Avoid sorting waiting queue & Minor code cleaning (#93)

cacheflow/core/scheduler.py

@@ -1,21 +1,16 @@
 import enum
-import os
-import pickle
 import time
-from typing import Any, Dict, List, Optional, Tuple
+from typing import Dict, List, Optional, Tuple
 
 from cacheflow.core.block_manager import BlockSpaceManager
 from cacheflow.logger import init_logger
 from cacheflow.core.policy import PolicyFactory
 from cacheflow.sampling_params import SamplingParams
-from cacheflow.sequence import Sequence
+from cacheflow.sequence import (Sequence, SequenceGroup, SequenceGroupMetadata,
-from cacheflow.sequence import SequenceGroup
+                                SequenceOutputs, SequenceStatus)
-from cacheflow.sequence import SequenceGroupMetadata
-from cacheflow.sequence import SequenceOutputs
-from cacheflow.sequence import SequenceStatus
-
 
 logger = init_logger(__name__)
+
 _LOGGING_INTERVAL_SEC = 10
 
 
@@ -129,7 +124,6 @@ class Scheduler:
 
         # Swap in the sequence groups in the SWAPPED state if possible.
         self.swapped = self.policy.sort_by_priority(now, self.swapped)
-        # FCFS
         while self.swapped and not blocks_to_swap_out:
             seq_group = self.swapped[0]
             # If the sequence group has been preempted in this step, stop.
@@ -162,7 +156,9 @@ class Scheduler:
         # This is because we want to bound the amount of CPU memory taken by
         # the swapped sequence groups.
         if not self.swapped:
-            self.waiting = self.policy.sort_by_priority(now, self.waiting)
+            # Optimization: We do not sort the waiting queue since the preempted
+            # sequence groups are added to the front and the new sequence groups
+            # are added to the back.
             while self.waiting:
                 seq_group = self.waiting[0]
                 # If the sequence group has been preempted in this step, stop.
@@ -347,7 +343,6 @@ class Scheduler:
         self.block_manager.allocate(seq_group)
         for seq in seq_group.seqs:
             seq.status = SequenceStatus.RUNNING
-        # FIXME(woosuk): Support interactive generation.
         if seq_group.group_id not in self.num_steps:
             self.num_steps[seq_group.group_id] = 0
 
@@ -404,7 +399,9 @@ class Scheduler:
         for seq in seqs:
             seq.status = SequenceStatus.WAITING
             self.block_manager.free(seq)
-        self.waiting.append(seq_group)
+        # NOTE: For FCFS, we insert the preempted sequence group to the front
+        # of the waiting queue.
+        self.waiting.insert(0, seq_group)
 
     def _preempt_by_swap(
         self,

cacheflow/core/server.py

@@ -17,7 +17,6 @@ from cacheflow.sequence import SequenceGroup
 from cacheflow.utils import get_gpu_memory, get_cpu_memory
 from cacheflow.worker.controller import Controller, DeviceID
 
-
 logger = init_logger(__name__)
 
 

cacheflow/frontend/simple_frontend.py

@@ -7,7 +7,6 @@ from cacheflow.sampling_params import SamplingParams
 from cacheflow.sequence import Sequence, SequenceGroup
 from cacheflow.utils import Counter
 
-
 logger = init_logger(__name__)
 
 

cacheflow/model_executor/memory_analyzer.py

@@ -4,7 +4,6 @@ from transformers import AutoConfig
 from cacheflow.logger import init_logger
 from cacheflow.model_executor.utils import get_dtype_size
 
-
 logger = init_logger(__name__)
 
 _GiB = 1 << 30