Open Directory opened 4 years ago
frankly i feel like its subtly dissing me. what a time we live in
Get more training data and train longer
Increase the training row pairs.
I trained it in 30 Million pairs for 100k steps and still getting all replies having "I'm not sure if " sentence. What should i do to improve the replies?
I know something like this isn't necessarily a bug but this behavior is weird. im using the default everything. the default datasets that came as test data etc. every question i ask it through the inference returns a similar response. "im not sure if your a good thing" Heres a log of some increadably intellectual and pressing questions i asked it.
Did anybody else get these types of responses using the default datasets? i havnt tweaked any settings or anything. im still very unfamilier with nmt, deep learning etc so a lot of the terminology in the README confused me. took me like a day to even get this far cause tensorflow was difficult to setup right. Honestly i would appriciate any general advice that could help me get pointed in the right direction.