Git Product home page Git Product logo

Comments (4)

freewym avatar freewym commented on August 26, 2024

Would it be faster if using 2 GPUs? Or does increasing batch size solve this issue? I am trying to figure out whether the problem is because of using more than 2 GPUs, or simply because the batch size you set is not large enough.

from espresso.

yfliao avatar yfliao commented on August 26, 2024

Sorry for the late replay.

The max-tokens and max-sentences were set to 32/24 and 39000/26000, respectively for the transformer/conv_lstm recipes. So it is not the problem. However, increasing the "update_freq" to 5 mades things a little better.


+-----------------------------------------------------------------------------+
| NVIDIA-SMI 450.51.05    Driver Version: 450.51.05    CUDA Version: 11.0     |
|-------------------------------+----------------------+----------------------+
| GPU  Name        Persistence-M| Bus-Id        Disp.A | Volatile Uncorr. ECC |
| Fan  Temp  Perf  Pwr:Usage/Cap|         Memory-Usage | GPU-Util  Compute M. |
|                               |                      |               MIG M. |
|===============================+======================+======================|
|   0  Tesla V100-SXM2...  On   | 00000000:85:00.0 Off |                    0 |
| N/A   51C    P0   263W / 300W |  13936MiB / 16160MiB |     97%      Default |
|                               |                      |                  N/A |
+-------------------------------+----------------------+----------------------+
|   1  Tesla V100-SXM2...  On   | 00000000:86:00.0 Off |                    0 |
| N/A   54C    P0   267W / 300W |  13842MiB / 16160MiB |     84%      Default |
|                               |                      |                  N/A |
+-------------------------------+----------------------+----------------------+
|   2  Tesla V100-SXM2...  On   | 00000000:89:00.0 Off |                    0 |
| N/A   59C    P0   277W / 300W |  13868MiB / 16160MiB |     81%      Default |
|                               |                      |                  N/A |
+-------------------------------+----------------------+----------------------+
|   3  Tesla V100-SXM2...  On   | 00000000:8A:00.0 Off |                    0 |
| N/A   53C    P0   274W / 300W |   8734MiB / 16160MiB |     88%      Default |
|                               |                      |                  N/A |
+-------------------------------+----------------------+----------------------+

from espresso.

freewym avatar freewym commented on August 26, 2024

Hmm, it seems that the bottle-neck is from when collecting gradients/stats across multiple GPUs in backprop. I don't have enough computing resources to try 4 GPUs, but I heard from someone else who tried 4 GTX 2080 GPUs in a SunGirdEngine environment that they didn't encounter such issue. So I am sorry I am unable to help out on this. But If a larger update_freq could help, maybe you can reduce the batch size accordingly so that the effective batch size for each update remain the same as when update_freq=1?

from espresso.

yfliao avatar yfliao commented on August 26, 2024

Thanks for the reminder. I will reduce the batch size.

from espresso.

Related Issues (20)

Recommend Projects

  • React photo React

    A declarative, efficient, and flexible JavaScript library for building user interfaces.

  • Vue.js photo Vue.js

    🖖 Vue.js is a progressive, incrementally-adoptable JavaScript framework for building UI on the web.

  • Typescript photo Typescript

    TypeScript is a superset of JavaScript that compiles to clean JavaScript output.

  • TensorFlow photo TensorFlow

    An Open Source Machine Learning Framework for Everyone

  • Django photo Django

    The Web framework for perfectionists with deadlines.

  • D3 photo 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.

  • Game

    Some thing interesting about game, make everyone happy.

Recommend Org

  • Facebook photo Facebook

    We are working to build community through open source technology. NB: members must have two-factor auth.

  • Microsoft photo Microsoft

    Open source projects and samples from Microsoft.

  • Google photo Google

    Google ❤️ Open Source for everyone.

  • D3 photo D3

    Data-Driven Documents codes.