-
I recompiled mmt from sources and successfully trained the example model.
Then I attempted "mmt start" but it did not start and no log was produced.
Here is a screen shot starting at the end of trai…
-
Using this PR https://github.com/modernmt/modernmt/pull/553 I wrote a small NMT live-tuning demonstrator:
http://biliedit.cubaix.com/
Here is a short video in action:
https://youtu.be/et3tyhMegKE
…
-
@davidecaroselli @nicolabertoldi
1. Back-up your existing engines so that you can copy them back later.
2. Reinstall Ubuntu 18.02 from a USB stick. Note that Ubuntu will not recognize your graphi…
-
1. Install Ubuntu 20.04 LTS (this does not work on 18.04)
2. Install NVIDIA driver
`sudo apt install nvidia-driver-455`
(I don't believe higher drivers like 460 will work.) There's a possibility th…
-
If the splitters specified in the command line do not exist this just fails silently and the script still runs through the entire corpus download.
-
I apologize if this is somehow fixed in a newer version (I am using 4.5.4) but it looks like no changes have been made to any queuing logic recently: https://github.com/modernmt/modernmt/tree/master/s…
-
Working on #436, I'm studying MMT code. I discovered that MMT is strictly serialising sentence processing, one sentence at a time. This is done using a decoder list in a blocking queue. These decoders…
-
Currently, the early stopping is implemented by looking at the TensorboardX loss at every N for `--save-interval-updates N`. This results in unexpected behavior if one tries to disable interval-based …
-
Hi!
What I'm doing is a bit particular: translating noisy text to denoised clean text (for NeuroSpell auto-corrector). On the principle, it should not cause the strange thing described below. It wo…
-
Hello. I've built a model with around 130,000 parallel lines of data with around 50% additional back translated data.
It's working quite well. But when I try to translate a single word the respons…