Skip to content

"Hey comma" to bookmark segments#35732

Closed
Quantizr wants to merge 33 commits intomasterfrom
heycomma
Closed

"Hey comma" to bookmark segments#35732
Quantizr wants to merge 33 commits intomasterfrom
heycomma

Conversation

@Quantizr
Copy link
Contributor

Saying "Hey comma" will let you bookmark a segment, in the same way that clicking the bookmark button in the sidebar currently does.

@Quantizr
Copy link
Contributor Author

Quantizr commented Jul 15, 2025

things im a little iffy on

  • apache attribution?
  • what core to set config_realtime_process
  • naming feedbackd vs something else like bookmarkd?
  • should this even be in modeld?

@Quantizr
Copy link
Contributor Author

I trolled and I thought the bad cores were the high index ones...
so actual CPU usage was 52% on the bad cores

quantized the embedding model to fp16, tinygrad LLVM has worse performance than fp32 on the bad core, so switched backend to CPU=1 and gets 42% CPU usage on bad cores

(however on the good core, LLVM seems to do better than CPU with the quantized model???)

there also seems to be possibly be some performance gap depending on if the net was compiled on the same type of core it is run on

@Quantizr
Copy link
Contributor Author

Quantizr commented Jul 15, 2025

17-18% little core CPU usage
roughly 2.5ms GPU usage / 80ms, 31 ms GPU / second, 3.1% GPU usage

@Quantizr
Copy link
Contributor Author

IMAGE=2 compiles the model to be ~ same time as BEAM=1 so this the previous cpu/gpu usage is the same, and compile times are normal

17-18% little core CPU usage
roughly 2.5ms GPU usage / 80ms, 31 ms GPU / second, 3.1% GPU usage

@Quantizr Quantizr requested a review from adeebshihadeh July 16, 2025 07:41
Copy link
Contributor

@adeebshihadeh adeebshihadeh left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Did a quick pass, looking into openWakeWord now

@commaai commaai deleted a comment from commaci-public Jul 17, 2025
@commaai commaai deleted a comment from github-actions bot Jul 17, 2025
@Quantizr Quantizr requested a review from adeebshihadeh July 17, 2025 04:02
@Quantizr Quantizr marked this pull request as draft July 22, 2025 08:13
@Quantizr
Copy link
Contributor Author

< 8% little core CPU usage... unsure how usable this is

model doesn't seem to get much better by scaling model size... perhaps real user data will fix it?

@github-actions
Copy link
Contributor

github-actions bot commented Aug 9, 2025

This PR has had no activity for 9 days. It will be automatically closed in 2 days if there is no activity.

@github-actions github-actions bot added the stale label Aug 9, 2025
@sshane sshane deleted the heycomma branch August 9, 2025 05:19
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

Projects

None yet

Development

Successfully merging this pull request may close these issues.

3 participants