Minor

cacheflow/master/block_manager.py

@@ -7,7 +7,7 @@ from cacheflow.sequence import SequenceStatus
 from cacheflow.utils import Device
 
 
-class BlockManager:
+class BlockAllocator:
 
     def __init__(
         self,
@@ -65,8 +65,8 @@ class BlockSpaceManager:
         self.num_total_gpu_blocks = num_gpu_blocks
         self.num_total_cpu_blocks = num_cpu_blocks
 
-        self.gpu_allocator = BlockManager(Device.GPU, block_size, num_gpu_blocks)
-        self.cpu_allocator = BlockManager(Device.CPU, block_size, num_cpu_blocks)
+        self.gpu_allocator = BlockAllocator(Device.GPU, block_size, num_gpu_blocks)
+        self.cpu_allocator = BlockAllocator(Device.CPU, block_size, num_cpu_blocks)
 
         # Mapping: seq_id -> BlockTable.
         self.block_tables: Dict[int, BlockTable] = {}

cacheflow/models/sample.py

@@ -8,6 +8,7 @@ from cacheflow.sampling_params import SamplingParams
 from cacheflow.sequence import SequenceOutputs
 from cacheflow.parallel_utils.tensor_parallel import gather_from_tensor_model_parallel_region
 
+
 class Sampler(nn.Module):
 
     def __init__(self) -> None:

cacheflow/sequence.py

@@ -30,7 +30,7 @@ class Sequence:
 
         self.status = SequenceStatus.PENDING
         self.output_logprobs: List[Dict[int, float]] = []
-        self.cumulative_logprobs = 1.0
+        self.cumulative_logprobs = 0.0
 
     def add_block(self) -> None:
         block = LogicalTokenBlock(