deepseek-ai/ESFT-vanilla-lite
Text Generation • 16B • Updated • 137 • 20
ESFT
models for paper expert-specialized fine-tuning
-
-
deepseek-ai/ESFT-token-law-lite
Text Generation • 1B • Updated • 108 • 6 -
deepseek-ai/ESFT-token-summary-lite
Text Generation • 1B • Updated • 104 • 4 -
deepseek-ai/ESFT-token-code-lite
Text Generation • 2B • Updated • 112 • 4 -
deepseek-ai/ESFT-token-intent-lite
Text Generation • 2B • Updated • 117 • 3 -
deepseek-ai/ESFT-token-translation-lite
Text Generation • 0.9B • Updated • 109 • 3 -
deepseek-ai/ESFT-token-math-lite
Text Generation • 1B • Updated • 109 • 4 -
deepseek-ai/ESFT-gate-law-lite
Text Generation • 2B • Updated • 87 • 4 -
deepseek-ai/ESFT-gate-summary-lite
Text Generation • 2B • Updated • 92 • 3 -
deepseek-ai/ESFT-gate-translation-lite
Text Generation • 1B • Updated • 92 • 4 -
deepseek-ai/ESFT-gate-code-lite
Text Generation • 2B • Updated • 96 • 3 -
deepseek-ai/ESFT-gate-intent-lite
Text Generation • 2B • Updated • 98 • 4 -
deepseek-ai/ESFT-gate-math-lite
Text Generation • 2B • Updated • 100 • 3 -
Let the Expert Stick to His Last: Expert-Specialized Fine-Tuning for Sparse Architectural Large Language Models
Paper • 2407.01906 • Published • 47