Examples: query, "exact match", wildcard*, wild?ard, wild*rd
Fuzzy search: cake~ (finds cakes, bake)
Term boost: "red velvet"^4, chocolate^2
Field grouping: tags:(+work -"fun-stuff")
Escaping: Escape characters +-&|!(){}[]^"~*?:\ with \, e.g. \+
Range search: properties.timestamp:[1587729413488 TO *] (inclusive), properties.title:{A TO Z}(excluding A and Z)
Combinations: chocolate AND vanilla, chocolate OR vanilla, (chocolate OR vanilla) NOT "vanilla pudding"
Field search: properties.title:"The Title" AND text
Answered
Hi All, I'M Looking For A Well-Built End-To-End Example For A Pytroch Project W\ Lightning W\ Trains. Any Suggestions? I'Ve Started From The Following Project Structure, But The Process Of Adding More And More Features (Many Customs Dataloaders, Early Sto

Hi all, I'm looking for a well-built end-to-end example for a pytroch project w\ lightning w\ trains.
any suggestions?
I've started from the following project structure, but the process of adding more and more features (many customs dataloaders, early stopping, loading checkpoints, more and more metrics, maybe gpu-paralleling soon, etc..) made my code less readable and modular.
https://github.com/codertimo/BERT-pytorch
Thanks in advance,
Nir

  
  
Posted 4 years ago
Votes Newest

Answers 17


Aren't the two lines enough for you? BTW why lightning and not ignite?

  
  
Posted 4 years ago

Hi, I just checked the two lines concept, didn't managed to get normal run-time (supposed to last a minute but it's not progressing) nor the entire logging..

https://demoapp.trains.allegro.ai/projects/215414aaade846db95a9eaac90a9a2d0/experiments/79512df8ec73413294c0f69d72243a7f/output/log

  
  
Posted 4 years ago

But what i meant was an entire project with a well built structure to bases on.. If it uses lightning/ignite/trains for scaling/simplicity reasons - even better (:

  
  
Posted 4 years ago

Looks like it is still running DeliciousSeaanemone40 , you're suggesting it is slower than usual? There are some messages there that I've never seen before

  
  
Posted 4 years ago

There are examples but nothing comes to mind when. Thinking about well fleshed out for Bert etc. Maybe someone here can correct me

  
  
Posted 4 years ago

i'm using logging.getLogger, tqdm, maybe it's not supported?

  
  
Posted 4 years ago

All should work, again - is it much slower than without trains?

  
  
Posted 4 years ago

yes

  
  
Posted 4 years ago

What's your code situation? Is it open enough to allow you to create an issue for this on our GitHub?

  
  
Posted 4 years ago

emm fortunately not (academic thesis which was not published yet), but I can refer to the https://github.com/codertimo/BERT-pytorch I've based on

  
  
Posted 4 years ago

If you can recreate the same problem with the original repo... 🤯 🤩

  
  
Posted 4 years ago

I guess it's not relevant but if it helps:
env settings:
Linux manjaro18 5.4.12-1-MANJARO
Conda env:
Python 3.8.1
torch 1.4

  
  
Posted 4 years ago

Not so much relevant, since it can be seen from your task 😄 but it would be interesting to find out if trains made something be much slower, and if so - how

  
  
Posted 4 years ago

yes I cat see that now ^^'
I'm not sure whether the problem can be recreate in the original BERT repo, but i'll try to look into it. never tried to run the original-repo's model/data.

  
  
Posted 4 years ago

also, for some reason these kind of images (wide, side-by-side) failed to upload (other formats suceeded):
https://demoapp.trains.allegro.ai/projects/215414aaade846db95a9eaac90a9a2d0/experiments/a967de8bd09145ca9e431d96ebb29345/output/log

  
  
Posted 4 years ago

Are you doing imshow or savefig? Is this the matplotlib oop or original subplot? Any warning message relevant to this?

  
  
Posted 4 years ago

i'm doing plt.savefig for both of the successful uploads and the failed ones.
I think the only difference is the failed ones use the plt.subplots.

  
  
Posted 4 years ago
801 Views
17 Answers
4 years ago
one year ago
Tags