xinyun/vllm
1
0
Fork 0
mirror of https://git.datalinker.icu/vllm-project/vllm.git synced 2026-05-31 13:47:08 +08:00
Code Issues Packages Projects Releases Wiki Activity

[Misc][Tools] make max-model-len a parameter in auto_tune script (#21321)

Browse Source 
Signed-off-by: Chengji Yao <chengjiyao@google.com>
Co-authored-by: gemini-code-assist[bot] <176961590+gemini-code-assist[bot]@users.noreply.github.com>
This commit is contained in:
Chengji Yao 2025-07-24 22:46:43 -07:00 committed by GitHub
parent fde60ee775
commit 947edd099e
No known key found for this signature in database
GPG Key ID: B5690EEEBB952194
2 changed files with 15 additions and 3 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

4
benchmarks/auto_tune/README.md
View File

@ -39,6 +39,7 @@ You must set the following variables at the top of the script before execution.
| `DOWNLOAD_DIR` | **Required.** Directory to download and load model weights from. | `""` (default download path) | | `DOWNLOAD_DIR` | **Required.** Directory to download and load model weights from. | `""` (default download path) |
| `INPUT_LEN` | **Required.** Request input length. | `4000` | | `INPUT_LEN` | **Required.** Request input length. | `4000` |
| `OUTPUT_LEN` | **Required.** Request output length. | `16` | | `OUTPUT_LEN` | **Required.** Request output length. | `16` |
| `MAX_MODEL_LEN` | **Required.** Max model length. | `4096` |
| `MIN_CACHE_HIT_PCT` | Prefix cache hit rate in percentage (0-100). Set to `0` to disable. | `60` | | `MIN_CACHE_HIT_PCT` | Prefix cache hit rate in percentage (0-100). Set to `0` to disable. | `60` |
| `MAX_LATENCY_ALLOWED_MS` | The maximum allowed P99 end-to-end latency in milliseconds. Set to a very large number (e.g., `100000000000`) to effectively ignore the latency constraint. | `500` | | `MAX_LATENCY_ALLOWED_MS` | The maximum allowed P99 end-to-end latency in milliseconds. Set to a very large number (e.g., `100000000000`) to effectively ignore the latency constraint. | `500` |
| `NUM_SEQS_LIST` | A space-separated string of `max-num-seqs` values to test. | `"128 256"` | | `NUM_SEQS_LIST` | A space-separated string of `max-num-seqs` values to test. | `"128 256"` |
@ -69,6 +70,7 @@ Here are a few examples of how to configure the script for different goals:
```bash ```bash
INPUT_LEN=1800 INPUT_LEN=1800
OUTPUT_LEN=20 OUTPUT_LEN=20
MAX_MODEL_LEN=2048
MIN_CACHE_HIT_PCT=0 MIN_CACHE_HIT_PCT=0
MAX_LATENCY_ALLOWED_MS=100000000000 # A very large number MAX_LATENCY_ALLOWED_MS=100000000000 # A very large number
``` ```
@ -80,6 +82,7 @@ MAX_LATENCY_ALLOWED_MS=100000000000 # A very large number
```bash ```bash
INPUT_LEN=1800 INPUT_LEN=1800
OUTPUT_LEN=20 OUTPUT_LEN=20
MAX_MODEL_LEN=2048
MIN_CACHE_HIT_PCT=0 MIN_CACHE_HIT_PCT=0
MAX_LATENCY_ALLOWED_MS=500 MAX_LATENCY_ALLOWED_MS=500
``` ```
@ -91,6 +94,7 @@ MAX_LATENCY_ALLOWED_MS=500
```bash ```bash
INPUT_LEN=1800 INPUT_LEN=1800
OUTPUT_LEN=20 OUTPUT_LEN=20
MAX_MODEL_LEN=2048
MIN_CACHE_HIT_PCT=60 MIN_CACHE_HIT_PCT=60
MAX_LATENCY_ALLOWED_MS=500 MAX_LATENCY_ALLOWED_MS=500
``` ```

14
benchmarks/auto_tune/auto_tune.sh
View File

@ -4,13 +4,15 @@
# See details in README (benchmarks/auto_tune/README.md). # See details in README (benchmarks/auto_tune/README.md).
TAG=$(date +"%Y_%m_%d_%H_%M") TAG=$(date +"%Y_%m_%d_%H_%M")
BASE="" SCRIPT_DIR=$( cd -- "$( dirname -- "${BASH_SOURCE[0]}" )" &> /dev/null && pwd )
BASE="$SCRIPT_DIR/../../.."
MODEL="meta-llama/Llama-3.1-8B-Instruct" MODEL="meta-llama/Llama-3.1-8B-Instruct"
SYSTEM="TPU" SYSTEM="TPU"
TP=1 TP=1
DOWNLOAD_DIR="" DOWNLOAD_DIR=""
INPUT_LEN=4000 INPUT_LEN=4000
OUTPUT_LEN=16 OUTPUT_LEN=16
MAX_MODEL_LEN=4096
MIN_CACHE_HIT_PCT=0 MIN_CACHE_HIT_PCT=0
MAX_LATENCY_ALLOWED_MS=100000000000 MAX_LATENCY_ALLOWED_MS=100000000000
NUM_SEQS_LIST="128 256" NUM_SEQS_LIST="128 256"
@ -36,6 +38,13 @@ current_hash=$(git rev-parse HEAD)
echo "hash:$current_hash" >> "$RESULT" echo "hash:$current_hash" >> "$RESULT"
echo "current_hash: $current_hash" echo "current_hash: $current_hash"
TOTAL_LEN=$((INPUT_LEN + OUTPUT_LEN))
RED='\033[0;31m'
if (( TOTAL_LEN > MAX_MODEL_LEN )); then
echo -e "${RED}FAILED: INPUT_LEN($INPUT_LEN) + OUTPUT_LEN($OUTPUT_LEN) = $TOTAL_LEN, which is > MAX_MODEL_LEN = $MAX_MODEL_LEN.\033[0m" >&2
exit 1
fi
best_throughput=0 best_throughput=0
best_max_num_seqs=0 best_max_num_seqs=0
best_num_batched_tokens=0 best_num_batched_tokens=0
@ -60,7 +69,7 @@ start_server() {
--enable-prefix-caching \ --enable-prefix-caching \
--load-format dummy \ --load-format dummy \
--download-dir "$DOWNLOAD_DIR" \ --download-dir "$DOWNLOAD_DIR" \
--max-model-len $(( INPUT_LEN+OUTPUT_LEN )) > "$vllm_log" 2>&1 & --max-model-len $MAX_MODEL_LEN > "$vllm_log" 2>&1 &
# wait for 10 minutes... # wait for 10 minutes...
server_started=0 server_started=0
@ -245,4 +254,3 @@ done
echo "finish permutations" echo "finish permutations"
echo "best_max_num_seqs: $best_max_num_seqs, best_num_batched_tokens: $best_num_batched_tokens, best_throughput: $best_throughput, profile saved in: $PROFILE_PATH" echo "best_max_num_seqs: $best_max_num_seqs, best_num_batched_tokens: $best_num_batched_tokens, best_throughput: $best_throughput, profile saved in: $PROFILE_PATH"
echo "best_max_num_seqs: $best_max_num_seqs, best_num_batched_tokens: $best_num_batched_tokens, best_throughput: $best_throughput, profile saved in: $PROFILE_PATH" >> "$RESULT" echo "best_max_num_seqs: $best_max_num_seqs, best_num_batched_tokens: $best_num_batched_tokens, best_throughput: $best_throughput, profile saved in: $PROFILE_PATH" >> "$RESULT"