20 linhas
470 B
Bash
20 linhas
470 B
Bash
|
#!/bin/bash
|
||
|
|
|
||
|
|
exec /app/llama-server $BASE_TOP_MODEL \
|
||
|
|
-c $BASE_CONTEXT_SIZE -ngl $BASE_GPU_LAYERS -n $BASE_MAX_TOKENS \
|
||
|
|
--temp 0.5 --top-p 0.9 --top-k 40 --repeat-penalty 1.1 \
|
||
|
|
--flash-attn auto --threads -1 --threads-batch -1 --threads-http -1 \
|
||
|
|
--jinja \
|
||
|
|
--timeout 900 --host 0.0.0.0 --port 8093 &
|
||
|
|
PID=$!
|
||
|
|
|
||
|
|
cleanup() {
|
||
|
|
echo "Stopping llama-server..."
|
||
|
|
kill $PID 2>/dev/null
|
||
|
|
wait $PID 2>/dev/null
|
||
|
|
exit 0
|
||
|
|
}
|
||
|
|
|
||
|
|
trap cleanup SIGTERM SIGINT
|
||
|
|
|
||
|
|
wait $PID
|