add tqdm, clean up logging

modeling.py

@@ -435,7 +435,6 @@ class BertForSequenceClassification(nn.Module):
 
         def init_weights(m):
             if isinstance(m, (nn.Linear, nn.Embedding)):
-                print("Initializing {}".format(m))
                 # Slight difference here with the TF version which uses truncated_normal
                 # cf https://github.com/pytorch/pytorch/pull/5617
                 m.weight.data.normal_(config.initializer_range)
@@ -481,7 +480,6 @@ class BertForQuestionAnswering(nn.Module):
 
         def init_weights(m):
             if isinstance(m, (nn.Linear, nn.Embedding)):
-                print("Initializing {}".format(m))
                 # Slight difference here with the TF version which uses truncated_normal for initialization
                 # cf https://github.com/pytorch/pytorch/pull/5617
                 m.weight.data.normal_(config.initializer_range)

run_squad.py

@@ -912,9 +912,9 @@ def main():
 
         model.eval()
         all_results = []
-        logger.info("Start evaulating")
+        logger.info("Start evaluating")
         #for input_ids, input_mask, segment_ids, label_ids, example_index in eval_dataloader:
-        for input_ids, input_mask, segment_ids, example_index in eval_dataloader:
+        for input_ids, input_mask, segment_ids, example_index in tqdm(eval_dataloader, descr="Evaluating"):
             if len(all_results) % 1000 == 0:
                 logger.info("Processing example: %d" % (len(all_results)))