Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

feat: pinning index in memory when building, second try #181

Merged
merged 2 commits into from
Feb 8, 2025

Conversation

usamoi
Copy link
Contributor

@usamoi usamoi commented Feb 6, 2025

issue: #117
first try: #150

@VoVAllen
Copy link
Member

VoVAllen commented Feb 6, 2025

Let's try it on 100M dataset and see how flamegraph goes

@usamoi
Copy link
Contributor Author

usamoi commented Feb 7, 2025

2c4t, on laion-100m:

  • estimated build time on main: ~60h
  • estimated build time on pinning-v2: ~50h

next steps:

  • merge two shared memory segments to one
  • make it optional

cc @VoVAllen

@VoVAllen
Copy link
Member

VoVAllen commented Feb 7, 2025

Why do we need to merge two segments together?

The optimization result is pretty close to the flamegraph result.

@VoVAllen
Copy link
Member

VoVAllen commented Feb 7, 2025

And what is 2c4t?

@usamoi
Copy link
Contributor Author

usamoi commented Feb 8, 2025

And what is 2c4t?

4 vcpu

Why do we need to merge two segments together?

not necessary; helpful to make it optional

@usamoi usamoi force-pushed the pinning-v2 branch 2 times, most recently from bc95918 to dd76752 Compare February 8, 2025 06:12
@usamoi
Copy link
Contributor Author

usamoi commented Feb 8, 2025

i4i.xlarge, laion-100m:

  • estimated build time on main: ~61h
  • estimated build time on pinning-v2: ~44h

@usamoi usamoi merged commit 6373f55 into tensorchord:main Feb 8, 2025
13 checks passed
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
None yet
Projects
None yet
Development

Successfully merging this pull request may close these issues.

2 participants