.gitignore

@@ -142,4 +142,6 @@ analyze_fail.dat
 src/__pycache__/
 src/cased_L-24_H-1024_A-16/
 data/
-kernel_meta/
+kernel_meta/
+output/
+log/

run_KTNET_squad.py

@@ -15,6 +15,7 @@ from utils.util import LossCallBack, make_directory, LoadNewestCkpt
 from src.reader.squad_twomemory import DataProcessor, write_predictions
 from src.dataset import create_squad_train_dataset, create_squad_dev_dataset
 
+import mindspore
 import mindspore.common.dtype as mstype
 from mindspore import context
 from mindspore import log as logger
@@ -215,13 +216,13 @@ def do_eval(processor, eval_concept_settings, eval_output_name='eval_result.json
         #     input_data.append(data[i])
         # input_mask, src_ids, pos_ids, sent_ids, wn_concept_ids, nell_concept_ids, unique_id = input_data
         
-        src_ids = np.squeeze(data[0])
-        pos_ids = np.squeeze(data[1])
-        sent_ids = np.squeeze(data[2])
-        wn_concept_ids = data[3]
-        nell_concept_ids = data[4]
-        input_mask = np.squeeze(data[5])
-        unique_id = data[6]
+        src_ids = Tensor(np.squeeze(data[0]), mindspore.int32)
+        pos_ids = Tensor(np.squeeze(data[1]), mindspore.int32)
+        sent_ids = Tensor(np.squeeze(data[2]), mindspore.int32)
+        wn_concept_ids = Tensor(data[3], mindspore.int32)
+        nell_concept_ids = Tensor(data[4], mindspore.int32)
+        input_mask = Tensor(np.squeeze(data[5]), mindspore.float32)
+        unique_id = Tensor(data[6], mindspore.int32)
 
         pad = ops.Pad(((0, 0), (0, 0), (0, 3), (0, 0)))
         nell_concept_ids = pad(nell_concept_ids)
@@ -244,7 +245,7 @@ def do_eval(processor, eval_concept_settings, eval_output_name='eval_result.json
                 start_logits=start_logits,
                 end_logits=end_logits))
             
-            logger.info("unique_id: %d" % unique_id)
+            # logger.info("unique_id: %d" % unique_id)
 
         # callback.update(logits, unique_id)
     if not os.path.exists(args.checkpoints):

scripts/run_squad_twomemory.sh

@@ -42,10 +42,10 @@ NELL_CPT_EMBEDDING_PATH=data/KB_embeddings/nell_concept2vec.txt
 
 python3 run_KTNET_squad.py \
   --device_target "Ascend" \
-  --device_id 4 \
+  --device_id 5 \
   --batch_size 6 \
-  --do_train true \
-  --do_predict false \
+  --do_train false \
+  --do_predict true \
   --do_lower_case false \
   --init_pretraining_params $BERT_DIR/params \
   --load_pretrain_checkpoint_path $BERT_DIR/roberta.ckpt \