Example: import spacy nlp = spacy.load("en_pytt_bertbaseuncased_lg") doc = nlp("Apple pie is delicious.") print(doc._.pytt_last_hidden_state.shape) You can check out the repo here:https://github.com/explosion/spacy-pytorch-transformers …
-
-
Prikaži ovu nitHvala. Twitter će to iskoristiti za poboljšanje vaše vremenske crte. PoništiPoništi
-
-
-
C'est le futur
@huggingface
Kraj razgovora
Novi razgovor -
-
Any recipes for using with prodigy? Awesome stuff!!!
-
Nothing official yet but we're working on it
If you want to play with it you could already write a recipe that loads a models, adds the pytt_textcat component with your labels & returns a simple scored stream & update callback. The tricky part is getting all the params right.
Kraj razgovora
Novi razgovor -
-
-
The post mentions a talk by Devlin at Google Berlin, is that available online anywhere?
-
No, the talk wasn't recorded. It was mostly a summary of the BERT paper and follow-up work such as the whole word masking. I'm sure he's given many similar talks though, so you may be able to find something online.
- Još 2 druga odgovora
Novi razgovor -
-
-
Exciting! Quick notes reg. the examples. 1: The import for "assert_almost_equal" is missing (is this from numpy or torch? does tensor comparison work out of the box here?). 2: the nlp.update param should be "sgd=optimizer". nlp.evaluate seems to take the default TextCategorizer.
-
Thanks, fixing! I think there was actually a small problem with the evaluation, so I'm fixing that as well.
- Još 5 drugih odgovora
Novi razgovor -
-
-
Such a beautifully written explanation on the design thinking and how it works under the hood! Thank you loads!! :)
Hvala. Twitter će to iskoristiti za poboljšanje vaše vremenske crte. PoništiPoništi
-
Čini se da učitavanje traje već neko vrijeme.
Twitter je možda preopterećen ili ima kratkotrajnih poteškoća u radu. Pokušajte ponovno ili potražite dodatne informacije u odjeljku Status Twittera.
BERT, XLNet & GPT-2 in your spaCy pipeline
Based on
Fine-tune pretrained models on your task
Model packages for English & German
Token alignment, similarity & more