fix Blip2加载和推理bug #1902 #1904 #1905 (#1958)

mindnlp/transformers/generation/utils.py

@@ -1297,6 +1297,13 @@ def _prepare_generated_length(
             and not self.config.is_encoder_decoder
         ):
             generation_config.max_length -= inputs_tensor.shape[1]
+        # by default let's always generate 20 new tokens
+        elif has_default_max_length:
+            if generation_config.max_length == GenerationConfig().max_length:
+                generation_config.max_length = generation_config.max_length + input_ids_length
+                max_position_embeddings = getattr(self.config, "max_position_embeddings", None)
+                if max_position_embeddings is not None:
+                    generation_config.max_length = min(generation_config.max_length, max_position_embeddings)
 
         # same for min length
         if generation_config.min_new_tokens is not None:

mindnlp/transformers/models/blip_2/modeling_blip_2.py

@@ -2881,31 +2881,45 @@ def generate(
             *language_model_inputs.shape[:-1], dtype=mindspore.int64
         )
         if input_ids is None:
-            input_ids = (
-                mindspore.Tensor([[self.config.text_config.bos_token_id]])
-                .repeat(batch_size, 1)
-            )
+            start_tokens = [self.config.text_config.bos_token_id]
+            if getattr(self.config, "image_token_index", None) is not None:
+                start_tokens = [self.config.image_token_index] * self.config.num_query_tokens + start_tokens
+            input_ids = ops.tile(mindspore.Tensor([start_tokens]), (batch_size, 1))
+
+        inputs_embeds = self.get_input_embeddings()(input_ids)
         if attention_mask is None:
             attention_mask = ops.ones_like(input_ids)
-        attention_mask = ops.cat([language_attention_mask, attention_mask], dim=1)
 
-        # concatenate query embeddings with prompt embeddings
-        inputs_embeds = self.get_input_embeddings()(input_ids)
-        inputs_embeds = ops.cat([language_model_inputs, inputs_embeds], dim=1)
+        # if the model already has "image_token_index" then the input is expanded to account for image embeds
+        # otherwise we expand manually by concatenating
+        if getattr(self.config, "image_token_index", None) is not None:
+            special_image_mask = (input_ids == self.config.image_token_index).unsqueeze(-1).expand_as(inputs_embeds)
+            inputs_embeds[special_image_mask] = language_model_inputs.flatten()
+        else:
+            logger.warning_once(
+                "Expanding inputs for image tokens in BLIP-2 should be done in processing. "
+                "Please follow instruction here (https://gist.github.com/zucchini-nlp/e9f20b054fa322f84ac9311d9ab67042) to update your BLIP-2 model. "
+                "Using processors without these attributes in the config is deprecated and will throw an error in v4.50."
+            )
+            inputs_embeds = ops.cat([language_model_inputs, inputs_embeds], dim=1)
+            attention_mask = ops.cat(
+                [language_attention_mask, attention_mask], dim=1
+            )
 
-        # add image_embeds length to max_length, so that the final max_length in counted only on token embeds
-        # -1 is to account for the prepended BOS after `generate.`
-        # TODO (joao, raushan): refactor `generate` to avoid these operations with VLMs
-        if not self.language_model.config.is_encoder_decoder:
-            generate_kwargs["max_length"] = generate_kwargs.get("max_length", 20) + language_model_inputs.shape[1] - 1
-            generate_kwargs["min_length"] = generate_kwargs.get("min_length", 0) + language_model_inputs.shape[1]
+            # add image_embeds length to max_length, so that the final max_length in counted only on token embeds
+            # -1 is to account for the prepended BOS after `generate.`
+            # TODO (joao, raushan): refactor `generate` to avoid these operations with VLMs
+            if not self.language_model.config.is_encoder_decoder:
+                generate_kwargs["max_length"] = (
+                    generate_kwargs.get("max_length", 20) + language_model_inputs.shape[1] - 1
+                )
+                generate_kwargs["min_length"] = generate_kwargs.get("min_length", 0) + language_model_inputs.shape[1]
 
-        outputs = self.language_model.generate(
-            inputs_embeds=inputs_embeds,
-            attention_mask=attention_mask,
-            **generate_kwargs,
-        )
+        inputs = {"inputs_embeds": inputs_embeds, "attention_mask": attention_mask}
+        if not self.language_model.config.is_encoder_decoder:
+            inputs["input_ids"] = input_ids
 
+        outputs = self.language_model.generate(**inputs, **generate_kwargs)
         return outputs
 
 __all__ = [

mindnlp/transformers/models/blip_2/processing_blip_2.py

@@ -20,7 +20,7 @@
 
 from ...image_utils import ImageInput
 from ...processing_utils import ProcessorMixin
-from ...tokenization_utils_base import BatchEncoding, PaddingStrategy, PreTokenizedInput, TextInput, TruncationStrategy
+from ...tokenization_utils_base import AddedToken, BatchEncoding, PaddingStrategy, PreTokenizedInput, TextInput, TruncationStrategy
 from ....utils import TensorType
 
 
@@ -36,13 +36,16 @@ class Blip2Processor(ProcessorMixin):
             An instance of [`BlipImageProcessor`]. The image processor is a required input.
         tokenizer (`AutoTokenizer`):
             An instance of ['PreTrainedTokenizer`]. The tokenizer is a required input.
+        num_query_tokens (`int`, *optional*):
+            Number of tokens used by the Qformer as queries, should be same as in model's config.
     """
     attributes = ["image_processor", "tokenizer"]
+    valid_kwargs = ["num_query_tokens"]
     image_processor_class = "BlipImageProcessor"
     tokenizer_class = "AutoTokenizer"
 
     # Copied from transformers.models.blip.processing_blip.BlipProcessor.__init__
-    def __init__(self, image_processor, tokenizer):
+    def __init__(self, image_processor, tokenizer, num_query_tokens=None, **kwargs):
         """
         Initializes a new instance of the Blip2Processor class.
         
@@ -53,16 +56,24 @@ def __init__(self, image_processor, tokenizer):
             tokenizer: An object representing the tokenizer to be used.
                 t should have the necessary methods and attributes required for tokenization.
                 The 'return_token_type_ids' attribute of the tokenizer will be set to False.
-        
+            num_query_tokens (`int`, *optional*):
+                Number of tokens used by the Qformer as queries, should be same as in model's config.
         Returns:
             None.
         
         Raises:
             None.
         """
         tokenizer.return_token_type_ids = False
+        self.current_processor = image_processor
+        if not hasattr(tokenizer, "image_token"):
+            self.image_token = AddedToken("<image>", normalized=False, special=True)
+            tokenizer.add_tokens([self.image_token], special_tokens=True)
+        else:
+            self.image_token = tokenizer.image_token
+        self.num_query_tokens = num_query_tokens
+
         super().__init__(image_processor, tokenizer)
-        self.current_processor = self.image_processor
 
     # Copied from transformers.models.blip.processing_blip.BlipProcessor.__call__
     def __call__(