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
Unanswered
Hi


Good news:
 new 

best_model

 is saved, add a tag 

best

,
Already supported, (you just can't see the tag, but it is there :))

Interesting! Could you point me where the tagging happens? Also, by "see" do you mean UI? I tried doing task.models["outputs"][-1].tag but there's no property tag ( AttributeError: 'Model' object has no attribute 'tag' , seems that only OutputModel s have the tag property?)

I'll answer specifically for ignite on your question regarding the interface:
Doing something like passing the tag to TrainSaver seems convenient. Same for the config / metadata etc
TrainsSaver(output_uri, tag="best", metadata={})

(btw, obviously if we know it's not good, why do we bother to store it in the first place...)

During training you find a new best model every some epochs, so you save that. I agree no need to save all the best models, only one is enough (that's why I have n_saved=1 in my ignite checkpoints for best models). So you save a best_model , tag it, add metadata to it. Then you find a new best_model , and overwrite the previous.

But in ignite they give the option to keep n_saved so you would need to remove the "best" tag from the previously saved best models and only add it on the last one.

I hope this makes sense 😄

  
  
Posted 3 years ago
77 Views
0 Answers
3 years ago
one year ago