I’m trying to make imdb classification training time shorter.
I have managed to make it 15 min per epoch, but still, it’s very slow.
There seem to be a bottle neck that has nothing to do with model complexity.
Any starting pointer of how to try to see what’s taking a long time?
I tried to use %prun but the result is too messy.
So I was gonna use %sankeviz, but port forwarding seems not to be working.
Can somebody take a look at my ssh setting and error message and tell me how to set it right?
ssh -L 9999:localhost:9999 -L 8080:localhost:8080 -L 8080:127.0.0.1:8080 paperspace@mypaperspacebox
where 9999 is used by jupyter and 8080 is reserved for snakeviz.
this is what I get when I run %snakeviz train_my_model
Usage: snakeviz [options] filename
snakeviz: error: no web browser found: could not locate runnable browser
snakeviz web server started on 127.0.0.1:8080; enter Ctrl-C to exit
but when I look at http://127.0.0.1:8080/snakeviz/%2Ftmp%2Ftmpdz4pglqa or even http://localhost:8080/snakeviz/%2Ftmp%2Ftmpdz4pglqa, nothing’s there.
Thoughts and ideas please. :’(