inclusionAI Releases Duo-1.0-distill Four-Step Distilled Weights

inclusionAI released Duo-1.0-distill, four-step distilled Realization weights for Duo 1.0, its full-duplex multimodal world model for real-time audio-visual interaction. The weights are published as a PyTorch Distributed Checkpoint overlay on the inclusionAI/Duo-1.0-causal-base bundle, which supplies the shared Intent, codecs, speech models and Realization backbone; users must download that full bundle and place the weights in the same directory. The release uses an Euler sampler with 4 steps and flow shift 5.0. A companion FP8-quantized version, Duo-1.0-distill-quant, is also available; its runtime must support the packaged ModelOpt FP8 format or an equivalent conversion path. According to inclusionAI, both sets of weights were initialized from MiniMax H3 and then trained by the Duo team.

สร้างจากรายงาน · อัปเดต 1 ชม. ที่แล้ว

ยังมีการเฝ้าดูต่อเนื่องไม่พอจะวาดแนวโน้ม