fix layer index (#93)

* specify layer_index via truncate_layer

* update doc

* fix test cases

angle_emb/angle.py

@@ -4,7 +4,6 @@
 import re
 import sys
 import json
-import copy
 import math
 from functools import partial
 from typing import Any, Dict, Optional, List, Union, Tuple, Callable
@@ -1213,8 +1212,6 @@ def __init__(self,
             pooling_strategy=self.pooling_strategy,
             padding_strategy=self.tokenizer.padding_side)
 
-        # full_backbone is used to Espresso inference
-        self.full_backbone = None
         self.__cfg = {
             'model_name_or_path': model_name_or_path,
             'max_length': max_length,
@@ -1474,12 +1471,22 @@ def evaluate(self, data: Dataset, batch_size: int = 32, metric: str = 'spearman_
             batch_size=batch_size,
         )(self)[metric]
 
+    def truncate_layer(self, layer_index: int):
+        """ truncate layer
+
+        :param layer_index: int. layers after layer_index will be truncated.
+        :return: self
+        """
+        if len(self.backbone.encoder.layer) < layer_index:
+            logger.info('current layer_index is larger than the number of layers, please check whether it is correct')
+        self.backbone.encoder.layer = self.backbone.encoder.layer[:layer_index]
+        return self
+
     def encode(self,
                inputs: Union[List[str], Tuple[str], List[Dict], str],
                max_length: Optional[int] = None,
                end_with_eos: bool = False,
                to_numpy: bool = True,
-               layer_index: int = -1,
                embedding_start: int = 0,
                embedding_size: Optional[int] = None,
                device: Optional[Any] = None,
@@ -1491,23 +1498,17 @@ def encode(self,
         :param inputs: Union[List[str], Tuple[str], List[Dict], str]. Input texts. Required.
         :param max_length: Optional[int]. Default None.
         :param to_numpy: bool. Default True.
-        :param layer_index: int. Obtain specific layer's sentence embeddings (for Espresso).
         :param embedding_start: int. Specify the start position of the embedding (for Espresso).
         :param embedding_size: Optional[int]. Specify embedding size (for Espresso).
             The embeddings from embedding_start to embedding_start+embedding_size will be returned.
         :param device: Optional[Any]. Default None.
         :param prompt: Optional[str]. Default None.
         :param normalize_embedding: bool. Default False.
         """
-        if layer_index != -1 and self.full_backbone is None:
-            self.full_backbone = copy.deepcopy(self.backbone)
-
-        if layer_index != -1:
-            self.backbone.encoder.layer = self.full_backbone.encoder.layer[:layer_index]
+        self.backbone.eval()
 
         if device is None:
             device = self.device
-        self.backbone.eval()
         if not isinstance(inputs, (tuple, list)):
             inputs = [inputs]
         if prompt is not None:
@@ -1537,7 +1538,6 @@ def encode(self,
         tok.to(device)
         with torch.no_grad():
             output = self.pooler(tok,
-                                 layer_index=layer_index,
                                  embedding_start=embedding_start,
                                  embedding_size=embedding_size)
         if normalize_embedding:

tests/test_loadding.py

@@ -16,14 +16,14 @@ def test_loadding():
     assert isinstance(vecs, np.ndarray)
 
 
-def test_2dmse_loadding():
+def test_ese_loadding():
     import numpy as np
     from angle_emb import AnglE
 
-    angle = AnglE.from_pretrained('WhereIsAI/UAE-Large-V1')
-    vecs = angle.encode('hello world', layer_index=20)
+    angle = AnglE.from_pretrained('WhereIsAI/UAE-Large-V1').truncate_layer(20)
+    vecs = angle.encode('hello world')
     assert isinstance(vecs, np.ndarray)
-    vecs = angle.encode(['hello world', 'hi there👋'], layer_index=20, embedding_size=512)
+    vecs = angle.encode(['hello world', 'hi there👋'], embedding_size=512)
     assert isinstance(vecs, np.ndarray)