grate/llama-server.sh at main · DataLabTechTV/grate · GitHub

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
#!/usr/bin/env sh
set -eux

model_alias() {
    model="$1"

    if [ -z "$model" ]; then
        echo
        return
    fi

    repo_model="${model%/*}"
    file="${model##*/}"
    file="${file%.gguf}"

    prefix="${repo_model##*/}"
    prefix="${prefix%-GGUF}"

    printf '%s:%s\n' "$repo_model" "${file#"$prefix"-}"
}

if [ ! -f /models/.initialized ]; then
    cp -r /usr/share/llama/models/. /models
    touch /models/.initialized
fi

LLAMA_MODEL_ALIAS=$(model_alias "${LLAMA_MODEL_PATH}")

/app/llama-server \
    --host 0.0.0.0 --port 8080 \
    --model "${LLAMA_MODEL_PATH}" \
    --alias "${LLAMA_MODEL_ALIAS}" \
    "$@" &

LLAMA_PID=$!

socat OPENSSL-LISTEN:8443,reuseaddr,fork,\
cert=/certs/wildcard.pem,key=/certs/wildcard-key.pem,\
verify=0 TCP:127.0.0.1:8080 &

SOCAT_PID=$!

cleanup() {
    kill "$SOCAT_PID" "$LLAMA_PID" 2>/dev/null || true
    wait "$SOCAT_PID" "$LLAMA_PID" 2>/dev/null || true
}

trap cleanup INT TERM EXIT

wait "$LLAMA_PID" "$SOCAT_PID"