zhangyiss/mlx
1
0
Fork 0
mirror of https://github.com/ml-explore/mlx.git synced 2025-06-24 17:31:16 +08:00
Code Issues Actions Packages Projects Releases Wiki Activity

Only fail when 10 consecutive socket errors occur (#1928)

Browse Source
This commit is contained in:
Angelos Katharopoulos 2025-03-05 13:16:19 -08:00 committed by GitHub
parent fd0d63ba5b
commit 0792ff02ff
No known key found for this signature in database
GPG Key ID: B5690EEEBB952194
2 changed files with 16 additions and 3 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

12
mlx/distributed/ring/ring.cpp
View File

@ -199,6 +199,7 @@ class SocketThread {
}
void worker() {
int error_count = 0;
bool delete_recv = false;
bool delete_send = false;
while (true) {
@ -235,10 +236,11 @@ class SocketThread {
task.buffer = static_cast<char*>(task.buffer) + r;
task.size -= r;
delete_recv = task.size == 0;
error_count = 0;
} else if (errno != EAGAIN) {
error_count++;
log_info(
true, "Receiving from socket", fd_, "failed with errno", errno);
return;
}
}
if (!sends_.empty()) {
@ -248,11 +250,17 @@ class SocketThread {
task.buffer = static_cast<char*>(task.buffer) + r;
task.size -= r;
delete_send = task.size == 0;
error_count = 0;
} else if (errno != EAGAIN) {
error_count++;
log_info(true, "Sending to socket", fd_, "failed with errno", errno);
return;
}
}
if (error_count >= 10) {
log_info(true, "Too many send/recv errors. Aborting...");
return;
}
}
}

5
python/mlx/distributed_run.py
View File

@ -112,7 +112,12 @@ def extract_rings(hosts, index):
break
if not ring:
break
try:
rings.append(normalize(concretize(ring, used_ports)))
except RuntimeError:
if len(rings) > 0:
return rings
raise
return rings