[PIR] Fix text_classification (#10497)

Author: hanlintang

slm/applications/text_classification/hierarchical/retrieval_based/deploy/python/predict.py

@@ -22,6 +22,10 @@
 
 from paddlenlp.data import Pad, Tuple
 from paddlenlp.transformers import AutoTokenizer
+from paddlenlp.utils.env import (
+    PADDLE_INFERENCE_MODEL_SUFFIX,
+    PADDLE_INFERENCE_WEIGHTS_SUFFIX,
+)
 
 sys.path.append(".")
 
@@ -114,8 +118,8 @@ def __init__(
         self.max_seq_length = max_seq_length
         self.batch_size = batch_size
 
-        model_file = model_dir + "/inference.get_pooled_embedding.pdmodel"
-        params_file = model_dir + "/inference.get_pooled_embedding.pdiparams"
+        model_file = model_dir + f"/inference.get_pooled_embedding{PADDLE_INFERENCE_MODEL_SUFFIX}"
+        params_file = model_dir + f"/inference.get_pooled_embedding{PADDLE_INFERENCE_WEIGHTS_SUFFIX}"
         if not os.path.exists(model_file):
             raise ValueError("not find model file path {}".format(model_file))
         if not os.path.exists(params_file):

slm/applications/text_classification/hierarchical/retrieval_based/utils/feature_extract.py

@@ -22,6 +22,10 @@
 
 import paddlenlp as ppnlp
 from paddlenlp.data import Pad, Tuple
+from paddlenlp.utils.env import (
+    PADDLE_INFERENCE_MODEL_SUFFIX,
+    PADDLE_INFERENCE_WEIGHTS_SUFFIX,
+)
 
 # fmt: off
 parser = argparse.ArgumentParser()
@@ -82,8 +86,8 @@ def __init__(
         self.max_seq_length = max_seq_length
         self.batch_size = batch_size
 
-        model_file = model_dir + "/inference.get_pooled_embedding.pdmodel"
-        params_file = model_dir + "/inference.get_pooled_embedding.pdiparams"
+        model_file = model_dir + f"/inference.get_pooled_embedding{PADDLE_INFERENCE_MODEL_SUFFIX}"
+        params_file = model_dir + f"/inference.get_pooled_embedding{PADDLE_INFERENCE_WEIGHTS_SUFFIX}"
         if not os.path.exists(model_file):
             raise ValueError("not find model file path {}".format(model_file))
         if not os.path.exists(params_file):

slm/applications/text_classification/multi_class/retrieval_based/deploy/python/predict.py

@@ -22,6 +22,10 @@
 
 from paddlenlp.data import Pad, Tuple
 from paddlenlp.transformers import AutoTokenizer
+from paddlenlp.utils.env import (
+    PADDLE_INFERENCE_MODEL_SUFFIX,
+    PADDLE_INFERENCE_WEIGHTS_SUFFIX,
+)
 
 sys.path.append(".")
 
@@ -114,8 +118,8 @@ def __init__(
         self.max_seq_length = max_seq_length
         self.batch_size = batch_size
 
-        model_file = model_dir + "/inference.get_pooled_embedding.pdmodel"
-        params_file = model_dir + "/inference.get_pooled_embedding.pdiparams"
+        model_file = model_dir + f"/inference.get_pooled_embedding{PADDLE_INFERENCE_MODEL_SUFFIX}"
+        params_file = model_dir + f"/inference.get_pooled_embedding{PADDLE_INFERENCE_WEIGHTS_SUFFIX}"
         if not os.path.exists(model_file):
             raise ValueError("not find model file path {}".format(model_file))
         if not os.path.exists(params_file):

slm/applications/text_classification/multi_class/retrieval_based/utils/feature_extract.py

@@ -22,6 +22,10 @@
 
 import paddlenlp as ppnlp
 from paddlenlp.data import Pad, Tuple
+from paddlenlp.utils.env import (
+    PADDLE_INFERENCE_MODEL_SUFFIX,
+    PADDLE_INFERENCE_WEIGHTS_SUFFIX,
+)
 
