##
## vLLM engine configuration — qwen05 (CPU smoke test)
##
## Consumed by `vllm serve --config` (see vllm-model@.container). Each key is a
## long CLI flag with the leading `--` stripped; store-true flags (e.g.
## --enforce-eager) become `true`. CLI args passed after --config still win, so
## the Quadlet keeps only `--config` and lets this file carry every parameter.
##
## Installed to /etc/quadlets/vllm/models/qwen05.yaml and mounted read-only into
## the container at /etc/vllm/config.yaml.
##

# Tiny real instruct model: has a chat template AND a tool-call parser, so it
# exercises /v1/chat/completions (what the Traefik /v1/* allowlist fronts).
model: Qwen/Qwen2.5-0.5B-Instruct
served-model-name: qwen05
dtype: bfloat16
max-model-len: 4096
max-num-seqs: 4

# --enforce-eager: this CPU image's torch inductor JIT fails at warmup; eager
# mode skips it (fine for a smoke test).
enforce-eager: true

# Tool-calling support (Qwen2 uses the hermes parser template).
enable-auto-tool-choice: true
tool-call-parser: hermes