Comments (2)
As far as I know, data migration generally does not take long. But if you use a lot of GPUs, nvlink is effective. In the demos, my GPU usage was maintained at 100%. Please check if there is any synchronization operation in your program causing other GPUs to wait longer.
from pytorch-distributed.
yons 10829 0.2 0.1 2991056 134104 pts/9 S+ 19:34 0:05 python3 -m torch.distributed.launch --nproc_per_node=2 train.py -net ir -depth 50 -b 200 -w 2 -lr 0.1 -num_gpus 2
yons 10882 101 4.9 31145112 6472044 pts/9 Sl+ 19:34 29:48 /usr/bin/python3 -u train.py --local_rank=0 -net ir -depth 50 -b 200 -w 2 -lr 0.1 -num_gpus 2
yons 10883 101 8.2 37004736 10941320 pts/9 Sl+ 19:34 29:51 /usr/bin/python3 -u train.py --local_rank=1 -net ir -depth 50 -b 200 -w 2 -lr 0.1 -num_gpus 2
yons 11082 39.5 4.1 21305204 5501024 pts/9 Dl+ 19:34 11:29 /usr/bin/python3 -u train.py --local_rank=0 -net ir -depth 50 -b 200 -w 2 -lr 0.1 -num_gpus 2
yons 11091 39.4 4.1 21308788 5506644 pts/9 Dl+ 19:34 11:27 /usr/bin/python3 -u train.py --local_rank=0 -net ir -depth 50 -b 200 -w 2 -lr 0.1 -num_gpus 2
yons 11101 39.6 7.5 27062080 9980424 pts/9 Dl+ 19:34 11:30 /usr/bin/python3 -u train.py --local_rank=1 -net ir -depth 50 -b 200 -w 2 -lr 0.1 -num_gpus 2
yons 11110 39.6 7.5 27060044 9977892 pts/9 Dl+ 19:34 11:30 /usr/bin/python3 -u train.py --local_rank=1 -net ir -depth 50 -b 200 -w 2 -lr 0.1 -num_gpus 2
训练一段时间后,取数据进程状态变成了Dl+,gpu利用率就忽高忽低了,这是为什么啊
from pytorch-distributed.
Related Issues (20)
- Does the test accuracy need to be synchronized in distributed.py? HOT 7
- 大佬,请问如何指定gpu训练
- apex 并行加速中的data_prefetcher normalize为什么做两次? HOT 1
- 您好,我想请问一下Pytorch 在多进程分布式训练的时候,一开始加载数据集dataloader的时候,由于是多进程的,就会导致多个进程一起加载dataloader,这些dataloder全部加载到内存中使得内存爆炸,请问有没有什么办法解决呢
- 多卡显存占用不均衡 HOT 1
- No need to divide with nprocs again HOT 1
- Windows 提示Distributed package doesn't have NCCL "Distributed package doesn't have NCCL built in HOT 1
- how can i modify it
- how can i modify it ?
- RuntimeError: Timed out initializing process group in store based barrier on rank: 2, for key: store_based_barrier_key:1 (world_size=2, worker_count=4, timeout=0:30:00) HOT 2
- Can't pickle <function main_worker
- Connections time out
- DDP掉精度 HOT 3
- distributed.py seed HOT 6
- [Discussion] mp: duplicate of torch.cuda.set_device(local_rank) and images = images.cuda(local_rank, non_blocking=True) HOT 4
- About saving model in the multi-GPU training. HOT 3
- 使用mp.spawn启动比torch.distributed.launch慢很多 HOT 2
- 关于损失backward问题 HOT 1
- 请问使用DistributedSampler,各个GPU的数据是如何分配的?是连续(互不相同)的还是相同的? HOT 1
Recommend Projects
-
React
A declarative, efficient, and flexible JavaScript library for building user interfaces.
-
Vue.js
🖖 Vue.js is a progressive, incrementally-adoptable JavaScript framework for building UI on the web.
-
Typescript
TypeScript is a superset of JavaScript that compiles to clean JavaScript output.
-
TensorFlow
An Open Source Machine Learning Framework for Everyone
-
Django
The Web framework for perfectionists with deadlines.
-
Laravel
A PHP framework for web artisans
-
D3
Bring data to life with SVG, Canvas and HTML. 📊📈🎉
-
Recommend Topics
-
javascript
JavaScript (JS) is a lightweight interpreted programming language with first-class functions.
-
web
Some thing interesting about web. New door for the world.
-
server
A server is a program made to process requests and deliver data to clients.
-
Machine learning
Machine learning is a way of modeling and interpreting data that allows a piece of software to respond intelligently.
-
Visualization
Some thing interesting about visualization, use data art
-
Game
Some thing interesting about game, make everyone happy.
Recommend Org
-
Facebook
We are working to build community through open source technology. NB: members must have two-factor auth.
-
Microsoft
Open source projects and samples from Microsoft.
-
Google
Google ❤️ Open Source for everyone.
-
Alibaba
Alibaba Open Source for everyone
-
D3
Data-Driven Documents codes.
-
Tencent
China tencent open source team.
from pytorch-distributed.