 # fmt: off
 parser = argparse.ArgumentParser()
@@ -83,8 +87,8 @@ def __init__(
         self.max_seq_length = max_seq_length
         self.batch_size = batch_size
 
-        model_file = model_dir + "/inference.get_pooled_embedding.pdmodel"
-        params_file = model_dir + "/inference.get_pooled_embedding.pdiparams"
+        model_file = model_dir + f"/inference.get_pooled_embedding{PADDLE_INFERENCE_MODEL_SUFFIX}"
+        params_file = model_dir + f"/inference.get_pooled_embedding{PADDLE_INFERENCE_WEIGHTS_SUFFIX}"
         if not os.path.exists(model_file):
             raise ValueError("not find model file path {}".format(model_file))
         if not os.path.exists(params_file):

slm/applications/text_classification/multi_label/README.md

@@ -355,7 +355,7 @@ python export_model.py --params_path ./checkpoint/ --output_path ./export --mult
 export/
 ├── float32.pdiparams
 ├── float32.pdiparams.info
-└── float32.pdmodel
+└── float32.json(PIR enabled)/float32.pdmodel(PIR disabled)
 ```
  导出模型之后用于部署，项目提供了基于 ONNXRuntime 的 [离线部署方案](./deploy/predictor/README.md) 和基于 Paddle Serving 的 [在线服务化部署方案](./deploy/predictor/README.md)。
 

slm/applications/text_classification/multi_label/retrieval_based/deploy/python/predict.py

@@ -22,6 +22,10 @@
 
 from paddlenlp.data import Pad, Tuple
 from paddlenlp.transformers import AutoTokenizer
+from paddlenlp.utils.env import (
+    PADDLE_INFERENCE_MODEL_SUFFIX,
+    PADDLE_INFERENCE_WEIGHTS_SUFFIX,
+)
 
 sys.path.append(".")
 
@@ -114,8 +118,8 @@ def __init__(
         self.max_seq_length = max_seq_length
         self.batch_size = batch_size
 
-        model_file = model_dir + "/inference.get_pooled_embedding.pdmodel"
-        params_file = model_dir + "/inference.get_pooled_embedding.pdiparams"
+        model_file = model_dir + f"/inference.get_pooled_embedding{PADDLE_INFERENCE_MODEL_SUFFIX}"
+        params_file = model_dir + f"/inference.get_pooled_embedding{PADDLE_INFERENCE_WEIGHTS_SUFFIX}"
         if not os.path.exists(model_file):
             raise ValueError("not find model file path {}".format(model_file))
         if not os.path.exists(params_file):

slm/applications/text_classification/multi_label/retrieval_based/utils/feature_extract.py

@@ -22,6 +22,10 @@
 
 import paddlenlp as ppnlp
 from paddlenlp.data import Pad, Tuple
+from paddlenlp.utils.env import (
+    PADDLE_INFERENCE_MODEL_SUFFIX,
+    PADDLE_INFERENCE_WEIGHTS_SUFFIX,
+)
 
 # fmt: off
 parser = argparse.ArgumentParser()
@@ -84,8 +88,8 @@ def __init__(
         self.max_seq_length = max_seq_length
         self.batch_size = batch_size
 
-        model_file = model_dir + "/inference.get_pooled_embedding.pdmodel"
-        params_file = model_dir + "/inference.get_pooled_embedding.pdiparams"
+        model_file = model_dir + f"/inference.get_pooled_embedding{PADDLE_INFERENCE_MODEL_SUFFIX}"
+        params_file = model_dir + f"/inference.get_pooled_embedding{PADDLE_INFERENCE_WEIGHTS_SUFFIX}"
         if not os.path.exists(model_file):
             raise ValueError("not find model file path {}".format(model_file))
         if not os.path.exists(params_file):