更新Trainer:

* 添加Trainer参数metric_key，指明用来做模型选择的指标的名字
* 在Trainer添加处理tester返回的评价指标的逻辑，选择当前最好的模型

fastNLP/core/trainer.py

@@ -1,39 +1,38 @@
-import itertools
 import os
 import time
 import warnings
-from collections import defaultdict
 from datetime import datetime
 from datetime import timedelta
 
 import torch
-from torch import nn
 from tensorboardX import SummaryWriter
+from torch import nn
 
 from fastNLP.core.batch import Batch
+from fastNLP.core.dataset import DataSet
+from fastNLP.core.losses import _prepare_losser
+from fastNLP.core.metrics import _prepare_metrics
 from fastNLP.core.optimizer import Optimizer
 from fastNLP.core.sampler import RandomSampler
 from fastNLP.core.sampler import SequentialSampler
 from fastNLP.core.tester import Tester
+from fastNLP.core.utils import CheckError
 from fastNLP.core.utils import _build_args
 from fastNLP.core.utils import _check_arg_dict_list
 from fastNLP.core.utils import _move_dict_value_to_device
 from fastNLP.core.utils import get_func_signature
-from fastNLP.core.dataset import DataSet
 
-from fastNLP.core.losses import LossBase
-from fastNLP.core.metrics import MetricBase
-from fastNLP.core.losses import _prepare_losser
-from fastNLP.core.metrics import _prepare_metrics
-from fastNLP.core.utils import CheckError
 
 class Trainer(object):
     """Main Training Loop
 
     """
-    def __init__(self, train_data, model, losser=None, metrics=None, n_epochs=3, batch_size=32, print_every=-1, validate_every=-1,
+
+    def __init__(self, train_data, model, losser=None, metrics=None, n_epochs=3, batch_size=32, print_every=-1,
+                 validate_every=-1,
                  dev_data=None, use_cuda=False, save_path="./save",
                  optimizer=Optimizer("Adam", lr=0.01, weight_decay=0), need_check_code=True,
+                 metric_key=None,
                  **kwargs):
         super(Trainer, self).__init__()
 
@@ -50,6 +49,13 @@ class Trainer(object):
 
         # prepare evaluate
         metrics = _prepare_metrics(metrics)
+
+        # parse metric_key
+        # increase_better is True. It means the exp result gets better if the indicator increases.
+        # It is true by default.
+        self.increase_better = False if metric_key[0] == "-" else True
+        self.metric_key = metric_key[1:] if metric_key[0] == "+" or metric_key[0] == "-" else metric_key
+
         # prepare loss
         losser = _prepare_losser(losser)
 
@@ -67,12 +73,10 @@ class Trainer(object):
         self.save_path = save_path
         self.print_every = int(print_every)
         self.validate_every = int(validate_every)
-        self._best_accuracy = 0
+        self.best_metric_indicator = None
 
         self._model_device = model.parameters().__next__().device
 
-        # TODO self._best_accuracy不能表现出当前的metric多种的情况
-
         if isinstance(optimizer, torch.optim.Optimizer):
             self.optimizer = optimizer
         else:
@@ -102,7 +106,7 @@ class Trainer(object):
             if torch.cuda.is_available() and self.use_cuda:
                 self.model = self.model.cuda()
 
-            self.mode(self.model, is_test=False)
+            self._mode(self.model, is_test=False)
 
             start = time.time()
             self.start_time = str(datetime.now().strftime('%Y-%m-%d-%H-%M-%S'))
@@ -112,7 +116,9 @@ class Trainer(object):
                     def __getattr__(self, item):
                         def pass_func(*args, **kwargs):
                             pass
+
                         return pass_func
+
                 self._summary_writer = psudoSW()
             else:
                 path = os.path.join(self.save_path, 'tensorboard_logs_{}'.format(self.start_time))
@@ -121,19 +127,20 @@ class Trainer(object):
             epoch = 1
             while epoch <= self.n_epochs:
 
