change expit to softmax (PaddlePaddle#5006)

applications/neural_search/ranking/cross_encoder/deploy/python/predict.py

@@ -16,52 +16,32 @@
 import os
 import sys
 
-from scipy.special import softmax
-from scipy.special import expit
 import numpy as np
-import pandas as pd
-
 import paddle
 from paddle import inference
-import paddle.nn.functional as F
-from paddlenlp.data import Stack, Tuple, Pad
+from scipy.special import softmax
+
+from paddlenlp.data import Pad, Tuple
 from paddlenlp.datasets import load_dataset
-from paddlenlp.utils.log import logger
 from paddlenlp.transformers import AutoTokenizer
+from paddlenlp.utils.log import logger
 
 sys.path.append(".")
 
 # yapf: disable
 parser = argparse.ArgumentParser()
-parser.add_argument("--model_dir", type=str, required=True,
-    help="The directory to static model.")
-
-parser.add_argument("--input_file", type=str, required=True,
-    help="The test set file.")
-parser.add_argument("--max_seq_length", default=128, type=int,
-    help="The maximum total input sequence length after tokenization. Sequences "
-    "longer than this will be truncated, sequences shorter will be padded.")
-parser.add_argument("--batch_size", default=32, type=int,
-    help="Batch size per GPU/CPU for training.")
-parser.add_argument('--device', choices=['cpu', 'gpu', 'xpu'], default="gpu",
-    help="Select which device to train model, defaults to gpu.")
-
-parser.add_argument('--use_tensorrt', default=False, type=eval, choices=[True, False],
-    help='Enable to use tensorrt to speed up.')
-parser.add_argument("--precision", default="fp32", type=str, choices=["fp32", "fp16", "int8"],
-    help='The tensorrt precision.')
-
-parser.add_argument('--cpu_threads', default=10, type=int,
-    help='Number of threads to predict when using cpu.')
-parser.add_argument('--enable_mkldnn', default=False, type=eval, choices=[True, False],
-    help='Enable to use mkldnn to speed up when using cpu.')
-
-parser.add_argument("--benchmark", type=eval, default=False,
-    help="To log some information about environment and running.")
-parser.add_argument("--save_log_path", type=str, default="./log_output/",
-    help="The file path to save log.")
+parser.add_argument("--model_dir", type=str, required=True, help="The directory to static model.")
+parser.add_argument("--input_file", type=str, required=True, help="The test set file.")
+parser.add_argument("--max_seq_length", default=128, type=int, help="The maximum total input sequence length after tokenization. Sequences longer than this will be truncated, sequences shorter will be padded.")
+parser.add_argument("--batch_size", default=32, type=int, help="Batch size per GPU/CPU for training.")
+parser.add_argument('--device', choices=['cpu', 'gpu', 'xpu'], default="gpu", help="Select which device to train model, defaults to gpu.")
+parser.add_argument('--use_tensorrt', default=False, type=eval, choices=[True, False], help='Enable to use tensorrt to speed up.')
+parser.add_argument("--precision", default="fp32", type=str, choices=["fp32", "fp16", "int8"], help='The tensorrt precision.')
+parser.add_argument('--cpu_threads', default=10, type=int, help='Number of threads to predict when using cpu.')
+parser.add_argument('--enable_mkldnn', default=False, type=eval, choices=[True, False], help='Enable to use mkldnn to speed up when using cpu.')
+parser.add_argument("--benchmark", type=eval, default=False, help="To log some information about environment and running.")
+parser.add_argument("--save_log_path", type=str, default="./log_output/", help="The file path to save log.")
 parser.add_argument('--model_name_or_path', default="rocketqa-base-cross-encoder", help="The pretrained model used for training")
-
 args = parser.parse_args()
 # yapf: enable
 
@@ -203,7 +183,7 @@ def predict(self, data, tokenizer):
         sim_score = self.output_handle.copy_to_cpu()
         if args.benchmark:
             self.autolog.times.stamp()
-        sim_score = expit(sim_score)[:, 1]
+        sim_score = softmax(sim_score)[:, 1]
 
         if args.benchmark:
             self.autolog.times.end(stamp=True)

applications/neural_search/ranking/cross_encoder/deploy/python/web_service.py

@@ -12,12 +12,10 @@
 # See the License for the specific language governing permissions and
 # limitations under the License.
 
-import numpy as np
-import sys
 import json
 
-from scipy.special import expit
-from paddle_serving_server.web_service import WebService, Op
+from paddle_serving_server.web_service import Op, WebService
+from scipy.special import softmax
 
 
 def convert_example(example, tokenizer, max_seq_length=512):
@@ -38,7 +36,7 @@ def init_op(self):
         self.tokenizer = AutoTokenizer.from_pretrained("rocketqa-base-cross-encoder")
 
     def preprocess(self, input_dicts, data_id, log_id):
-        from paddlenlp.data import Stack, Tuple, Pad
+        from paddlenlp.data import Pad, Tuple
 
         ((_, input_dict),) = input_dicts.items()
         print("input dict", input_dict)
@@ -60,7 +58,7 @@ def preprocess(self, input_dicts, data_id, log_id):
 
     def postprocess(self, input_dicts, fetch_dict, data_id, log_id):
         new_dict = {}
-        sim_score = expit(fetch_dict["predict"])[:, 1]
+        sim_score = softmax(fetch_dict["predict"])[:, 1]
         new_dict["predict"] = str(sim_score)
         return new_dict, None, ""