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.

Sign up for GitHub

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

Jump to bottom

Regression of log-transformed energy and pt, training checkpoints #343

Merged
merged 39 commits into from
Sep 20, 2024
Merged

Regression of log-transformed energy and pt, training checkpoints #343

merged 39 commits into from
Sep 20, 2024

Conversation

jpata
Copy link
Owner

@jpata jpata commented Sep 12, 2024 •
edited
Loading

Primary changes:

  • use log-transformed energy and pt for regression
  • split classification to binary and PID multiclass losses
  • last attention layer queries configurable, now set to input embeddings
  • additional CMS stats in v2.3
  • generate CLIC WW->fullhad, ZH->tautau, Z->tautau in v2.1
  • ensure training works on AMD MI250X (LUMI)
  • split and monitor regression losses separately
  • add CLD training
  • option to save attention matrices for visualization

Result checkpoints:

@jpata jpata changed the title Fixes sep6 Model improvements Sep 12, 2024
jpata and others added 16 commits September 13, 2024 13:06
@jpata
key_padding_mask
5034203
@jpata
save attention
bced81a
@jpata
fixes for distributed
bf1364d
@jpata
up
948b8e8
@jpata
up
c1acc52
@jpata
mass term
7987761
@jpata
add jet eta plot
bb6336c
@jpata
change mass loss coef
26f70cd
@jpata
runs of sep14
8b0432f
@jpata
restrict to pos mass
8c0bdb2
@jpata
disable layernorm
f1c36a0
@jpata
change jet ptcut
949dcf1
@jpata
split apart reg losses
28406b1
@jpata
rocm 6.2 (#344)
3788d5c 
* rocm

* up

* update lumi scripts

* remove correct dir

* update samples

* add samples

* up

* fix supervised key

* resubmit training

* add missing

* Merge

* add mpgun

* plt target and gen separately

* separate submission scripts

* add stats

* fix softmax bug

* add CLD sample

* add CLD sample

* add finetuning script

* add genjob script

* add cld

* up

* up

* up

* up

* up

---------

Co-authored-by: Joosep Pata <joosep.pata@kbfi.ee>
up
7870957
up
2b0430f
@jpata jpata changed the title Model improvements Regression of log-transformer energy and pt Sep 20, 2024
@jpata jpata changed the title Regression of log-transformer energy and pt Regression of log-transformer energy and pt, training checkpoints Sep 20, 2024
@jpata jpata changed the title Regression of log-transformer energy and pt, training checkpoints Regression of log-transformed energy and pt, training checkpoints Sep 20, 2024
@jpata jpata merged commit fba4288 into main Sep 20, 2024
5 checks passed
This was referenced Sep 20, 2024
Closed
Merged
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Reviewers
No reviews
Assignees
No one assigned
Labels
None yet
Projects
None yet
Milestone
No milestone
Development

Successfully merging this pull request may close these issues.
1 participant
@jpata