Gokul NC

Related work: **Multilingual Abusive Comment Detection (MACD) at Scale for Indic Languages** Dataset: https://github.com/ShareChatAI/MACD Langs: Hindi, Tamil, Telugu, Malayalam and Kannada Count: 30k samples per lang on avg.

BanFakeNews: https://github.com/Rowan1697/FakeNews Paper: https://arxiv.org/pdf/2004.08789.pdf

Yeah, even I don't find the trained model that the repo author has quoted in his README. @qidiso Can you please upload the trained model for the accuracies that you...

@quocnhat : Here's the new link: https://github.com/aidlearning/AidLearning-FrameWork/tree/master/src/facencnn/models All I had to do was browse the repo patiently for a minute. :)

Thanks. OK, I will try that. It might also be that monolingual data using which I generated back-translations is very diverse compared to the learned SPM vocab from the 2M...

Thanks @emjotde I finally managed to find it. I think some kind of **SQL(ite) Injection** has happened because of a noisy line in the data. So my corpus was perfectly...

I guess the gradients seems to be exploding when using `--fp16` aware training. ``` [2021-03-09 13:30:04] NaN/Inf percentage 1.00 in 10 gradient updates, but cost-scaling factor 7.62939e-06 is already at...

@emjotde I am calling it using `marian-server` and the `-c decoder.yml` config after replacing the model name with the converted `model.bin`. Also I just tried using `marian-decoder` too (using **`intgemm8`**...

Please find attached the log (which has the command and model config) [server.log](https://github.com/marian-nmt/marian-dev/files/6128067/server.log) decoder.yml: ``` relative-paths: false models: - model.bin vocabs: - model/vocab.en.spm - model/vocab.hi.spm beam-size: 8 normalize: 0 word-penalty:...

Yes, the trained model.npz model works perfect.