if (len(self.datasets)>0 or len(self.testers)>0 ) and self.trainer.dev_data is None: + if (len(self.datasets) > 0 or len(self.testers) > 0) and self.trainer.dev_data is None: raise RuntimeError("Trainer has no dev data, you cannot pass extra data to do evaluation.") - - if len(self.datasets)>0: + + if len(self.datasets) > 0: for key, data in self.datasets.items(): tester = Tester(data=data, model=self.model, batch_size=self.batch_size, metrics=self.trainer.metrics, verbose=0) self.testers[key] = tester fitlog.add_progress(total_steps=self.n_steps) - + def on_backward_begin(self, loss): fitlog.add_loss(loss.item(), name='loss', step=self.step, epoch=self.epoch) - + def on_valid_end(self, eval_result, metric_key, optimizer, better_result): if better_result: eval_result = deepcopy(eval_result) @@ -494,11 +498,11 @@ class FitlogCallback(Callback): eval_result['epoch'] = self.epoch fitlog.add_best_metric(eval_result) fitlog.add_metric(eval_result, step=self.step, epoch=self.epoch) - if len(self.testers)>0: + if len(self.testers) > 0: for key, tester in self.testers.items(): try: eval_result = tester.test() - if self.verbose!=0: + if self.verbose != 0: self.pbar.write("Evaluation on DataSet {}:".format(key)) self.pbar.write(tester._format_eval_results(eval_result)) fitlog.add_metric(eval_result, name=key, step=self.step, epoch=self.epoch) @@ -506,10 +510,10 @@ class FitlogCallback(Callback): fitlog.add_best_metric(eval_result, name=key) except Exception: self.pbar.write("Exception happens when evaluate on DataSet named `{}`.".format(key)) - + def on_train_end(self): fitlog.finish() - + def on_exception(self, exception): fitlog.finish(status=1) if self._log_exception: