Hugging Face's logo Hugging Face
  • Models
  • Datasets
  • Spaces
  • Buckets new
  • Docs
  • Enterprise
  • Pricing
    • Website
      • Tasks
      • HuggingChat
      • Collections
      • Languages
      • Organizations
    • Community
      • Blog
      • Posts
      • Daily Papers
      • Hardware
      • Learn
      • Discord
      • Forum
      • GitHub
    • Solutions
      • Team & Enterprise
      • Hugging Face PRO
      • Enterprise Support
      • Inference Providers
      • Inference Endpoints
      • Storage Buckets

  • Log In
  • Sign Up

Spaces:
Umong
/
repro-foam-state-folding
Running

App Files Files Community
repro-foam-state-folding / pages
504 kB
Ctrl+K
Ctrl+K
  • 1 contributor
History: 1 commit
Umong's picture
Umong
Update logbook: Repro - FOAM: Blocked State Folding for Memory-Efficient LLM Training
7b815f3 verified about 1 month ago
  • claim-1-foam-stores-the-predicted-compressed-number-of-optimizer-states
    Update logbook: Repro - FOAM: Blocked State Folding for Memory-Efficient LLM Training about 1 month ago
  • claim-2-fold-unfold-residuals-recover-the-full-gradient-without-persistent-state
    Update logbook: Repro - FOAM: Blocked State Folding for Memory-Efficient LLM Training about 1 month ago
  • claim-3-the-official-foam-update-matches-the-paper-equations
    Update logbook: Repro - FOAM: Blocked State Folding for Memory-Efficient LLM Training about 1 month ago
  • claim-4-foam-retains-the-primary-log-t-over-square-root-t-convergence-rate
    Update logbook: Repro - FOAM: Blocked State Folding for Memory-Efficient LLM Training about 1 month ago
  • conclusion
    Update logbook: Repro - FOAM: Blocked State Folding for Memory-Efficient LLM Training about 1 month ago
  • foam-62070-repro
    Update logbook: Repro - FOAM: Blocked State Folding for Memory-Efficient LLM Training about 1 month ago
  • sources-protocol-and-integrity
    Update logbook: Repro - FOAM: Blocked State Folding for Memory-Efficient LLM Training about 1 month ago
  • index.md
    865 Bytes
    Update logbook: Repro - FOAM: Blocked State Folding for Memory-Efficient LLM Training about 1 month ago