@fchollet I don't know much about LSTMs, but folks at CVPR said many devs don't run the training long enough. How long are you training?
Badly-formed LSTNietzsche: "preacious of self-mentifed to the god that it no love and so morality and souls" -A few more epochs necessary...
-
-
-
@quantombone 2-3 hours on a EC2 GPU. You can tell when the training has converged simply by monitoring the training loss against test data.. - Show replies
New conversation -
-
-
.
@fchollet I'd love to hear your opinion on this char level rnn fad; my gut feeling is lstms are huge overkill vs ngram markov models -
@mat_kelcey LSTM is often overkill vs. GRU, but can do much more than markov models, like remembering to close a long-open parenthesis - Show replies
New conversation -
-
-
@fchollet are you using vanilla Keras to do so ?Thanks. Twitter will use this to make your timeline better. UndoUndo
-
Loading seems to be taking a while.
Twitter may be over capacity or experiencing a momentary hiccup. Try again or visit Twitter Status for more information.