[Bugfix] More type hint fixes for py 3.8 (#4039)

vllm/executor/executor_base.py

@@ -39,7 +39,7 @@ class ExecutorBase(ABC):
         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/worker/cpu_worker.py

 """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 @@ class CPUWorker(LoraNotSupportedWorkerBase):
     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

 """A Neuron worker class."""
-from typing import List, Optional
+from typing import List, Optional, Tuple
 
 import torch
 import torch.distributed
@@ -40,7 +40,7 @@ class NeuronWorker(LoraNotSupportedWorkerBase):
     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

 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 @@ class WorkerBase(ABC):
         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.