new

app.py

@@ -1,7 +1,6 @@
 import gradio as gr
 from transformers import AutoConfig, AutoTokenizer
 from bert_graph import BertForMultipleChoice
-import ipdb
 import torch
 import copy
 from itertools import chain
@@ -13,14 +12,12 @@ def preprocess_function_exp(examples, tokenizer):
 
         # Flatten out
         pair_list = examples
-        # ipdb.set_trace()
         pair_len = [len(item) for item in pair_list]
 
         first_sentences = []
         second_sentences = []
         for line_list in pair_list:
             for line in line_list:
-            # ipdb.set_trace()
                 sent_item = line.strip().split('\t')
                 first_sentences.append(sent_item[0].strip())
                 second_sentences.append(sent_item[1].strip())
@@ -100,8 +97,6 @@ def max_vote(logits1, logits2, pred1, pred2):
         # torch.topk(soft_logits1, n=2)
         values_1, _ = soft_logits1.topk(k=2)
         values_2, _ = soft_logits2.topk(k=2)
-        # import ipdb
-        # ipdb.set_trace()
         # if (values_1[0] - values_2[0]) > (values_1[1] - values_2[1]):
         #     pred_res.append(int(pred1[i].detach().cpu().numpy()))
         # else:
@@ -136,7 +131,6 @@ def model_infer(input_a, input_b):
     examples = [[input_a+'\t'+input_a, input_a+'\t'+input_b, input_b+'\t'+input_a, input_b+'\t'+input_b]]
     tokenized_inputs = preprocess_function_exp(examples, tokenizer)
     tokenized_inputs = DCForMultipleChoice(tokenized_inputs, tokenizer)
-    # ipdb.set_trace()
     outputs = model(**tokenized_inputs)
     predictions, scores = max_vote(outputs.logits[0], outputs.logits[1], outputs.logits[0].argmax(dim=-1), outputs.logits[1].argmax(dim=-1))
 
@@ -148,7 +142,6 @@ def model_infer(input_a, input_b):
     label_b_a = label_space[prediction_b_a]
 
     return 'Head Argument {} Tail Argument'.format(label_a_b, label_b_a)
-    # ipdb.set_trace()
 
 
 with gr.Blocks() as demo: