分类模型的f1值大概是多少_准确率召回率f1分数

欢迎您对PaddleHub提出建议，非常感谢您对PaddleHub的贡献！

目前PaddleHub已实现您提出的需求，请安装PaddleHub>=1.7.0, PaddlePaddle>=1.8.0快速体验。请您在定义task任务时增加metrics_choices=[‘f1’]选项，即可实现多分类F1 Score评估指标，示例如下：

task = hub.ImageClassifierTask(

data_reader=data_reader,

feed_list=feed_list,

feature=feature_map,

num_classes=dataset.num_labels,

metrics_choices=[‘f1’], #PaddleHub同时支持’f1’和’acc’评价标准,可使用metrics_choices=[‘f1’, ‘acc’]快速实现

config=config)

发现一个新的问题，ImageClassifierTask设定f1作为metrics时，多分类任务会报错，

metrics_choices = [‘f1’]

错误信息：

[2020-08-07 11:13:35,971] [ INFO] – PaddleHub finetune start

—————————————————————————TypeError Traceback (most recent call last) in

—-> 1 run_states = task.finetune_and_eval()

/opt/conda/envs/python35-paddle120-env/lib/python3.7/site-packages/paddlehub/finetune/task/base_task.py in finetune_and_eval(self)

943

944 def finetune_and_eval(self):

–> 945 return self.finetune(do_eval=True)

946

947 def finetune(self, do_eval=False):

/opt/conda/envs/python35-paddle120-env/lib/python3.7/site-packages/paddlehub/finetune/task/base_task.py in finetune(self, do_eval)

964 while self.current_epoch <= self.config.num_epoch:

965 self.config.strategy.step()

–> 966 run_states = self._run(do_eval=do_eval)

967 self.env.current_epoch += 1

968

/opt/conda/envs/python35-paddle120-env/lib/python3.7/site-packages/paddlehub/finetune/task/base_task.py in _run(self, do_eval)

1222 if self.is_train_phase:

1223 if self.current_step % self.config.log_interval == 0:

-> 1224 self._log_interval_event(period_run_states)

1225 global_run_states += period_run_states

1226 period_run_states = []

/opt/conda/envs/python35-paddle120-env/lib/python3.7/site-packages/paddlehub/finetune/task/base_task.py in hook_function(self, *args)

706 for name, func in self._hooks[hook_type].items():

707 if inspect.ismethod(func):

–> 708 func(*args)

709 else:

710 partial(func, self)(*args)

/opt/conda/envs/python35-paddle120-env/lib/python3.7/site-packages/paddlehub/finetune/task/base_task.py in default_log_interval_event(self, run_states)

849 tag=”{}{}”.format(metric, self.phase),

850 value=scores[metric],

–> 851 step=self._envs[‘train’].current_step)

852 log_scores += “%s=%.5f ” % (metric, scores[metric])

853 logger.train(“step %d / %d: loss=%.5f %s[step/sec: %.2f]” %

/opt/conda/envs/python35-paddle120-env/lib/python3.7/site-packages/visualdl/writer/writer.py in add_scalar(self, tag, value, step, walltime)

133 walltime = round(time.time()) if walltime is None else walltime

134 self._get_file_writer().add_record(

–> 135 scalar(tag=tag, value=value, step=step, walltime=walltime))

136

137 def add_image(self, tag, img, step, walltime=None):

/opt/conda/envs/python35-paddle120-env/lib/python3.7/site-packages/visualdl/component/base_component.py in scalar(tag, value, step, walltime)

31 Package with format of record_pb2.Record

32 “””

—> 33 value = float(value)

34 return Record(values=[

35 Record.Value(id=step, tag=tag, timestamp=walltime, value=value)

TypeError: float() argument must be a string or a number, not ‘tuple’

查看了代码，问题可能出在PaddleHub/paddlehub/finetune/task/classifier_task.py

calculate_metrics()调用的calculate_f1_np()函数应该是只能对2分类任务计算f1、precision和recall。