add score to output
Browse files
__pycache__/bert_for_sequence_classification.cpython-37.pyc
CHANGED
|
Binary files a/__pycache__/bert_for_sequence_classification.cpython-37.pyc and b/__pycache__/bert_for_sequence_classification.cpython-37.pyc differ
|
|
|
__pycache__/bibert_multitask_classification.cpython-37.pyc
CHANGED
|
Binary files a/__pycache__/bibert_multitask_classification.cpython-37.pyc and b/__pycache__/bibert_multitask_classification.cpython-37.pyc differ
|
|
|
__pycache__/handler.cpython-37.pyc
CHANGED
|
Binary files a/__pycache__/handler.cpython-37.pyc and b/__pycache__/handler.cpython-37.pyc differ
|
|
|
handler.py
CHANGED
|
@@ -32,6 +32,7 @@ tasks = [
|
|
| 32 |
|
| 33 |
|
| 34 |
idtolabel = {"0":"Negative", "1":"Negative", "2": "Neutral", "3":"Positive", "4": "Positive" }
|
|
|
|
| 35 |
|
| 36 |
class EndpointHandler():
|
| 37 |
def __init__(self, path=""):
|
|
@@ -72,12 +73,12 @@ class EndpointHandler():
|
|
| 72 |
|
| 73 |
for idx, x in enumerate(inputs):
|
| 74 |
label = classifier_pol[idx]['label']
|
| 75 |
-
|
| 76 |
|
| 77 |
-
if label == '0' and
|
| 78 |
-
prediction_res.append({"label":"Neutral"})
|
| 79 |
else:
|
| 80 |
-
prediction_res.append({"label":idtolabel.get(classifier_subj[idx]['label'])})
|
| 81 |
elapsed = 1000 * (perf_counter() - t0)
|
| 82 |
logger.info("Model prediction time: %d ms.", elapsed)
|
| 83 |
return prediction_res
|
|
|
|
| 32 |
|
| 33 |
|
| 34 |
idtolabel = {"0":"Negative", "1":"Negative", "2": "Neutral", "3":"Positive", "4": "Positive" }
|
| 35 |
+
idtoscore = {"0": -1, "1": -1, "2": 0, "3": 1, "4": 1 }
|
| 36 |
|
| 37 |
class EndpointHandler():
|
| 38 |
def __init__(self, path=""):
|
|
|
|
| 73 |
|
| 74 |
for idx, x in enumerate(inputs):
|
| 75 |
label = classifier_pol[idx]['label']
|
| 76 |
+
prob = classifier_pol[idx]['probability']
|
| 77 |
|
| 78 |
+
if label == '0' and prob >= 0.75:
|
| 79 |
+
prediction_res.append({"label":"Neutral", "score":0})
|
| 80 |
else:
|
| 81 |
+
prediction_res.append({"label":idtolabel.get(classifier_subj[idx]['label']), "score": idtoscore.get(classifier_subj[idx]['label'])})
|
| 82 |
elapsed = 1000 * (perf_counter() - t0)
|
| 83 |
logger.info("Model prediction time: %d ms.", elapsed)
|
| 84 |
return prediction_res
|