List your company

Decoding on Multi-task Evaluation Suite Llama-3.2-1B (test)

394.81MT Throughput (token/s)FR-Spec
254.4308290.8754327.32363.7646Feb 20, 2025

Evaluation results

MethodLinks
FR-SpecConfiguration=16k, Temperature=0, Token-frequency statistics=SlimPajama2025.02394.81443326.75476.47394.47375.7402.07401.91.59
FR-SpecConfiguration=8k, Temperature=0, Token-frequency statistics=SlimPajama2025.02386.97428.94319.83462.98382.75363.5392.13391.011.55
FR-SpecConfiguration=32k, Temperature=0, Token-frequency statistics=SlimPajama2025.02378.9428.75317.68467.53378.39363.7395.95390.131.55
FR-SpecConfiguration=64k, Temperature=0, Token-frequency statistics=SlimPajama2025.02349.12406.14297.62427.14350.08338.81390.78365.671.45
EAGLE-2Temperature=0, Token-frequency statistics=SlimPajama2025.02306.04358.37266.84372.37305.52294.82360.6323.511.28
VanillaTemperature=0, Token-frequency statistics=SlimPajama2025.02259.83255.89220.25263.34260.13248.15256.64252.031