def predict():
# Empty cache to ensure larger batch can be loaded for testing
if request.method == "POST":
text = request.form['textInput']
data = [tokenize(text)]
else:
text = "Barack Obama went to Paris and never returned to the USA."
text1 = "Stan Lee was a legend who developed Spiderman and the Avengers movie series."
text2 = "I just learned about donald drumph through john oliver. #JohnOliverShow such an awesome show."
data = [text, text1, text2]
data = [tokenize(text) for text in data]
torch.cuda.empty_cache()
tokens = [obj["value"] for obj in data]
output = list(
get_model_output(model, tokens, args, readers, vocab, test_iterator))
idx = 0
df = output_to_df(tokens[idx], output[idx], vocab)
for k in data[idx].keys():
if k != "value":
df[k] = data[idx][k]
df = df.set_index("tokens")
return render_template(
'sequence_tagging.html',
text=text,
df_html=df.T.to_html(classes="table table-sm table-hover"))
def predict_json():
# Empty cache to ensure larger batch can be loaded for testing
if request.method == "POST":
text = request.form['textInput']
data = [tokenize(text)]
else:
text = "Barack Obama went to Paris and never returned to the USA."
text1 = "Stan Lee was a legend who developed Spiderman and the Avengers movie series."
text2 = "I just learned about donald drumph through john oliver. #JohnOliverShow such an awesome show."
data = [text, text1, text2]
data = [tokenize(text) for text in data]
torch.cuda.empty_cache()
tokens = [obj["value"] for obj in data]
output = list(
get_model_output(model, tokens, args, readers, vocab, test_iterator))
idx = 0
df = output_to_df(tokens[idx], output[idx], vocab)
for k in data[idx].keys():
if k != "value":
df[k] = data[idx][k]
#df = df.set_index("tokens")
output_json = df.to_json(orient='table')
return app.response_class(response=output_json,
status=200,
mimetype='application/json')
def predict_df(texts=None):
# Empty cache to ensure larger batch can be loaded for testing
if texts:
data = [tokenize(text) for text in texts]
else:
text = "Barack Obama went to Paris and never returned to the USA."
text1 = "Stan Lee was a legend who developed Spiderman and the Avengers movie series."
text2 = "I just learned about donald drumph through john oliver. #JohnOliverShow such an awesome show."
texts = [text, text1, text2]
data = [tokenize(text) for text in texts]
torch.cuda.empty_cache()
tokens = [obj["value"] for obj in data]
output = list(
get_model_output(model, tokens, args, readers, vocab, test_iterator))
idx = 0
def _get_data_values(d):
return {k: d[k] for k in d.keys() if k != "value"}
#df = output_to_df(tokens[idx], output[idx], vocab)
df = pd.concat([
output_to_df(tokens[i], output[i],
vocab).assign(**_get_data_values(d)).assign(data_idx=i)
for i, d in enumerate(data)
])
# for k in data[idx].keys():
# if k != "value":
# df[k] = data[idx][k]
return df