Giles Thomas

gpjt's activity

• All
• Models
• Datasets
• Spaces
• Buckets
• Papers
• Collections
• Community
• Posts
• Upvotes
• Likes
• Articles

New activity in AndrewThompson1233/maba-v1-architecture 2 days ago

Factorized embeddings intervention for GPT-2

17

#1 opened 13 days ago by

gpjt

updated a collection 8 days ago

LoRE

Collection

Models using low-rank bottlenecks for the embeddings or the output head. See https://www.gilesthomas.com/2026/10/low-rank-vocab-matrices • 11 items • Updated 8 days ago

updated a model 8 days ago

gpjt/8xa100m40-lore-9-new-seed-output-only

Text Generation • 0.1B • Updated 8 days ago • 180

published a model 8 days ago

gpjt/8xa100m40-lore-9-new-seed-output-only

Text Generation • 0.1B • Updated 8 days ago • 180

updated a model 8 days ago

gpjt/8xa100m40-lore-8-new-seed-both-smart-init

Text Generation • 0.1B • Updated 8 days ago • 173

published a model 8 days ago

gpjt/8xa100m40-lore-8-new-seed-both-smart-init

Text Generation • 0.1B • Updated 8 days ago • 173

updated a model 8 days ago

gpjt/8xa100m40-lore-7-new-seed-input-only

Text Generation • 0.1B • Updated 8 days ago • 167

published a model 8 days ago

gpjt/8xa100m40-lore-7-new-seed-input-only

Text Generation • 0.1B • Updated 8 days ago • 167

updated a model 8 days ago

gpjt/8xa100m40-lore-6-new-seed-baseline-disabled

Text Generation • 0.2B • Updated 8 days ago • 159

published a model 8 days ago

gpjt/8xa100m40-lore-6-new-seed-baseline-disabled

Text Generation • 0.2B • Updated 8 days ago • 159