search

mgrankin / ru_transformers

Apache License 2.0
776 stars 108 forks source link

Usage of service #4

Closed AngryJKirk closed 4 years ago

AngryJKirk commented 4 years ago

Hello!

Superb work, I've been looking for thing like that for quite a long time :)

But I am very new in Python and especially in DL, so I have 2 questions:

1) How to use it to get text output?

2) What is the requirements to model? For example, I have a lot of Russian dialogs, usually not very connected to each other, will it fit well?

mgrankin commented 4 years ago
  1. Start with looking at the code located in rest.py, that's an example.
  2. You should try and only when you'll know. Just be careful with overfitting, it's very easy to overfit on small dataset.
gpt2ent commented 4 years ago

Do you detect overfitting by looking at samples or something else (perplexity scores, etc)?

mgrankin commented 4 years ago

I'm aware of only one way to detect overfitting - by using a validation set.

mrcatman commented 4 years ago

Hi, I'm also very new to ML, and I want to ask a few questions: 1) Is there a simple way to train the model on plain TXT files? 2) How big should the model be to give out enough meaningful results? Thanks in advance.

mgrankin commented 4 years ago
  1. Yes, the README is all about that.
  2. There are two models trained in this repo. Try both to find which one is works for you.