Add supports_gradient_checkpointing

configuration_internvl_chat.py

@@ -64,6 +64,8 @@ class InternVLChatConfig(PretrainedConfig):
         self.ps_version = ps_version  # pixel shuffle version
         self.min_dynamic_patch = min_dynamic_patch
         self.max_dynamic_patch = max_dynamic_patch
+        # By default, we use tie_word_embeddings=False for models of all sizes.
+        self.tie_word_embeddings = self.llm_config.tie_word_embeddings
 
         logger.info(f'vision_select_layer: {self.select_layer}')
         logger.info(f'ps_version: {self.ps_version}')

modeling_intern_vit.py

@@ -364,6 +364,7 @@ class InternVisionEncoder(nn.Module):
 class InternVisionModel(PreTrainedModel):
     main_input_name = 'pixel_values'
     _supports_flash_attn_2 = True
+    supports_gradient_checkpointing = True
     config_class = InternVisionConfig
     _no_split_modules = ['InternVisionEncoderLayer']
 

modeling_internvl_chat.py

@@ -38,6 +38,7 @@ class InternVLChatModel(PreTrainedModel):
     main_input_name = 'pixel_values'
     base_model_prefix = 'language_model'
     _supports_flash_attn_2 = True
+    supports_gradient_checkpointing = True
     _no_split_modules = ['InternVisionModel', 'LlamaDecoderLayer', 'Phi3DecoderLayer']
 
     def __init__(self, config: InternVLChatConfig, vision_model=None, language_model=None, use_flash_attn=True):
@@ -347,3 +348,13 @@ class InternVLChatModel(PreTrainedModel):
         )
 
         return outputs
+
+    @property
+    def lm_head(self):
+        return self.language_model.get_output_embeddings()
+
+    def get_input_embeddings(self):
+        return self.language_model.get_input_embeddings()
+
+    def get_output_embeddings(self):
+        return self.language_model.get_output_embeddings()

modeling_phi3.py

@@ -1336,7 +1336,7 @@ class Phi3ForCausalLM(Phi3PreTrainedModel):
             if isinstance(past_key_values, Cache):
                 cache_length = past_key_values.get_seq_length()
                 past_length = past_key_values.seen_tokens
-                max_cache_length = past_key_values.get_max_length() if hasattr(past_key_values, "get_max_length") else past_key_values.get_max_cache_shape()
+                max_cache_length = past_key_values.get_max_length()
             else:
                 cache_length = past_length = past_key_values[0][0].shape[2]
                 max_cache_length = None