Add ConvBert model and pass ut. (#991)

mindnlp/transformers/models/__init__.py

@@ -44,6 +44,7 @@
     byt5,
     clip,
     codegen,
+    convbert,
     cpm,
     cpmant,
     cpmbee,
@@ -131,6 +132,7 @@
 from .byt5 import *
 from .clip import *
 from .codegen import *
+from .convbert import *
 from .cpm import *
 from .cpmant import *
 from .cpmbee import *
@@ -218,6 +220,7 @@
 __all__.extend(byt5.__all__)
 __all__.extend(clip.__all__)
 __all__.extend(codegen.__all__)
+__all__.extend(convbert.__all__)
 __all__.extend(cpm.__all__)
 __all__.extend(cpmant.__all__)
 __all__.extend(cpmbee.__all__)

mindnlp/transformers/models/auto/modeling_auto.py

@@ -55,8 +55,9 @@
         ("chatglm", "ChatGLMModel"),
         ("clip", "CLIPModel"),
         ("clip_vision_model", "CLIPVisionModel"),
+        ("convbert", "ConvBertModel"),
         ("deberta", "DebertaModel"),
-        ('encodec','EncodecModel'),
+        ('encodec', 'EncodecModel'),
         ("esm", "EsmModel"),
         ("ernie", "ErnieModel"),
         ("ernie_m", "ErnieMModel"),
@@ -69,7 +70,7 @@
         ("layoutlmv2", "LayoutLMv2Model"),
         ("longformer", "LongformerModel"),
         ("mamba", "MambaModel"),
-        ('mbart','MBartModel'),
+        ('mbart', 'MBartModel'),
         ('minicpm', 'MiniCPMModel'),
         ("mistral", "MistralModel"),
         ("mixtral", "MixtralModel"),
@@ -797,9 +798,12 @@
 )
 
 MODEL_MAPPING = _LazyAutoMapping(CONFIG_MAPPING_NAMES, MODEL_MAPPING_NAMES)
-MODEL_FOR_PRETRAINING_MAPPING = _LazyAutoMapping(CONFIG_MAPPING_NAMES, MODEL_FOR_PRETRAINING_MAPPING_NAMES)
-MODEL_WITH_LM_HEAD_MAPPING = _LazyAutoMapping(CONFIG_MAPPING_NAMES, MODEL_WITH_LM_HEAD_MAPPING_NAMES)
-MODEL_FOR_CAUSAL_LM_MAPPING = _LazyAutoMapping(CONFIG_MAPPING_NAMES, MODEL_FOR_CAUSAL_LM_MAPPING_NAMES)
+MODEL_FOR_PRETRAINING_MAPPING = _LazyAutoMapping(
+    CONFIG_MAPPING_NAMES, MODEL_FOR_PRETRAINING_MAPPING_NAMES)
+MODEL_WITH_LM_HEAD_MAPPING = _LazyAutoMapping(
+    CONFIG_MAPPING_NAMES, MODEL_WITH_LM_HEAD_MAPPING_NAMES)
+MODEL_FOR_CAUSAL_LM_MAPPING = _LazyAutoMapping(
+    CONFIG_MAPPING_NAMES, MODEL_FOR_CAUSAL_LM_MAPPING_NAMES)
 
 MODEL_FOR_IMAGE_CLASSIFICATION_MAPPING = _LazyAutoMapping(
     CONFIG_MAPPING_NAMES, MODEL_FOR_IMAGE_CLASSIFICATION_MAPPING_NAMES
@@ -818,20 +822,24 @@
 MODEL_FOR_VIDEO_CLASSIFICATION_MAPPING = _LazyAutoMapping(
     CONFIG_MAPPING_NAMES, MODEL_FOR_VIDEO_CLASSIFICATION_MAPPING_NAMES
 )
-MODEL_FOR_VISION_2_SEQ_MAPPING = _LazyAutoMapping(CONFIG_MAPPING_NAMES, MODEL_FOR_VISION_2_SEQ_MAPPING_NAMES)
+MODEL_FOR_VISION_2_SEQ_MAPPING = _LazyAutoMapping(
+    CONFIG_MAPPING_NAMES, MODEL_FOR_VISION_2_SEQ_MAPPING_NAMES)
 MODEL_FOR_VISUAL_QUESTION_ANSWERING_MAPPING = _LazyAutoMapping(
     CONFIG_MAPPING_NAMES, MODEL_FOR_VISUAL_QUESTION_ANSWERING_MAPPING_NAMES
 )
 MODEL_FOR_DOCUMENT_QUESTION_ANSWERING_MAPPING = _LazyAutoMapping(
     CONFIG_MAPPING_NAMES, MODEL_FOR_DOCUMENT_QUESTION_ANSWERING_MAPPING_NAMES
 )
-MODEL_FOR_MASKED_LM_MAPPING = _LazyAutoMapping(CONFIG_MAPPING_NAMES, MODEL_FOR_MASKED_LM_MAPPING_NAMES)
+MODEL_FOR_MASKED_LM_MAPPING = _LazyAutoMapping(
+    CONFIG_MAPPING_NAMES, MODEL_FOR_MASKED_LM_MAPPING_NAMES)
 
