format

Signed-off-by: Varun Sundar Rabindranath <[email protected]>

vllm/v1/core/kv_cache_manager.py

@@ -85,7 +85,8 @@ def get_computed_blocks(self, request: Request) -> List[KVCacheBlock]:
 
         # TODO(rickyx): potentially we could cache this so we don't have to
         # recompute it every time.
-        metadata_hash = None if not request.lora_request else hash(request.lora_request.lora_int_id)
+        metadata_hash = (None if not request.lora_request else
+                         request.lora_request.lora_int_id)
         block_hashes = hash_request_tokens(self.block_size,
                                            request.all_token_ids,
                                            parent_hash=metadata_hash)
@@ -379,8 +380,9 @@ def _cache_full_blocks(
             prev_block: The previous block in the chain.
         """
         # Update the new blocks with the block hashes through the chain.
-        metadata_hash = None if request.lora_request is None else request.lora_request.lora_int_id 
-        parent_hash = metadata_hash 
+        metadata_hash = (None if request.lora_request is None else
+                         request.lora_request.lora_int_id)
+        parent_hash = metadata_hash
         if prev_block is not None:
             # Previous block must have a block hash because it must be
             # a full, cached block.

vllm/v1/core/kv_cache_utils.py

@@ -182,9 +182,10 @@ def hash_block_tokens(parent_hash: Optional[int],
                          tuple(curr_block_token_ids))
 
 
-def hash_request_tokens(block_size: int,
-                        token_ids: Sequence[int],
-                        parent_hash: Optional[int] = None) -> List[BlockHashType]:
+def hash_request_tokens(
+        block_size: int,
+        token_ids: Sequence[int],
+        parent_hash: Optional[int] = None) -> List[BlockHashType]:
     """Computes hash values of a chain of blocks given a sequence of
     token IDs. The hash value is used for prefix caching.
 
@@ -204,8 +205,7 @@ def hash_request_tokens(block_size: int,
         # Do not hash the block if it is not full.
         if len(block_token_ids) < block_size:
             break
-        block_hash = hash_block_tokens(parent_hash,
-                                       block_token_ids)
+        block_hash = hash_block_tokens(parent_hash, block_token_ids)
         ret.append(block_hash)
         parent_hash = block_hash.hash_value
     return ret