Experimenting with few-shot classification.

app.py

@@ -74,7 +74,7 @@ few_shot_model = SetFitModel.from_pretrained("moshew/bge-small-en-v1.5_setfit-ss
 candidate_labels = ["true", "false"]
 reference_dataset = load_dataset("emotion")
 dummy_dataset = Dataset.from_dict({})
-train_dataset = get_templated_dataset(dummy_dataset, candidate_labels=candidate_labels, sample_size=8, template="Based on the Given passage, the hypothesis is {}.")
+train_dataset = get_templated_dataset(dummy_dataset, candidate_labels=candidate_labels, sample_size=8, template="This statement is {}.")
 args = TrainingArguments(
     batch_size=32,
     num_epochs=1
@@ -110,14 +110,18 @@ def zero_shot_classification(data):
     return response_string
 
 def create_sequences(data):
-    return ['###Given:\n' + data['sequence'] + '\n###End Given\n###Hypothesis:\n' + data['hypothesis_template'].format(label) + "\n###End Hypothesis" for label in data['candidate_labels']]
+    # return ['###Given:\n' + data['sequence'] + '\n###End Given\n###Hypothesis:\n' + data['hypothesis_template'].format(label) + "\n###End Hypothesis" for label in data['candidate_labels']]
+    return [data['sequence'] + '\n' + data['hypothesis_template'].format(label) + for label in data['candidate_labels']]
 
 def few_shot_classification(data):
     sequences = create_sequences(data)
     print(sequences)
     # results = onnx_few_shot_model(sequences)
     probs = onnx_few_shot_model.predict_proba(sequences)
-    response_string = json.dumps(probs.tolist())
+    scores = [true[0] for true in probs]
+    response_dict = {'scores': scores, 'labels': data.candidate_labels}
+    print()response_dict)
+    response_string = json.dumps(response_dict)
     return response_string
 
 gradio_interface = gradio.Interface(