Tuning as in the reproducibility section #208
sorgfresser
started this conversation in
General
Replies: 1 comment
-
Hi @sorgfresser
I used only the dev split for tuning. The performance reported in the paper (and README) is on test
I got AMI from the same repo. It also resulted in that file being out of any split. |
Beta Was this translation helpful? Give feedback.
0 replies
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
-
Hey, since I did not find any further information on this so far: how exactly did you tune the hyperparams stated in the reproducibility section? I'm currently messing around with
diart.tune
for AMI with the new pyannote models but would like a fair comparison against the old DER values.I will post any questions coming up regarding the tuning process here to have one concise place for them.
As of now:
1.) Did you use the whole dataset or just the dev set to tune?
2.) I used the AMI dataset by pyannote/AMI-diarization-setup. At least for me this resulted in IB4005.wav lacking the matching rttm. Where did you get AMI from?
Beta Was this translation helpful? Give feedback.
All reactions