fxmeng/DSV2-Lite-IndexCache-Warmup-400

Same weights, repacked vs deepseek-ai/DeepSeek-V2-Lite. Same weights in different files (re-saved, re-sharded or new metadata), plus 45 new tensors.

Tensors changed: lm_head 0/1, Embeddings 0/1, Norms 0/82, MLP 0/2763, Attention 0/108

On the Hub · Downloads on Model Pulse