[Feature] [PD Disaggregation] simplify configuration for pd-disaggregated deployment, and refactor post-init and usage for all ports (#5415)

* [feat] simplify configuration for pd-disaggregated deployment, and refactor post-init and usage for all ports

* [fix] fix some bugs

* [fix] fix rdma port for cache manager/messager

* [fix] temporarily cancel port availability check to see if it can pass ci test

* [feat] simplify args for multi api server

* [fix] fix dp

* [fix] fix port for xpu

* [fix] add tests for ports post processing & fix ci

* [test] fix test_multi_api_server

* [fix] fix rdma_comm_ports args for multi_api_server

* [fix] fix test_common_engine

* [fix] fix test_cache_transfer_manager

* [chore] automatically setting FD_ENABLE_MULTI_API_SERVER

* [fix] avoid api server from creating engine_args twice

* [fix] fix test_run_batch

* [fix] fix test_metrics

* [fix] fix splitwise connector init

* [test] add test_rdma_transfer and test_expert_service

* [fix] fix code syntax

* [fix] fix test_rdma_transfer and build wheel with rdma script
This commit is contained in:
Yonghua Li
2025-12-17 15:50:42 +08:00
committed by GitHub
parent cdc0004894
commit 0c8c6369ed
34 changed files with 1323 additions and 409 deletions

View File

@@ -167,7 +167,7 @@ class PrefixCacheManager:
device_ids,
pod_ip,
engine_worker_queue_port,
pid_suffix,
ipc_suffix,
create_cache_tensor,
):
"""
@@ -184,7 +184,7 @@ class PrefixCacheManager:
)
self.cache_task_queue = EngineCacheQueue(
address=(pod_ip, cache_config.cache_queue_port),
address=(pod_ip, cache_config.local_cache_queue_port),
authkey=b"cache_queue_service",
is_server=False,
num_client=tensor_parallel_size,
@@ -210,7 +210,7 @@ class PrefixCacheManager:
val_cache_shape,
pod_ip,
engine_worker_queue_port,
pid_suffix,
ipc_suffix,
)
if cache_messager_processes is None:
raise RuntimeError("Launch cache messager failed")
@@ -269,15 +269,15 @@ class PrefixCacheManager:
+ f" --cache_dtype {cache_config.cache_dtype}"
+ f" --key_cache_shape {key_cache_shape}"
+ val_cache_arg_str
+ f" --cache_queue_port {cache_config.cache_queue_port}"
+ f" --cache_queue_port {cache_config.local_cache_queue_port}"
+ f" --enable_splitwise {int(self.enable_splitwise)}"
+ f" --pod_ip {pod_ip}"
+ f" --engine_worker_queue_port {engine_worker_queue_port}"
+ f" --num_cpu_blocks {cache_config.num_cpu_blocks}"
+ f" --engine_pid {pid_suffix}"
+ f" --ipc_suffix {ipc_suffix}"
+ f" --protocol {cache_config.cache_transfer_protocol}"
+ f" --local_data_parallel_id {self.local_data_parallel_id}"
+ f" --rdma_port {cache_config.rdma_comm_ports[i] if cache_config.rdma_comm_ports is not None else '0'}"
+ f" --rdma_port {cache_config.local_rdma_comm_ports[i] if cache_config.local_rdma_comm_ports is not None else '0'}"
+ f" --speculative_config '{self.speculative_config.to_json_string()}'"
+ f" --default_dtype '{self.config.model_config.dtype}'"
+ (" --create_cache_tensor" if create_cache_tensor else "")
@@ -321,7 +321,7 @@ class PrefixCacheManager:
value_cache_shape,
pod_ip,
engine_worker_queue_port,
pid_suffix,
ipc_suffix,
):
"""
launch_cache_messager function used to initialize the cache messager.
@@ -334,7 +334,7 @@ class PrefixCacheManager:
name="cache_ready_signal",
array=cache_ready_signal_data,
dtype=np.int32,
suffix=pid_suffix,
suffix=ipc_suffix,
create=False,
)
@@ -366,12 +366,12 @@ class PrefixCacheManager:
+ f" --key_cache_shape {key_cache_shape}"
+ val_cache_arg_str
+ f" --pod_ip {pod_ip}"
+ f" --cache_queue_port {cache_config.cache_queue_port}"
+ f" --cache_queue_port {cache_config.local_cache_queue_port}"
+ f" --engine_worker_queue_port {engine_worker_queue_port}"
+ f" --protocol {cache_config.cache_transfer_protocol}"
+ f" --local_data_parallel_id {self.local_data_parallel_id}"
+ f" --engine_pid {pid_suffix}"
+ f" --rdma_port {cache_config.rdma_comm_ports[i] if cache_config.rdma_comm_ports is not None else '0'}"
+ f" --ipc_suffix {ipc_suffix}"
+ f" --rdma_port {cache_config.local_rdma_comm_ports[i] if cache_config.local_rdma_comm_ports is not None else '0'}"
+ f" --speculative_config '{self.speculative_config.to_json_string()}'"
+ f" >{log_dir}/launch_cache_messager_tprank{i}.log 2>&1"
)