apps/FastDeploy
1
0
Fork 0
mirror of https://github.com/PaddlePaddle/FastDeploy.git synced 2025-10-05 08:37:06 +08:00
Code Issues Actions 2 Packages Projects Releases Wiki Activity

polish code with new pre-commit rule (#2923)

Browse Source
This commit is contained in:
Zero Rains
2025-07-19 23:19:27 +08:00
committed by GitHub
parent b8676d71a8
commit 25698d56d1
424 changed files with 14307 additions and 13518 deletions
Download Patch File Download Diff File Expand all files Collapse all files

32
fastdeploy/cache_manager/transfer_factory/rdma_cache_transfer.py
View File

@@ -24,13 +24,24 @@ class RDMACommManager:
RDMACommManager to manage rdma communication
"""
def __init__(self, splitwise_role, rank, gpu_id, cache_k_ptr_list, \
cache_v_ptr_list, max_block_num, block_bytes, rdma_port):
def __init__(
self,
splitwise_role,
rank,
gpu_id,
cache_k_ptr_list,
cache_v_ptr_list,
max_block_num,
block_bytes,
rdma_port,
):
try:
import rdma_comm
except:
logger.error(f"The installation of the RDMA library failed." \
"Confirm whether your network card supports RDMA transmission.")
logger.error(
"The installation of the RDMA library failed."
"Confirm whether your network card supports RDMA transmission."
)
return
self.messager = rdma_comm.RDMACommunicator(
splitwise_role,
@@ -50,7 +61,7 @@ class RDMACommManager:
Connect to remote gpu and write cache.
"""
assert self.splitwise_role == "prefill", "only prefill can call this method"
addr = f"{ip}:{str(port)}"
addr = f"{ip}:{port!s}"
if addr in self.connected_rdma:
return True
ret = self.messager.is_connected(ip, str(port))
@@ -59,18 +70,13 @@ class RDMACommManager:
return True
ret = self.messager.connect(ip, str(port))
logger.info(
f"connect to remote rdma address {ip}:{port} status is {ret}")
logger.info(f"connect to remote rdma address {ip}:{port} status is {ret}")
if ret == 0:
self.connected_rdma.add(addr)
return ret == 0
def write_cache(self, ip, port, local_block_ids, remote_block_ids,
layer_idx):
def write_cache(self, ip, port, local_block_ids, remote_block_ids, layer_idx):
"""
Connect to remote gpu and write cache.
"""
return self.messager.write_cache(ip, str(port), local_block_ids,
remote_block_ids, layer_idx)
return self.messager.write_cache(ip, str(port), local_block_ids, remote_block_ids, layer_idx)