pablo-reyes8/kimi-k3-pytorch: Build, train and experiment with a research-scale Kimi K3 in PyTorch — implementing KDA, Gated MLA, AttnRes, LatentMoE, multimodality, Muon, long-context training, KV-cache inference and ViT.

Build, train and experiment with a research-scale Kimi K3 in PyTorch — implementing KDA, Gated MLA, AttnRes, LatentMoE, multimodality, Muon, long-context training, KV-cache inference and ViT.

Read Original

Related