Repository navigation
feat: support SSH-based multi-node training launch - #260
Open
JYMiracle305 wants to merge 4 commits into
Open
JYMiracle305 wants to merge 4 commits into
JYMiracle305 wants to merge 4 commits into
Conversation
JYMiracle305
force-pushed
the
feat/ssh-multinode-launcher
branch
2 times, most recently
from
November 3, 2026 13:28
af20bc2 to
e473e35
Compare
JYMiracle305
force-pushed
the
feat/ssh-multinode-launcher
branch
from
November 3, 2026 14:11
e473e35 to
b33bf43
Compare
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
说明
127.0.0.1临时端口,支持同机并发任务;多机由 node rank 0 监听显式的MASTER_ADDR:MASTER_PORT。InfiniTrain::rendezvous基础库,协议由服务端和 CCL 客户端共享。使用方式
控制节点通过
scripts/multinode_ssh/launch_multinode_ssh.sh启动任务,可使用--hosts或 hostfile 指定节点顺序。常用参数包括:--nproc-per-node:每节点进程数。--master-addr/--master-port:多机 Rendezvous 地址。--rdzv-timeout/--job-timeout:初始化及任务超时。--local-host:指定控制节点直接本地执行。--remote-docker-container:在远程容器中执行。--env/--check-path:透传环境变量并检查数据或权重路径。--follow-output:在控制端持续显示最后一个 global rank 的 stderr。启动前会检查代码 Git SHA、训练程序和
infini_run的 SHA-256、数据路径、GPU 可见数量及 NVIDIA 环境。控制日志记录完整命令、元数据和预检结果,rank 日志写入:本地回归测试位于:
验证结果
已在两个训练节点、每节点 8 张 NVIDIA A100-SXM4-80GB 的环境完成验证:
TP4 × PP2 × DP2、20 step 训练。4.468029、2.951596、3.553573。0.000001、0.004184、0.002242。NET/IBext_v10和 GDRDMA。0/17/124/130。infini_run、infini_train构建、分布式 rank 测试、clang-format 和全量 IWYU 检查通过。已知限制
--tokenizer_bin,当前文本生成路径要求 pipeline parallel size 为 1。