Skip to content

server: /v1/models returns boolean vocab_type instead of integer #49

Description

@am17an

Upstream: ggml-org#29091

GET /v1/models returns meta.vocab_type: true instead of the documented integer vocabulary enum.

Reproduced on spark-a934.local, Linux aarch64, CPU Release, upstream 2b1847030cef76ef315eaee0b7ae0cdcd4fb15ff, Qwen2.5-0.5B-Instruct Q4_K_M.

llama-server -m qwen2.5-0.5b-instruct-q4_k_m.gguf -ngl 0 -c 512 -t 2 --host 127.0.0.1 --port 18919
curl -s localhost:18919/v1/models | jq '.data[0].meta.vocab_type'

Observed: true (JSON boolean). Expected: 2 (BPE).

Cause: get_res_model_info() passes enum llama_vocab_type directly to common_json_value. Its integer constructor excludes enums, leaving the bool overload. Upstream PR ggml-org#28518 identifies d9f918d as the introducing change; not independently bisected. Proposed upstream fixes: ggml-org#28518 / ggml-org#29097.

Activity

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Metadata

Metadata

Assignees

No one assigned

    Labels

    bugSomething isn't working

    Projects

    No projects

      Milestone

      No milestone

      Relationships

      None yet

      Development

      No branches or pull requests

      Issue actions