vlm: increase the default max_num_batched_tokens for multimodal mod…

…els (#973)

aphrodite/common/config.py

@@ -35,6 +35,7 @@
 APHRODITE_USE_MODELSCOPE = envs.APHRODITE_USE_MODELSCOPE
 
 _EMBEDDING_MODEL_MAX_NUM_BATCHED_TOKENS = 32768
+_MULTIMODAL_MODEL_MAX_NUM_BATCHED_TOKENS = 4096
 
 _PP_SUPPORTED_MODELS = [
     "AquilaModel",
@@ -722,6 +723,10 @@ def is_embedding_model(self) -> bool:
         """Extract the embedding model flag."""
         return self.embedding_mode
 
+    @property
+    def is_multimodal_model(self) -> bool:
+        return self.multimodal_config is not None
+
 
 class CacheConfig:
     """Configuration for the KV cache.
@@ -1118,28 +1123,35 @@ def __init__(self,
                  num_lookahead_slots: int = 0,
                  delay_factor: float = 0.0,
                  enable_chunked_prefill: bool = False,
-                 embedding_mode: Optional[bool] = False,
+                 embedding_mode: bool = False,
+                 is_multimodal_model: bool = False,
                  preemption_mode: Optional[str] = None,
                  num_scheduler_steps: int = 1,
                  send_delta_data: bool = False,
                  single_user_mode: bool = False) -> None:
-        if max_num_batched_tokens is not None:
-            self.max_num_batched_tokens = max_num_batched_tokens
-        else:
+        if max_num_batched_tokens is None:
             if enable_chunked_prefill:
-                if not HAS_TRITON:
-                    raise ValueError("Triton is not installed, "
-                                     "chunked prefill will not work.")
-                # For chunked prefill, choose the well-tuned batch size.
-                self.max_num_batched_tokens = 768
-            elif embedding_mode:
-                # For embedding, choose specific value for higher throughput
-                self.max_num_batched_tokens = max(
-                    max_model_len, _EMBEDDING_MODEL_MAX_NUM_BATCHED_TOKENS)
+                # It is the values that have the best balance between ITL
+                # and TTFT on A100. Note it is not optimized for throughput.
+                max_num_batched_tokens = 512
             else:
                 # If max_model_len is too short, use 2048 as the default value
                 # for higher throughput.
-                self.max_num_batched_tokens = max(max_model_len, 2048)
+                max_num_batched_tokens = max(max_model_len, 2048)
+            if embedding_mode:
+                # For embedding, choose specific value for higher throughput
+                max_num_batched_tokens = max(
+                    max_num_batched_tokens,
+                    _EMBEDDING_MODEL_MAX_NUM_BATCHED_TOKENS,
+                )
+            if is_multimodal_model:
+                # The value needs to be at least the number of multimodal tokens
+                max_num_batched_tokens = max(
+                    max_num_batched_tokens,
+                    _MULTIMODAL_MODEL_MAX_NUM_BATCHED_TOKENS,
+                )
+        self.max_num_batched_tokens = max_num_batched_tokens
+
         if enable_chunked_prefill:
             logger.info(
                 "Chunked prefill is enabled with "

aphrodite/engine/aphrodite_engine.py

@@ -1849,7 +1849,7 @@ def _validate_model_inputs(self, inputs: Union[LLMInputs,
             prompt_ids = inputs.get("prompt_token_ids")
         if prompt_ids is None or len(prompt_ids) == 0:
             raise ValueError("Prompt cannot be empty")
-        if self.model_config.multimodal_config is not None:
+        if self.model_config.is_multimodal_model:
             max_prompt_len = self.model_config.max_model_len
             if len(prompt_ids) > max_prompt_len:
                 raise ValueError(
@@ -1859,6 +1859,9 @@ def _validate_model_inputs(self, inputs: Union[LLMInputs,
                     "number of text tokens plus multimodal tokens. For image "
                     "inputs, the number of image tokens depends on the number "
                     "of images, and possibly their aspect ratios as well.")
+            # TODO: Find out how many placeholder tokens are there so we can
+            # check that chunked prefill does not truncate them
+            # max_batch_len = self.scheduler_config.max_num_batched_tokens
 
 
 setup_logger()

aphrodite/engine/args_tools.py

@@ -1061,6 +1061,7 @@ def create_engine_config(self, ) -> EngineConfig:
             delay_factor=self.scheduler_delay_factor,
             enable_chunked_prefill=self.enable_chunked_prefill,
             embedding_mode=model_config.embedding_mode,
+            is_multimodal_model=model_config.is_multimodal_model,
             preemption_mode=self.preemption_mode,
             num_scheduler_steps=self.num_scheduler_steps,
             send_delta_data=(APHRODITE_USE_RAY_SPMD_WORKER and

aphrodite/task_handler/utils.py

@@ -39,7 +39,7 @@ def assert_enc_dec_mr_supported_scenario(
         raise NotImplementedError(
             STR_NOT_IMPL_ENC_DEC_ERR_STRS['STR_NOT_IMPL_ENC_DEC_PP'])
 
-    if enc_dec_mr.model_config.multimodal_config is not None:
+    if enc_dec_mr.model_config.is_multimodal_model:
         raise NotImplementedError(
             STR_NOT_IMPL_ENC_DEC_ERR_STRS['STR_NOT_IMPL_ENC_DEC_MM'])