server : disable embeddings/pooling on the speculative draft/MTP context

[liminfei-amd]

Overview

MTP (and speculative draft) models fail to load in llama-server while the same
model loads and runs in llama-cli (#24443). The server aborts in load_model
when creating the speculative context.

Root cause

When the server builds the speculative draft context and the MTP context,
it reuses the target context params. Those carry embeddings = true and a
pooling type. The draft/MTP graphs only produce draft logits and have no
embeddings output, so creating their context with embeddings/pooling enabled
fails. llama-cli does not copy these target-side params into the speculative
context, which is why it is unaffected. (This matches the maintainer note on
#24480 that MTP contexts do not need to set the embeddings tensor.)

Fix

In tools/server/server-context.cpp, before llama_init_from_model() for the
draft context and for the MTP context, set:

cparams.embeddings   = false;
cparams.pooling_type = LLAMA_POOLING_TYPE_NONE;

so the speculative context is created for logits only. The target context is
unchanged.

Validation

On an RDNA3.5 (gfx1151) ROCm/Vulkan build:

• cli: loads & runs the MTP model (--spec-type draft-mtp) — unaffected.
• server, unpatched: aborts in load_model (reproduced; gdb stack bottoms at
  server.cpp ctx_server.load_model).
• server, patched: loads, and serves MTP speculative decoding end to end
  (draft-mtp: #gen drafts = 15, #acc drafts = 5, #acc tokens = 8, mean acc len = 1.53), slots release cleanly. Output matches the cli run.
• git apply --check clean on current master.

Scope / related

#24480 (closed, unmerged) addressed a Gemma4-model-specific facet in
src/models/gemma4-assistant.cpp; this PR is the server-side context-params fix
and is independent of it. The issue thread also mentions hardware-specific
variants; this PR fixes the embeddings/pooling load failure that reproduces on
the speculative draft/MTP path, which is the common failure mode in the report.

Requirements

• I have read the contributing guidelines.
• AI usage disclosure: this change was prepared with AI assistance; a human
  reviewed and verified it and can explain every line.

[ggml-gh-bot]

Hi @liminfei-amd, thanks for your contribution!

Per our contribution guidelines, the automated PR checker found the following issue(s) that need your attention:

• Multiple open PRs from a new contributor: We limit new contributors (those without a previously merged PR) to 1 open PR at a time. You currently have 5 open PRs.

• AI-generated content: This project does not accept PRs, descriptions or commit messages that are fully or predominantly AI-generated. If you have used AI to assist you in writing code, please make sure to disclose that explicitly.

---

Please note that maintainers reserve the right to make final decisions on PRs. If you believe there is a mistake, please comment below.

[AndASM]

60bc886 appears to cause a new break that blocks this fix. With that commit reverted this fix works for me.

That's this pull I think #24980