|
import gradio as gr |
|
from os import environ |
|
from transformers import pipeline |
|
|
|
api_key = environ.get("api_key") |
|
|
|
def app(input, filter_strategy, relevence_threshold, k ): |
|
classifier_pipeline = pipeline(model="gianma/testModel", tokenizer='gianma/testModel', use_auth_token=api_key) |
|
|
|
kwargs = {'padding':True,'truncation':True,'max_length':512} |
|
|
|
if filter_strategy=='top_k': |
|
kwargs['top_k'] = k |
|
|
|
res = classifier_pipeline(input, **kwargs) |
|
|
|
res = {el['label']:el['score'] for el in res} |
|
|
|
if filter_strategy == 'soglia_di_confidenza': |
|
res = {k:v for k,v in res.items() if k >= relevence_threshold} |
|
|
|
return res |
|
|
|
iface = gr.Interface(fn=app, |
|
inputs=["text", gr.Radio(['soglia_di_confidenza', 'top_k'], value='soglia_di_confidenza', label='come vuoi fiiltrare i risultati'), gr.Slider(0, 1, value=0.5,label='soglia di confidenza'), gr.Slider(1, 21, step= 1,, value=4, label='k')], |
|
outputs="label") |
|
|
|
iface.launch() |
|
|
|
|
|
|
