hazyresearch/based-360m
Text Generation • Updated
• 16 • 2
based
These language model checkpoints are trained at the 360M and 1.3Bn parameter scales for up to 50Bn tokens on the Pile corpus, for research purposes.
-
-
hazyresearch/attn-360m
Updated • 6 • 1 -
hazyresearch/mamba-360m
Updated • 5 -
hazyresearch/based-1b
Text Generation • Updated • 6 • 8 -
hazyresearch/mamba-1b
Updated • 4 • 1 -
Simple linear attention language models balance the recall-throughput tradeoff
Paper • 2402.18668 • Published • 20 -
hazyresearch/attn-1b
Updated • 4 -
hazyresearch/based-fda
Viewer • Updated • 1.1k • 1.43k • 3 -
hazyresearch/based-squad
Viewer • Updated • 2.98k • 1.01k • 2 -
hazyresearch/based-swde
Viewer • Updated • 1.11k • 1.18k • 4 -
hazyresearch/based-1b-50b
Updated • 2 • 1 -
hazyresearch/mamba-1b-50b
Updated -
hazyresearch/attn-1b-50bn
Updated • 1 -
hazyresearch/my-awesome-model
Updated • 4 • 2