Article
Shipping a Trillion Parameters With a Hub Bucket: Delta Weight Sync in TRL
- +5
aminediroHF, qgallouedec, kashif, lewtun, edbeeching, albertvillanova, lvwerra
• • 28Amine Dirhoussi
aminediroHF
AI & ML interests
None yet
Recent Activity
published an article 3 days ago
Shipping a Trillion Parameters With a Hub Bucket: Delta Weight Sync in TRL updated a Space 7 days ago
aminediroHF/delta-weight-sync-figures updated a Space 7 days ago
aminediroHF/delta-weight-sync-anchor-deltaOrganizations
published an article 3 days ago
updated 5
Spaces 7 days ago
Delta Weight Sync Figures
🔄
Interactive figures for the TRL delta weight sync blogpost
Delta Weight Sync · Anchor/Delta Cadence
📦
Watch a bucket fill with anchors and sparse deltas
Delta Weight Sync · bf16 ULP Demo
📐
Interactive bf16 ULP visualization for RL updates
Delta Weight Sync · Diff Transfer
🧬
End-to-end animation of one weight sync, delta vs NCCL
Delta Weight Sync · Timeline
🔄
NCCL vs delta-over-bucket weight sync timeline comparison
published 4
Spaces 7 days ago
Delta Weight Sync · Diff Transfer
🧬
End-to-end animation of one weight sync, delta vs NCCL
Delta Weight Sync · Anchor/Delta Cadence
📦
Watch a bucket fill with anchors and sparse deltas
Delta Weight Sync · bf16 ULP Demo
📐
Interactive bf16 ULP visualization for RL updates
Delta Weight Sync · Timeline
🔄
NCCL vs delta-over-bucket weight sync timeline comparison
published a
Space 9 days ago
upvoted an article 10 days ago
Article
PaddleOCR 3.5: Running OCR and Document Parsing Tasks with a Transformers Backend
PaddlePaddle
• • 32 upvoted an article 14 days ago
Article
Unlocking asynchronicity in continuous batching
- +1
ror, pcuenq, ariG23498
• • 55 updated a
Space 16 days ago
published a
Space 16 days ago
updated a
bucket 16 days ago
published a
bucket 16 days ago
updated a
Space 16 days ago
published a
Space 16 days ago
updated a
bucket 16 days ago