All pull requests

Pull requests

• Prevent activation_checkpoint_cpu_offload from silently no-op'ing on transformers >=5.0

  #373·xylian86 opened May 11, 2026·

  ·

• Fix incorrect validation for attention implementations

  #371·sfc-gh-sbekman opened Apr 23, 2026Collaborator·

  ·

• Add the custom Comm for creating new custom communication collectives

  #370·sfc-gh-reyazda opened Apr 3, 2026Contributor·

  ·

• Feature: Support more experiment tracking libraries

  #365·sfc-gh-mwyatt opened Mar 24, 2026Collaborator·

  ·
  4

• Enhancement: Better Metrics

  #362·sfc-gh-jrasley opened Mar 12, 2026Collaborator·

  ·
  8

• Add GRPO-based RL training for Text-to-SQL

  #350·VNitish opened Feb 9, 2026·

  ·

• adding input output format

  #349·sfc-gh-zhyao opened Feb 2, 2026Contributor·

  ·

• Improve sample packing

  #347·sfc-gh-truwase opened Jan 29, 2026Collaborator·

  ·
  19

• new feature: On policy distillation

  #344·sfc-gh-thonguyen opened Jan 28, 2026Collaborator·

  ·
  2

• Fix NaN loss in SFT by using token-based label masking

  #339·sfc-gh-aponnusamy opened Jan 9, 2026·

  ·

• Fix NaN validation loss in SFT training by correcting label masking logic

  #335·sfc-gh-aponnusamy opened Jan 5, 2026·

  ·
  5

• Ac sft debug

  #333·sfc-gh-aponnusamy opened Dec 23, 2025·

  ·

• Feature: TransformerEngine FP8 support

  #331·sfc-gh-mwyatt opened Dec 18, 2025Collaborator·

  ·
  2

• Zhewei smart packing

  #327·sfc-gh-zhyao opened Dec 15, 2025Contributor·

  ·
  4

• Update ArcticSpeculator README

  #321·sfc-gh-jaelee opened Dec 5, 2025Contributor·

  ·

• testing modal ci with pull_request_target

  #318·sfc-gh-sbekman opened Dec 2, 2025Collaborator·

  ·

• get modal CI to run on PRs from forks

  #317·sfc-gh-sbekman opened Dec 2, 2025Collaborator·

  ·
  1

• chunk

  #313·sfc-gh-zhyao opened Nov 21, 2025Contributor·

  ·

• try to fix dpo traiing

  #300·sfc-gh-zhyao opened Oct 30, 2025Contributor·

  ·

• Pyu/patch arctic embed examples 1

  #283·sfc-gh-pyu opened Sep 19, 2025Collaborator·

  ·

• Arctic-embed-SPLADE

  #274·sfc-gh-pyu opened Sep 3, 2025Collaborator·

  ·

• [DO NOT MERGE] GPU Actions Workflow

  #207·sfc-gh-mwyatt opened Jun 17, 2025Collaborator·

  ·

• Better FA2 FLOPs estimate

  #171·sfc-gh-mwyatt opened Apr 28, 2025Collaborator·

  ·

• Add Arctic Embed minimal example

  #138·sfc-gh-lmerrick opened Apr 9, 2025Contributor·

  ·
  5

• Allow CLI overrides of yaml config values

  #101·sfc-gh-mwyatt opened Mar 19, 2025Collaborator·

  ·
  1