-                data_iterator = Batch(self.train_data, batch_size=self.batch_size, sampler=RandomSampler(), as_numpy=False)
+                data_iterator = Batch(self.train_data, batch_size=self.batch_size, sampler=RandomSampler(),
+                                      as_numpy=False)
 
-                self._train_epoch(data_iterator, self.model, epoch, self.dev_data, start)
+                self._train_epoch(data_iterator, self.model, epoch, start)
 
                 # validate_every override validation at end of epochs
                 if self.dev_data and self.validate_every <= 0:
-                    self.do_validation()
+                    self._do_validation()
                 epoch += 1
         finally:
             self._summary_writer.close()
             del self._summary_writer
 
-    def _train_epoch(self, data_iterator, model, epoch, dev_data, start, **kwargs):
+    def _train_epoch(self, data_iterator, model, epoch, start):
         """Training process in one epoch.
 
             kwargs should contain:
@@ -144,10 +151,10 @@ class Trainer(object):
         for batch_x, batch_y in data_iterator:
             # TODO 这里可能会遇到问题，万一用户在model内部修改了prediction的device就会有问题
             _move_dict_value_to_device(self._model_device, batch_x, batch_y)
-            prediction = self.data_forward(model, batch_x)
-            loss = self.get_loss(prediction, batch_y)
-            self.grad_backward(loss)
-            self.update()
+            prediction = self._data_forward(model, batch_x)
+            loss = self._compute_loss(prediction, batch_y)
+            self._grad_backward(loss)
+            self._update()
             self._summary_writer.add_scalar("loss", loss.item(), global_step=self.step)
             for name, param in self.model.named_parameters():
                 if param.requires_grad:
@@ -162,18 +169,19 @@ class Trainer(object):
                 print(print_output)
 
             if self.validate_every > 0 and self.step % self.validate_every == 0:
-                self.do_validation()
+                self._do_validation()
 
             self.step += 1
 
-    def do_validation(self):
+    def _do_validation(self):
         res = self.tester.test()
         for name, num in res.items():
             self._summary_writer.add_scalar("valid_{}".format(name), num, global_step=self.step)
-        if self.save_path is not None and self.best_eval_result(res):
-            self.save_model(self.model, 'best_model_' + self.start_time)
+        if self.save_path is not None and self._better_eval_result(res):
+            self.save_model(self.model,
+                            "best_" + "_".join([self.model.__class__.__name__, self.metric_key, self.start_time]))
 
-    def mode(self, model, is_test=False):
+    def _mode(self, model, is_test=False):
         """Train mode or Test mode. This is for PyTorch currently.
 
         :param model: a PyTorch model
@@ -185,20 +193,20 @@ class Trainer(object):
         else:
             model.train()
 
-    def update(self):
+    def _update(self):
         """Perform weight update on a model.
 
         """
         self.optimizer.step()
 
-    def data_forward(self, network, x):
+    def _data_forward(self, network, x):
         x = _build_args(network.forward, **x)
         y = network(**x)
         if not isinstance(y, dict):
             raise TypeError(f"The return value of {get_func_signature(network.forward)} should be dict, got {type(y)}.")
         return y
 
-    def grad_backward(self, loss):
+    def _grad_backward(self, loss):
         """Compute gradient with link rules.
 
         :param loss: a scalar where back-prop starts
@@ -208,7 +216,7 @@ class Trainer(object):
         self.model.zero_grad()
         loss.backward()
 
-    def get_loss(self, predict, truth):
+    def _compute_loss(self, predict, truth):
         """Compute loss given prediction and ground truth.
 
         :param predict: prediction dict, produced by model.forward
@@ -224,27 +232,52 @@ class Trainer(object):
         else:
             torch.save(model, model_name)
 
-    def best_eval_result(self, metrics):
+    def _better_eval_result(self, metrics):
         """Check if the current epoch yields better validation results.
 
