disagg.sh 1.13 KB
Newer Older
1
2
3
4
5
6
7
8
9
10
11
12
13
14
#!/bin/bash
# SPDX-FileCopyrightText: Copyright (c) 2025 NVIDIA CORPORATION & AFFILIATES. All rights reserved.
# SPDX-License-Identifier: Apache-2.0

# Setup cleanup trap
cleanup() {
    echo "Cleaning up background processes..."
    kill $DYNAMO_PID $PREFILL_PID 2>/dev/null || true
    wait $DYNAMO_PID $PREFILL_PID 2>/dev/null || true
    echo "Cleanup complete."
}
trap cleanup EXIT INT TERM

# run clear_namespace
15
python3 -m dynamo.sglang.utils.clear_namespace --namespace dynamo
16
17

# run ingress
18
python3 -m dynamo.frontend --http-port=8000 &
19
20
21
DYNAMO_PID=$!

# run prefill worker
22
python3 -m dynamo.sglang \
23
24
  --model-path Qwen/Qwen3-0.6B \
  --served-model-name Qwen/Qwen3-0.6B \
25
26
27
28
29
30
31
32
33
  --page-size 16 \
  --tp 1 \
  --trust-remote-code \
  --skip-tokenizer-init \
  --disaggregation-mode prefill \
  --disaggregation-transfer-backend nixl &
PREFILL_PID=$!

# run decode worker
34
CUDA_VISIBLE_DEVICES=1 python3 -m dynamo.sglang \
35
36
  --model-path Qwen/Qwen3-0.6B \
  --served-model-name Qwen/Qwen3-0.6B \
37
38
39
40
41
  --page-size 16 \
  --tp 1 \
  --trust-remote-code \
  --skip-tokenizer-init \
  --disaggregation-mode decode \
42
  --disaggregation-transfer-backend nixl