use block size 128 or 64 when possible

serve/mlc_serve/model/tvm_model.py

@@ -461,10 +461,23 @@ def init_tvm_model(
 ) -> Tuple[TextGenerator, CacheManager]:
     dev = tvm.device("cuda", 0)
 
+    num_kv_heads = (
+        model_artifact_config.num_key_value_heads // model_artifact_config.num_shards
+    )
+    head_size = (
+        model_artifact_config.hidden_size // model_artifact_config.num_attention_heads
+    )
+
     if model_artifact_config.paged_kv_cache_type == "flash-decoding":
         allocate_func_name = "tvm.contrib.flash_attn.allocate_kv_cache"
         copy_blocks_func_name = "tvm.contrib.flash_attn.copy_blocks"
-        block_size = 256
+        # This needs to match with the model definition in llama_batched_vllm.py
+        if head_size <= 64:
+            block_size = 256
+        elif head_size <= 128:
+            block_size = 128
+        else:
+            block_size = 64
     else:
         allocate_func_name = "tvm.contrib.vllm.allocate_kv_cache"
         copy_blocks_func_name = "tvm.contrib.vllm.copy_blocks"
@@ -475,13 +488,6 @@ def init_tvm_model(
     if model_artifact_config.num_shards > 1:
         model.disco_session.sync_worker_0()
 
-    num_kv_heads = (
-        model_artifact_config.num_key_value_heads // model_artifact_config.num_shards
-    )
-    head_size = (
-        model_artifact_config.hidden_size // model_artifact_config.num_attention_heads
-    )
-
     if engine_config.max_num_batched_tokens > 0:
         LOG.info("Running memory profiling.")
         num_blocks = get_num_cache_blocks(