-MODEL_FOR_OBJECT_DETECTION_MAPPING = _LazyAutoMapping(CONFIG_MAPPING_NAMES, MODEL_FOR_OBJECT_DETECTION_MAPPING_NAMES)
+MODEL_FOR_OBJECT_DETECTION_MAPPING = _LazyAutoMapping(
+    CONFIG_MAPPING_NAMES, MODEL_FOR_OBJECT_DETECTION_MAPPING_NAMES)
 MODEL_FOR_ZERO_SHOT_OBJECT_DETECTION_MAPPING = _LazyAutoMapping(
     CONFIG_MAPPING_NAMES, MODEL_FOR_ZERO_SHOT_OBJECT_DETECTION_MAPPING_NAMES
 )
-MODEL_FOR_DEPTH_ESTIMATION_MAPPING = _LazyAutoMapping(CONFIG_MAPPING_NAMES, MODEL_FOR_DEPTH_ESTIMATION_MAPPING_NAMES)
+MODEL_FOR_DEPTH_ESTIMATION_MAPPING = _LazyAutoMapping(
+    CONFIG_MAPPING_NAMES, MODEL_FOR_DEPTH_ESTIMATION_MAPPING_NAMES)
 MODEL_FOR_SEQ_TO_SEQ_CAUSAL_LM_MAPPING = _LazyAutoMapping(
     CONFIG_MAPPING_NAMES, MODEL_FOR_SEQ_TO_SEQ_CAUSAL_LM_MAPPING_NAMES
 )
@@ -847,40 +855,51 @@
 MODEL_FOR_TOKEN_CLASSIFICATION_MAPPING = _LazyAutoMapping(
     CONFIG_MAPPING_NAMES, MODEL_FOR_TOKEN_CLASSIFICATION_MAPPING_NAMES
 )
-MODEL_FOR_MULTIPLE_CHOICE_MAPPING = _LazyAutoMapping(CONFIG_MAPPING_NAMES, MODEL_FOR_MULTIPLE_CHOICE_MAPPING_NAMES)
+MODEL_FOR_MULTIPLE_CHOICE_MAPPING = _LazyAutoMapping(
+    CONFIG_MAPPING_NAMES, MODEL_FOR_MULTIPLE_CHOICE_MAPPING_NAMES)
 MODEL_FOR_NEXT_SENTENCE_PREDICTION_MAPPING = _LazyAutoMapping(
     CONFIG_MAPPING_NAMES, MODEL_FOR_NEXT_SENTENCE_PREDICTION_MAPPING_NAMES
 )
 MODEL_FOR_AUDIO_CLASSIFICATION_MAPPING = _LazyAutoMapping(
     CONFIG_MAPPING_NAMES, MODEL_FOR_AUDIO_CLASSIFICATION_MAPPING_NAMES
 )
-MODEL_FOR_CTC_MAPPING = _LazyAutoMapping(CONFIG_MAPPING_NAMES, MODEL_FOR_CTC_MAPPING_NAMES)
-MODEL_FOR_SPEECH_SEQ_2_SEQ_MAPPING = _LazyAutoMapping(CONFIG_MAPPING_NAMES, MODEL_FOR_SPEECH_SEQ_2_SEQ_MAPPING_NAMES)
+MODEL_FOR_CTC_MAPPING = _LazyAutoMapping(
+    CONFIG_MAPPING_NAMES, MODEL_FOR_CTC_MAPPING_NAMES)
+MODEL_FOR_SPEECH_SEQ_2_SEQ_MAPPING = _LazyAutoMapping(
+    CONFIG_MAPPING_NAMES, MODEL_FOR_SPEECH_SEQ_2_SEQ_MAPPING_NAMES)
 MODEL_FOR_AUDIO_FRAME_CLASSIFICATION_MAPPING = _LazyAutoMapping(
     CONFIG_MAPPING_NAMES, MODEL_FOR_AUDIO_FRAME_CLASSIFICATION_MAPPING_NAMES
 )
-MODEL_FOR_AUDIO_XVECTOR_MAPPING = _LazyAutoMapping(CONFIG_MAPPING_NAMES, MODEL_FOR_AUDIO_XVECTOR_MAPPING_NAMES)
+MODEL_FOR_AUDIO_XVECTOR_MAPPING = _LazyAutoMapping(
+    CONFIG_MAPPING_NAMES, MODEL_FOR_AUDIO_XVECTOR_MAPPING_NAMES)
 
