mirror of
https://github.com/ggerganov/llama.cpp.git
synced 2026-02-05 13:53:23 +02:00
* graph : avoid branches between embedding and token inputs * models : make deepstack graphs (e.g. Qwen3 VL) have constant topology * ci : enable -DGGML_SCHED_NO_REALLOC=ON for server CI * cont : pad token embeddings to n_embd_inp