GPU on apply function colab

user_5276544 0
Viewed 510

I use this apply function on my code:

def entities_extraction(text):
    doc = nlp(text)
    entities= [ent.text for sentence in doc.sentences for ent in sentence.entities if ent.type in {"PERSON", "ORG", "GPE", "NORP", "FAC", "LOC", "PRODUCT", "EVENT", "WORK_OF_ART", "LAW", "LANGUAGE", "MISC"}]
    return entities

df["entities"] = df['text'].progress_apply(lambda x: entities_extraction(x))

The problem is that, at the moment, it is too slow (it take nearly 12 hours)

So I tried to modify it to use colab gpu:

@cuda.jit
def entities_extraction(text):
    doc = nlp(text)
    entities= [ent.text for sentence in doc.sentences for ent in sentence.entities if ent.type in {"PERSON", "ORG", "GPE", "NORP", "FAC", "LOC", "PRODUCT", "EVENT", "WORK_OF_ART", "LAW", "LANGUAGE", "MISC"}]
    return entities

But I get this error:

ValueError: 
Kernel launch configuration was not specified. Use the syntax:

kernel_function[blockspergrid, threadsperblock](arg0, arg1, ..., argn)

See https://numba.pydata.org/numba-doc/latest/cuda/kernels.html#kernel-invocation for help.

Do you know how to solve it, or you have a better implementation for gpu on apply functions?

Sorry if I made a lot of mistakes, I'm new to the argument of gpu for faster code. Thank you for your help!

google-colaboratory stanford-nlp cuda gpu python
1 Answers

You should write a code that cuda.jit decorator expects as shown in the link or you could try to use cuDF.apply. However, I am unsure if the cuDF supports for your function in lambda.

Alternatively, you could use Stanza Pipeline API with use_gpu=True parameter. See the basic example section, and you should pay attention to the note in there.

user_18420980
Related