sahandfer / cem Goto Github PK

There's problem how to open embedding file in ge_embeddings function.
That folder at src/models/common.py in line 700. This code below explain it more:

def gen_embeddings(vocab):
      ...
      for line in open(config.emb_file, encoding='utf-8').readlines() # to solve the opening pretrained embedding issue
      ... 
      return embeddings 

and also, don't forget to download the nltk stopwords first by this syntax

import nltk
nltk.download('stopwords')

Hi Sahand,

Just seeing your AAAA2022 pre-speaking and your paper, and in my mind, COMET plays a critical role on generating cognitive empathy based on your paper.

Sure, I agree with you and commonsense reasoning is very very important for reaching human thinking. My question is : whether having any 中文COMET or 类似中文ATOMIC-2020 dataset for chineses empathy dialog generating ?

Thanks!

return mu_positive, logvar_positive, mu_negative, logvar_positive
I think it should be
return mu_positive, logvar_positive, mu_negative, logvar_negative

Hi , Sahandfer!
I read your paper carefully, but I have some questions about the reproduction of your experimental results. In the automatic evaluation experiment, I completely configured according to your experimental environment, but there is some gap between my running results (ACC, PPL) and those provided by you.

What adjustments do I need to make to the Settings to achieve the effect in your paper? Hyperparameter, environment configuration?

Thank you very much

environment setting:

myself results:

Hello, excellent work!
About line 544 in src/models/EMPDG/model.py, the original line is "weights_best = state["models_g"]", I think maybe it should be "weights_best = state["model"]".

Hi Sahandfer,

Great work for CEM!

Could you please tell us when to upload codes and data, thanks !