-        :return: bool, True means current results on dev set is the best.
+        :return bool value: True means current results on dev set is the best.
         """
         if isinstance(metrics, tuple):
             loss, metrics = metrics
 
         if isinstance(metrics, dict):
             if len(metrics) == 1:
-                accuracy = list(metrics.values())[0]
+                # only single metric, just use it
+                metric_dict = list(metrics.values())[0]
+                metrics_name = list(metrics.keys())[0]
             else:
-                accuracy = metrics[self.eval_sort_key]
-        else:
-            accuracy = metrics
-
-        if accuracy > self._best_accuracy:
-            self._best_accuracy = accuracy
-            return True
-        else:
-            return False
+                metrics_name = self.metrics[0].__class__.__name__
+                if metrics_name not in metrics:
+                    raise RuntimeError(f"{metrics_name} is chosen to do validation, but got {metrics}")
+                metric_dict = metrics[metrics_name]
+
+            if len(metric_dict) == 1:
+                indicator_val, indicator = list(metric_dict.values())[0], list(metric_dict.keys())[0]
+            elif len(metric_dict) > 1 and self.metric_key is None:
+                raise RuntimeError(
+                    f"Got multiple metric keys: {metric_dict}, but metric_key is not set. Which one to use?")
+            else:
+                # metric_key is set
+                if self.metric_key not in metric_dict:
+                    raise RuntimeError(f"matric key {self.metric_key} not found in {metric_dict}")
+                indicator_val = metric_dict[self.metric_key]
+
+            is_better = True
+            if self.best_metric_indicator is None:
+                # first-time validation
+                self.best_metric_indicator = indicator_val
+            else:
+                if self.increase_better is True:
+                    if indicator_val > self.best_metric_indicator:
+                        self.best_metric_indicator = indicator_val
+                    else:
+                        is_better = False
+                else:
+                    if indicator_val < self.best_metric_indicator:
+                        self.best_metric_indicator = indicator_val
+                    else:
+                        is_better = False
+            return is_better
 
 
 DEFAULT_CHECK_BATCH_SIZE = 2
@@ -254,6 +287,7 @@ IGNORE_CHECK_LEVEL = 0
 WARNING_CHECK_LEVEL = 1
 STRICT_CHECK_LEVEL = 2
 
+
 def _check_code(dataset, model, losser, metrics, batch_size=DEFAULT_CHECK_BATCH_SIZE,
                 dev_data=None,
                 check_level=WARNING_CHECK_LEVEL):
@@ -264,7 +298,7 @@ def _check_code(dataset, model, losser, metrics, batch_size=DEFAULT_CHECK_BATCH_
     for batch_count, (batch_x, batch_y) in enumerate(batch):
         _move_dict_value_to_device(model_devcie, batch_x, batch_y)
         # forward check
-        if batch_count==0:
+        if batch_count == 0:
             _check_forward_error(model_func=model.forward, check_level=check_level,
                                  batch_x=batch_x)
 
@@ -285,17 +319,17 @@ def _check_code(dataset, model, losser, metrics, batch_size=DEFAULT_CHECK_BATCH_
         if batch_count == 0:
             if not isinstance(loss, torch.Tensor):
                 raise TypeError(f"The return value of {get_func_signature(losser.__call__)} should be `torch.Tensor`, "
-                                 f"but got `{type(loss)}`.")
-            if len(loss.size())!=0:
+                                f"but got `{type(loss)}`.")
+            if len(loss.size()) != 0:
                 raise ValueError(f"The size of return value of {get_func_signature(losser.__call__)} is {loss.size()}, "
                                  f"should be torch.size([])")
         loss.backward()
         model.zero_grad()
-        if batch_count+1>=DEFAULT_CHECK_NUM_BATCH:
+        if batch_count + 1 >= DEFAULT_CHECK_NUM_BATCH:
             break
 
     if dev_data is not None:
-        tester = Tester(data=dataset[:batch_size*DEFAULT_CHECK_NUM_BATCH], model=model, metrics=metrics,
+        tester = Tester(data=dataset[:batch_size * DEFAULT_CHECK_NUM_BATCH], model=model, metrics=metrics,
                         batch_size=batch_size, verbose=-1)
         tester.test()
 
