Skip to content

Aligning sasrec implementation with CIKM2020-S3Rec#5

Draft
iskbaga wants to merge 1 commit into
NonameUntitled:mainfrom
iskbaga:sasrec-refactored
Draft

Aligning sasrec implementation with CIKM2020-S3Rec#5
iskbaga wants to merge 1 commit into
NonameUntitled:mainfrom
iskbaga:sasrec-refactored

Conversation

@iskbaga

@iskbaga iskbaga commented Jun 17, 2026

Copy link
Copy Markdown
Collaborator

Align our SASRec implementation with baseline from TIGER paper (from S3-Rec - Zhou et al. CIKM '20, github.com/aHuiWang/CIKM2020-S3Rec):

  • tiger/modeling/models/sasrec.py - embedding layerNorm + dropout added
  • tiger/modeling/models/base.py - updated weight init
  • tiger/modeling/dataset/samplers.py - negative sampling rejects items from user's history
  • tiger/modeling/dataset/base.py - num_items to sampler
  • notebooks/DatasetProcessing.ipynb - add _row_idx for join determinism
  • tiger/configs/sasrec_train_config.json - params to S3-Rec defaults
  • tiger/train_sasrec.py - Adam (not AdamW) and drop_last=False

@iskbaga iskbaga marked this pull request as draft June 17, 2026 22:24
@iskbaga iskbaga force-pushed the sasrec-refactored branch 5 times, most recently from 4321511 to 03b1e49 Compare June 18, 2026 20:20
@iskbaga iskbaga force-pushed the sasrec-refactored branch from 03b1e49 to 2f7abfd Compare June 18, 2026 20:33
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.

1 participant