[Bugfix] More type hint fixes for py 3.8

vllm/executor/cpu_executor.py

@@ -1,5 +1,5 @@
 import os
-from typing import Dict, List, Optional
+from typing import Dict, List, Optional, Tuple
 
 import torch
 
@@ -61,7 +61,7 @@ def _init_worker(self):
         self.driver_worker.init_device()
         self.driver_worker.load_model()
 
-    def determine_num_available_blocks(self) -> tuple[int, int]:
+    def determine_num_available_blocks(self) -> Tuple[int, int]:
         """Determine the number of available KV blocks by invoking the
         underlying worker.
         """

vllm/executor/executor_base.py

@@ -1,5 +1,5 @@
 from abc import ABC, abstractmethod
-from typing import Dict, List, Optional
+from typing import Dict, List, Optional, Tuple
 
 from vllm.config import (CacheConfig, DeviceConfig, LoRAConfig, ModelConfig,
                          ParallelConfig, SchedulerConfig, SpeculativeConfig,
@@ -31,15 +31,15 @@ def __init__(
         raise NotImplementedError
 
     @abstractmethod
-    def determine_num_available_blocks(self) -> tuple[int, int]:
+    def determine_num_available_blocks(self) -> Tuple[int, int]:
         """Determine the number of available blocks for the GPU KV cache and
         swappable CPU KV cache.
 
         Normally, this should simply delegate to the underlying Worker. Some
         ExecutorBase may require modification of the result, e.g. to ensure the
         selected cache sizes are compatible with all workers.
 
-        Returns a tuple[num_gpu_blocks, num_cpu_blocks], where num_gpu_blocks
+        Returns a Tuple[num_gpu_blocks, num_cpu_blocks], where num_gpu_blocks
         are blocks that are "active" on the device and can be appended to.
         num_cpu_blocks refers to "swapped" blocks in CPU memory and cannot be
         appended to.

vllm/executor/gpu_executor.py

@@ -1,4 +1,4 @@
-from typing import Dict, List, Optional
+from typing import Dict, List, Optional, Tuple
 
 from vllm.config import (CacheConfig, DeviceConfig, LoRAConfig, ModelConfig,
                          ParallelConfig, SchedulerConfig, SpeculativeConfig,
@@ -66,7 +66,7 @@ def _init_worker(self):
         self.driver_worker.init_device()
         self.driver_worker.load_model()
 
-    def determine_num_available_blocks(self) -> tuple[int, int]:
+    def determine_num_available_blocks(self) -> Tuple[int, int]:
         """Determine the number of available KV blocks by invoking the
         underlying worker.
         """

vllm/executor/neuron_executor.py

@@ -1,4 +1,4 @@
-from typing import Dict, List, Optional
+from typing import Dict, List, Optional, Tuple
 
 from vllm.config import (CacheConfig, DeviceConfig, LoRAConfig, ModelConfig,
                          ParallelConfig, SchedulerConfig, SpeculativeConfig,
@@ -47,7 +47,7 @@ def _init_worker(self):
         self.driver_worker.init_device()
         self.driver_worker.load_model()
 
-    def determine_num_available_blocks(self) -> tuple[int, int]:
+    def determine_num_available_blocks(self) -> Tuple[int, int]:
         """Determine the number of available KV blocks by invoking the
         underlying worker.
         """

vllm/executor/ray_gpu_executor.py

@@ -3,7 +3,7 @@
 import os
 import pickle
 from collections import defaultdict
-from typing import TYPE_CHECKING, Any, Dict, List, Optional
+from typing import TYPE_CHECKING, Any, Dict, List, Optional, Tuple
 
 from vllm.config import (CacheConfig, DeviceConfig, LoRAConfig, ModelConfig,
                          ParallelConfig, SchedulerConfig, SpeculativeConfig,
@@ -197,15 +197,15 @@ def _init_workers_ray(self, placement_group: "PlacementGroup",
             max_parallel_loading_workers,
         )
 
-    def determine_num_available_blocks(self) -> tuple[int, int]:
+    def determine_num_available_blocks(self) -> Tuple[int, int]:
         """Determine the number of available KV blocks.
 
         This invokes `determine_num_available_blocks` on each worker and takes
         the min of the results, guaranteeing that the selected cache sizes are
         compatible with all workers.
 
         Returns:
-            - tuple[num_gpu_blocks, num_cpu_blocks]
+            - Tuple[num_gpu_blocks, num_cpu_blocks]
         """
         # Get the maximum number of blocks that can be allocated on GPU and CPU.
         num_blocks = self._run_workers("determine_num_available_blocks", )

vllm/worker/cpu_worker.py

@@ -1,5 +1,5 @@
 """A CPU worker class."""
-from typing import Dict, List, Optional
+from typing import Dict, List, Optional, Tuple
 
 import torch
 import torch.distributed
@@ -157,7 +157,7 @@ def init_device(self) -> None:
     def load_model(self):
         self.model_runner.load_model()
 
-    def determine_num_available_blocks(self) -> tuple[int, int]:
+    def determine_num_available_blocks(self) -> Tuple[int, int]:
         """Determine the number of blocks available for the KV cache.
 
         This determines how many KV blocks can fit into the configured CPU

vllm/worker/neuron_worker.py

@@ -1,5 +1,5 @@
 """A Neuron worker class."""
-from typing import List, Optional
+from typing import List, Optional, Tuple
 
 import torch
 import torch.distributed
@@ -40,7 +40,7 @@ def init_device(self) -> None:
     def load_model(self):
         self.model_runner.load_model()
 
-    def determine_num_available_blocks(self) -> tuple[int, int]:
+    def determine_num_available_blocks(self) -> Tuple[int, int]:
         """Determine the number of available KV blocks.
 
         Swapping is not yet supported, so always return num_cpu_blocks=0.

vllm/worker/worker_base.py

@@ -1,5 +1,5 @@
 from abc import ABC, abstractmethod
-from typing import Dict, List
+from typing import Dict, List, Tuple
 
 from vllm.lora.request import LoRARequest
 from vllm.sequence import SamplerOutput, SequenceGroupMetadata
@@ -18,14 +18,14 @@ def init_device(self) -> None:
         raise NotImplementedError
 
     @abstractmethod
-    def determine_num_available_blocks(self) -> tuple[int, int]:
+    def determine_num_available_blocks(self) -> Tuple[int, int]:
         """Determine the number of available blocks for the GPU KV cache and
         swappable CPU KV cache.
 
         The implementation may run profiling or other heuristics to determine
         the size of caches.
 
-        Returns a tuple[num_gpu_blocks, num_cpu_blocks], where num_gpu_blocks
+        Returns a Tuple[num_gpu_blocks, num_cpu_blocks], where num_gpu_blocks
         are blocks that are "active" on the device and can be appended to.
         num_cpu_blocks refers to "swapped" blocks in CPU memory and cannot be
         appended to.