 MODEL_FOR_TEXT_TO_SPECTROGRAM_MAPPING = _LazyAutoMapping(
     CONFIG_MAPPING_NAMES, MODEL_FOR_TEXT_TO_SPECTROGRAM_MAPPING_NAMES
 )
 
-MODEL_FOR_TEXT_TO_WAVEFORM_MAPPING = _LazyAutoMapping(CONFIG_MAPPING_NAMES, MODEL_FOR_TEXT_TO_WAVEFORM_MAPPING_NAMES)
+MODEL_FOR_TEXT_TO_WAVEFORM_MAPPING = _LazyAutoMapping(
+    CONFIG_MAPPING_NAMES, MODEL_FOR_TEXT_TO_WAVEFORM_MAPPING_NAMES)
 
-MODEL_FOR_BACKBONE_MAPPING = _LazyAutoMapping(CONFIG_MAPPING_NAMES, MODEL_FOR_BACKBONE_MAPPING_NAMES)
+MODEL_FOR_BACKBONE_MAPPING = _LazyAutoMapping(
+    CONFIG_MAPPING_NAMES, MODEL_FOR_BACKBONE_MAPPING_NAMES)
 
-MODEL_FOR_MASK_GENERATION_MAPPING = _LazyAutoMapping(CONFIG_MAPPING_NAMES, MODEL_FOR_MASK_GENERATION_MAPPING_NAMES)
+MODEL_FOR_MASK_GENERATION_MAPPING = _LazyAutoMapping(
+    CONFIG_MAPPING_NAMES, MODEL_FOR_MASK_GENERATION_MAPPING_NAMES)
 
-MODEL_FOR_TEXT_ENCODING_MAPPING = _LazyAutoMapping(CONFIG_MAPPING_NAMES, MODEL_FOR_TEXT_ENCODING_MAPPING_NAMES)
+MODEL_FOR_TEXT_ENCODING_MAPPING = _LazyAutoMapping(
+    CONFIG_MAPPING_NAMES, MODEL_FOR_TEXT_ENCODING_MAPPING_NAMES)
 
-MODEL_FOR_IMAGE_TO_IMAGE_MAPPING = _LazyAutoMapping(CONFIG_MAPPING_NAMES, MODEL_FOR_IMAGE_TO_IMAGE_MAPPING_NAMES)
+MODEL_FOR_IMAGE_TO_IMAGE_MAPPING = _LazyAutoMapping(
+    CONFIG_MAPPING_NAMES, MODEL_FOR_IMAGE_TO_IMAGE_MAPPING_NAMES)
 
-MODEL_FOR_IMAGE_MAPPING = _LazyAutoMapping(CONFIG_MAPPING_NAMES, MODEL_FOR_IMAGE_MAPPING_NAMES)
+MODEL_FOR_IMAGE_MAPPING = _LazyAutoMapping(
+    CONFIG_MAPPING_NAMES, MODEL_FOR_IMAGE_MAPPING_NAMES)
 
 MODEL_FOR_SEMANTIC_SEGMENTATION_MAPPING = _LazyAutoMapping(
     CONFIG_MAPPING_NAMES, MODEL_FOR_SEMANTIC_SEGMENTATION_MAPPING_NAMES
 )
 
+
 class AutoModelForMaskGeneration(_BaseAutoModelClass):
     _model_mapping = MODEL_FOR_MASK_GENERATION_MAPPING
 
@@ -901,6 +920,8 @@ class AutoModelForPreTraining(_BaseAutoModelClass):
     _model_mapping = MODEL_FOR_PRETRAINING_MAPPING
 
 # Private on purpose, the public class will add the deprecation warnings.
+
+
 class _AutoModelWithLMHead(_BaseAutoModelClass):
     _model_mapping = MODEL_WITH_LM_HEAD_MAPPING
 
@@ -937,7 +958,6 @@ class AutoModelForDocumentQuestionAnswering(_BaseAutoModelClass):
     _model_mapping = MODEL_FOR_DOCUMENT_QUESTION_ANSWERING_MAPPING
 
 
-
 class AutoModelForTokenClassification(_BaseAutoModelClass):
     _model_mapping = MODEL_FOR_TOKEN_CLASSIFICATION_MAPPING
 

mindnlp/transformers/models/convbert/__init__.py

@@ -15,3 +15,14 @@
 """
 ConvBert Model.
 """
+from . import convbert, convbert_config, convbert_tokenizer, convbert_tokenizer_fast
+from .convbert import *
+from .convbert_config import *
+from .convbert_tokenizer import *
+from .convbert_tokenizer_fast import *
+
+__all__ = []
+__all__.extend(convbert.__all__)
+__all__.extend(convbert_config.__all__)
+__all__.extend(convbert_tokenizer.__all__)
+__all__.extend(convbert_tokenizer_fast.__all__)