fix(container): update image ghcr.io/ggml-org/llama.cpp (server-cuda-b10362 ➔ server-cuda-b10380)

[drag0n141-bot]

This PR contains the following updates:

Package	Update	Change
ghcr.io/ggml-org/llama.cpp	patch	server-cuda-b10362 → server-cuda-b10380

---

Configuration

📅 Schedule: (in timezone Europe/Berlin)

• Branch creation
  • At any time (no schedule defined)
• Automerge
  • At any time (no schedule defined)

🚦 Automerge: Disabled by config. Please merge this manually once you are satisfied.

♻ Rebasing: Whenever PR becomes conflicted, or you tick the rebase/retry checkbox.

🔕 Ignore: Close this PR and you won't be reminded about these updates again.

---

• If you want to rebase/retry this PR, check this box

---

This PR has been generated by Mend Renovate CLI.

[claude]

ghcr.io/ggml-org/llama.cpp: server-cuda-b10362 → server-cuda-b10380

Verdict: Safe to merge

Patch bump spanning 18 upstream commits (roughly two weeks of development). Reviewed every commit in the range; none affect this deployment's usage (CUDA server image, dense Qwen3.5-4B GGUF model pulled via LLAMA_ARG_HF_REPO/LLAMA_ARG_MMPROJ_URL, jinja templating, discrete NVIDIA GPU via runtime: nvidia):

• llama: add default load-mode auto only changes mmap defaults for non-Metal iGPUs — not applicable to a discrete-GPU deployment.
• CUDA: only disable CUDA graphs when mul_mat_id actually needs a stream sync improves throughput for MoE models; Qwen3.5-4B is dense, so no measurable effect.
• mtmd: support pocket-tts, the RWKV wkv7 kernel, HIP register-spill list, and Windows CI additions are unrelated to this image's usage.
• Notable incidental fix: common/chat: tighten bare function parsing for Qwen models corrects a grammar-trigger bug where literal text such as #include <functional> was falsely matched as a tool call when tools were provided — directly benefits the Qwen model served by docker/nas01/llama/docker-compose.yaml, requires no config change on our end.

No breaking changes, deprecations, or required repo changes identified.

Sources consulted:

• ggml-org/llama.cpp@b10362...b10380
• ggml-org/llama.cpp#26793
• ggml-org/llama.cpp#26802
• ggml-org/llama.cpp#26081
• ggml-org/llama.cpp#26871