List your company

RL Training on LongReason

80Peak Memory (GB)DualKV
78.9685.9893100.02May 14, 2026

Evaluation results

MethodLinks
DualKVmb/GPU=42026.058031363—231606—59.385.3
FA2mb/GPU=8, SP=42026.05924,1154475072145,28480.2——
DualKVmb/GPU=82026.059324263—237542—77.486.8
DualKVmb/GPU=8, SP=12026.051031,0781291422151,56477.4——
FA2mb/GPU=42026.05106597153—236987—30.984.8
FA3mb/GPU=42026.05106543142—234919—3485.9