LayerSkip: Enabling Early Exit Inference and Self-Speculative Decoding
Paper
•
2404.16710
•
Published
•
74
LayerSkip
Models continually pretrained using LayerSkip - https://arxiv.org/abs/2404.16710
-
-
facebook/layerskip-llama2-7B
Text Generation • Updated • 506 • 11 -
facebook/layerskip-llama2-13B
Text Generation • Updated • 71 • 5 -
facebook/layerskip-llama2-70B
Text Generation • Updated • 71 • 4 -
facebook/layerskip-codellama-7B
Text Generation • Updated • 35 • 6 -
facebook/layerskip-codellama-34B
Text Generation • Updated • 18 • 4 -
facebook/layerskip-llama3-8B
Text Generation • Updated • 310 • 18 -
facebook/layerskip-llama3.2-1B
Text Generation • Updated • 1.21k • 17