@@ -305,18 +339,18 @@ def _check_forward_error(model_func, check_level, batch_x):
     _missing = ''
     _unused = ''
     func_signature = get_func_signature(model_func)
-    if len(check_res['missing'])!=0:
+    if len(check_res['missing']) != 0:
         _missing = "Function {} misses {}, only provided with {}, " \
                    ".\n".format(func_signature, check_res.missing,
-                                             list(batch_x.keys()))
-    if len(check_res['unused'])!=0:
+                                list(batch_x.keys()))
+    if len(check_res['unused']) != 0:
         if len(check_res.unused) > 1:
             _unused = "{} are not used ".format(check_res.unused)
         else:
             _unused = "{} is not used ".format(check_res.unused)
         _unused += "in function {}.\n".format(func_signature)
     if _missing:
-        if len(_unused)>0 and STRICT_CHECK_LEVEL:
+        if len(_unused) > 0 and STRICT_CHECK_LEVEL:
             _error_str = "(1).{}\n(2).{}".format(_missing, _unused)
         else:
             _error_str = _missing
@@ -329,38 +363,40 @@ def _check_forward_error(model_func, check_level, batch_x):
         elif check_level == WARNING_CHECK_LEVEL:
             warnings.warn(message=_unused)
 
-def _check_loss_evaluate(prev_func, func, check_res, output, batch_y, check_level):
+
+def _check_loss_evaluate(prev_func, func, check_level, output, batch_y):
+    check_res = _check_arg_dict_list(func, [output, batch_y])
     _missing = ''
     _unused = ''
     _duplicated = ''
     func_signature = get_func_signature(func)
     prev_func_signature = get_func_signature(prev_func)
-    if len(check_res.missing)>0:
+    if len(check_res.missing) > 0:
         _missing = "function {} misses argument {}, \n\t only provided with {}(from {}) and " \
                    "{}(from target in Dataset)." \
-                   .format(func_signature, check_res.missing,
-                            list(output.keys()), prev_func_signature,
-                           list(batch_y.keys()))
-    if len(check_res.unused)>0:
+            .format(func_signature, check_res.missing,
+                    list(output.keys()), prev_func_signature,
+                    list(batch_y.keys()))
+    if len(check_res.unused) > 0:
         if len(check_res.unused) > 1:
             _unused = "{} are not used ".format(check_res.unused)
         else:
             _unused = "{} is not used ".format(check_res.unused)
         _unused += "in function {}.\n".format(func_signature)
-    if len(check_res.duplicated)>0:
+    if len(check_res.duplicated) > 0:
         if len(check_res.duplicated) > 1:
             _duplicated = "duplicated keys {} are detected when calling function {}. \n\tDon't set {} as target and output " \
                           "them in {} at the same time.".format(check_res.duplicated,
-                                                                            func_signature,
-                                                                            check_res.duplicated,
-                                                                  prev_func_signature)
-        else:
-            _duplicated = "duplicated key {} is detected when calling function {}. \n\tDon't set {} as target and output " \
-                          "it in {} at the same time.".format(check_res.duplicated,
                                                                 func_signature,
                                                                 check_res.duplicated,
                                                                 prev_func_signature)
-    _number_errs = int(len(_missing)!=0) + int(len(_duplicated)!=0) + int(len(_unused)!=0)
+        else:
+            _duplicated = "duplicated key {} is detected when calling function {}. \n\tDon't set {} as target and output " \
+                          "it in {} at the same time.".format(check_res.duplicated,
+                                                              func_signature,
+                                                              check_res.duplicated,
+                                                              prev_func_signature)
+    _number_errs = int(len(_missing) != 0) + int(len(_duplicated) != 0) + int(len(_unused) != 0)
     if _number_errs > 0:
         _error_strs = []
         if _number_errs > 1: