You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
{{ message }}
This repository has been archived by the owner on Dec 29, 2022. It is now read-only.
Hi,
Thanks for releasing the code for active-qa.
After browsing the code, I did not find Monte-Carlo Sampling in the training stage. It seems that each training instance consists of only one 「query, reformulated_query, reward」 tuple. Therefore, the reward is the same for each token in one reformulated query.
I don't know whether the suspicion is right. If it is right, what will model perform with or without Monte-Carlo sampling? Maybe using only one instance for Monte Carlo sampling is like the relation between stochastic gradient descent and gradient descent?
Thank you
The text was updated successfully, but these errors were encountered:
me the same, have you finger out the problem?or get a new version code? @fangkuann
I didn't try that. But we apply the paper's training method in our query rewrite module, then the retrieve performance could be enhanced. We further improved this method by adding a value network, for details you may refer to this Chinese technic article https://www.6aiq.com/article/1577969687897.
me the same, have you finger out the problem?or get a new version code? @fangkuann
I didn't try that. But we apply the paper's training method in our query rewrite module, then the retrieve performance could be enhanced. We further improved this method by adding a value network, for details you may refer to this Chinese technic article https://www.6aiq.com/article/1577969687897.
@fangkuann Yes, I follow this article to here. May I ask some question by email? I couldn't found a way to concat you. Send a message to yanggodfly1994@gmail.com if it's ok, thanks a lot
Sign up for freeto subscribe to this conversation on GitHub.
Already have an account?
Sign in.
Hi,
Thanks for releasing the code for active-qa.
After browsing the code, I did not find Monte-Carlo Sampling in the training stage. It seems that each training instance consists of only one 「query, reformulated_query, reward」 tuple. Therefore, the reward is the same for each token in one reformulated query.
I don't know whether the suspicion is right. If it is right, what will model perform with or without Monte-Carlo sampling? Maybe using only one instance for Monte Carlo sampling is like the relation between stochastic gradient descent and gradient descent?
Thank you
The text was updated successfully, but these errors were encountered: