Drop 0.11.0 support (#4377)

There is a lot hack code for v0.11.0, which makes the code hard to
upgrade to newer vLLM version. Since v0.11.0 will release soon. Let's
drop v0.11.0 support first. Then we'll upgrade to v0.11.2 soon.


- vLLM version: v0.11.0
- vLLM main:
2918c1b49c

Signed-off-by: wangxiyuan <wangxiyuan1007@gmail.com>

vllm_ascend/distributed/cpu_offload_manager/metadata.py

@@ -10,17 +10,12 @@ import vllm.envs as envs
 import zmq
 from vllm.config import KVTransferConfig, VllmConfig
 from vllm.utils import logger
+from vllm.utils.network_utils import make_zmq_socket
+from vllm.utils.torch_utils import get_dtype_size
 from vllm.v1.kv_cache_interface import AttentionSpec
 
 from vllm_ascend.distributed.cpu_offload_manager.cpu_kv_cache_manager import \
     CPUKVCacheManager
-from vllm_ascend.utils import vllm_version_is
-
-if vllm_version_is("0.11.0"):
-    from vllm.utils import get_dtype_size, make_zmq_socket
-else:
-    from vllm.utils.network_utils import make_zmq_socket
-    from vllm.utils.torch_utils import get_dtype_size
 
 
 @dataclass

vllm_ascend/distributed/llmdatadist_c_mgr_connector.py

@@ -33,17 +33,13 @@ from vllm.v1.request import Request, RequestStatus
 import vllm_ascend.envs as envs_ascend
 from vllm_ascend.distributed.utils import get_transfer_timeout_value
 from vllm_ascend.utils import (AscendSocVersion, get_ascend_soc_version,
-                               prefill_context_parallel_enable,
-                               vllm_version_is)
+                               prefill_context_parallel_enable)
 
 if prefill_context_parallel_enable():
     from vllm.distributed.parallel_state import \
         get_prefill_context_model_parallel_rank
 
-if vllm_version_is("0.11.0"):
-    from vllm.utils import get_ip
-else:
-    from vllm.utils.network_utils import get_ip
+from vllm.utils.network_utils import get_ip
 
 TORCH_DTYPE_TO_NPU_DTYPE = {
     torch.half: llm_datadist.DataType.DT_FLOAT16,

vllm_ascend/distributed/mooncake/config_data.py

@@ -10,14 +10,7 @@ import torch
 from vllm.distributed.kv_transfer.kv_connector.v1.base import \
     KVConnectorMetadata
 from vllm.utils import logger
-
-from vllm_ascend.utils import vllm_version_is
-
-if vllm_version_is("0.11.0"):
-    from vllm.utils import cdiv
-else:
-    from vllm.utils.math_utils import cdiv
-
+from vllm.utils.math_utils import cdiv
 from vllm.v1.core.sched.output import NewRequestData
 
 DEFAULT_GLOBAL_SEGMENT_SIZE = 3355443200  # 3.125 GiB

vllm_ascend/distributed/mooncake/mooncake_engine.py

@@ -8,6 +8,7 @@ from typing import Generator, List, Optional, Union
 import torch
 from vllm.config import VllmConfig
 from vllm.utils import logger
+from vllm.utils.torch_utils import get_kv_cache_torch_dtype
 
 from vllm_ascend.distributed.mooncake.config_data import (
     ChunkedTokenDatabase, LasyerMultiBlockReqMeta, MooncakeConnectorMetadata,
@@ -16,12 +17,6 @@ from vllm_ascend.distributed.mooncake.kv_transfer import (
     KVCacheStoreLayerRecvingThread, KVCacheStoreLayerSendingThread,
     KVCacheStoreRecvingThread, KVCacheStoreSendingThread, KVTransferThread)
 from vllm_ascend.distributed.mooncake.mooncake_store import Mooncakestore
-from vllm_ascend.utils import vllm_version_is
-
-if vllm_version_is("0.11.0"):
-    from vllm.utils import get_kv_cache_torch_dtype
-else:
-    from vllm.utils.torch_utils import get_kv_cache_torch_dtype
 
 
 class MooncakeEngine:

vllm_ascend/distributed/mooncake/mooncake_store.py

@@ -6,18 +6,13 @@ from mooncake.store import ReplicateConfig  # type: ignore
 from vllm.config import ParallelConfig
 from vllm.distributed.parallel_state import get_tensor_model_parallel_rank
 from vllm.utils import logger
+from vllm.utils.network_utils import get_ip
 
 from vllm_ascend.distributed.mooncake.config_data import MooncakeEngineKey
 from vllm_ascend.distributed.mooncake.transfer_engine import get_global_te
-from vllm_ascend.utils import vllm_version_is
 
 from .config_data import MooncakeStoreConfig
 
-if vllm_version_is("0.11.0"):
-    from vllm.utils import get_ip
-else:
-    from vllm.utils.network_utils import get_ip
-
 METADATA_BYTES_LEN = 24
 BASE_PORT = int(os.getenv("VLLM_BASE_PORT", "8790"))
 

vllm_ascend/distributed/mooncake/mooncake_store_connector_v1.py

@@ -10,6 +10,7 @@ from vllm.distributed.kv_transfer.kv_connector.v1.base import (
     KVConnectorBase_V1, KVConnectorMetadata, KVConnectorRole)
 from vllm.forward_context import ForwardContext
 from vllm.utils import logger
+from vllm.utils.network_utils import make_zmq_socket
 from vllm.v1.core.kv_cache_manager import KVCacheBlocks
 from vllm.v1.core.sched.output import SchedulerOutput
 from vllm.v1.request import Request
@@ -18,12 +19,6 @@ from vllm.v1.serial_utils import MsgpackDecoder, MsgpackEncoder
 from vllm_ascend.distributed.mooncake.config_data import (
     LoadSpec, MooncakeConnectorMetadata, ReqMeta, RequestTracker)
 from vllm_ascend.distributed.mooncake.mooncake_engine import MooncakeEngine
-from vllm_ascend.utils import vllm_version_is
-
-if vllm_version_is("0.11.0"):
-    from vllm.utils import make_zmq_socket
-else:
-    from vllm.utils.network_utils import make_zmq_socket
 
 
 class MooncakeConnectorV1(KVConnectorBase_V1):

vllm_ascend/distributed/mooncake_connector.py

@@ -37,7 +37,7 @@ import vllm_ascend.envs as envs_ascend
 from vllm_ascend.ascend_config import get_ascend_config, init_ascend_config
 from vllm_ascend.distributed.mooncake.transfer_engine import get_global_te
 from vllm_ascend.distributed.utils import get_transfer_timeout_value
-from vllm_ascend.utils import prefill_context_parallel_enable, vllm_version_is
+from vllm_ascend.utils import prefill_context_parallel_enable
 
 # isort: off
 if prefill_context_parallel_enable():
@@ -46,10 +46,7 @@ if prefill_context_parallel_enable():
                                   )
 # isort: on
 
-if vllm_version_is("0.11.0"):
-    from vllm.utils import get_ip, make_zmq_path, make_zmq_socket
-else:
-    from vllm.utils.network_utils import get_ip, make_zmq_path, make_zmq_socket
+from vllm.utils.network_utils import get_ip, make_zmq_path, make_zmq_socket
 
 if TYPE_CHECKING:
     from vllm.attention.backends.abstract import AttentionMetadata

vllm_ascend/distributed/mooncake_layerwise_connector.py

@@ -28,6 +28,7 @@ from vllm.distributed.kv_transfer.kv_connector.v1.base import (
 from vllm.distributed.parallel_state import (get_tensor_model_parallel_rank,
                                              get_tp_group, get_world_group)
 from vllm.utils import logger
+from vllm.utils.network_utils import get_ip, make_zmq_path, make_zmq_socket
 from vllm.v1.core.sched.output import SchedulerOutput
 
 import vllm_ascend.envs as envs_ascend
@@ -35,12 +36,6 @@ from vllm_ascend.ascend_config import get_ascend_config
 from vllm_ascend.distributed.utils import (align_memory,
                                            get_transfer_timeout_value,
                                            kv_alltoall_and_rearrange)
-from vllm_ascend.utils import vllm_version_is
-
-if vllm_version_is("0.11.0"):
-    from vllm.utils import get_ip, make_zmq_path, make_zmq_socket
-else:
-    from vllm.utils.network_utils import get_ip, make_zmq_path, make_zmq_socket
 
 if TYPE_CHECKING:
     from vllm.attention.backends.abstract import AttentionMetadata