BigHeartC / Al_challenger_2018_sentiment_analysis

AI Challenger 2018 细粒度用户评论情感分析,排名17th,基于Aspect Level 思路的解决方案
327 stars 84 forks source link

f1分数为什么只有0.38? #2

Open yiqingyang2012 opened 5 years ago

yiqingyang2012 commented 5 years ago

单独跑了你的GCAE模型,只跑了20个分类中的一个分类的四分类情形,f1分数为什么只有0.38哟

col_idx=0, mean_f1=0.384092749704, f1=[0.9183772 0.        0.        0.6179938]
>>>>>>>>>> epoch 4_1: train-loss=2.696633; train-f1=0.000000; test-loss=2.653187; test-f1=0.384093; <<<<<<<<<<
HBox(children=(IntProgress(value=0, max=1641), HTML(value=u'')))

HBox(children=(IntProgress(value=0, max=469), HTML(value=u'')))

col_idx=0, mean_f1=0.383931036479, f1=[0.91848635 0.         0.         0.6172378 ]
>>>>>>>>>> epoch 4_2: train-loss=2.666327; train-f1=0.000000; test-loss=2.636907; test-f1=0.383931; <<<<<<<<<<
HBox(children=(IntProgress(value=0, max=1641), HTML(value=u'')))

HBox(children=(IntProgress(value=0, max=469), HTML(value=u'')))

col_idx=0, mean_f1=0.389377065007, f1=[0.91867899 0.         0.         0.63882927]
>>>>>>>>>> epoch 5_1: train-loss=2.647028; train-f1=0.000000; test-loss=2.666511; test-f1=0.389377; <<<<<<<<<<
HBox(children=(IntProgress(value=0, max=1641), HTML(value=u'')))

HBox(children=(IntProgress(value=0, max=469), HTML(value=u'')))

col_idx=0, mean_f1=0.388486282112, f1=[0.91941822 0.         0.         0.63452691]
>>>>>>>>>> epoch 5_2: train-loss=2.621379; train-f1=0.000000; test-loss=2.629981; test-f1=0.388486; <<<<<<<<<<
HBox(children=(IntProgress(value=0, max=1641), HTML(value=u'')))

HBox(children=(IntProgress(value=0, max=469), HTML(value=u'')))

col_idx=0, mean_f1=0.386024691301, f1=[0.91755923 0.         0.         0.62653953]
>>>>>>>>>> epoch 6_1: train-loss=2.613969; train-f1=0.000000; test-loss=2.659768; test-f1=0.386025; <<<<<<<<<<
HBox(children=(IntProgress(value=0, max=1641), HTML(value=u'')))

HBox(children=(IntProgress(value=0, max=469), HTML(value=u'')))

col_idx=0, mean_f1=0.38731240782, f1=[0.91939526 0.         0.         0.62985437]
>>>>>>>>>> epoch 6_2: train-loss=2.586844; train-f1=0.000000; test-loss=2.624157; test-f1=0.387312; <<<<<<<<<<
The max f1 of val data is 0.3893770650065502 of step 5_1
bluesky632 commented 5 years ago

您好 ,请问你有遇到AttributeError: 'NoneType' object has no attribute 'session'错误吗?

yiqingyang2012 commented 5 years ago

不要浪费时间,分数上不去,有其他开源的分数很高的模型

bbingy commented 5 years ago

请问有哪些模型分数很高呢 @yiqingyang2012

zwy18 commented 5 years ago

不要浪费时间,分数上不去,有其他开源的分数很高的模型

没有啊 这个效果能到0.7啊

BigHeartC commented 5 years ago

这跑了是 0.7+ 的,你确定没做什么修改吗?