@tduncan / Nvidia-Optimization-Stack-VLLM Mirror · read-only / commits / 067c34a155
Fork 0

A high-throughput and memory-efficient inference and serving engine for LLMs

Mirrored from https://github.com/tduncan-lum/Nvidia-Optimization-Stack---VLLM.git

Synced Nov 14, 22:35 UTC

docs: remove deprecated disable-log-requests flag (#22113)

Signed-off-by: Roger Wang <hey@rogerw.me>

Roger Wang committed Aug 2, 2025 at 00:19 UTC 067c34a1559400e956311f067ddd185f54207a2b
19 files changed +14 -41
.buildkite/scripts/tpu/run_bm.sh
-1
@@ -44,7 +44,6 @@ echo
44
45 VLLM_USE_V1=1 vllm serve $MODEL \
46 --seed 42 \
47 - --disable-log-requests \
47 --max-num-seqs $MAX_NUM_SEQS \
48 --max-num-batched-tokens $MAX_NUM_BATCHED_TOKENS \
49 --tensor-parallel-size $TENSOR_PARALLEL_SIZE \
benchmarks/README.md
+5 -5
@@ -91,7 +91,7 @@ become available.
91 First start serving your model
92
93 ```bash
94 -vllm serve NousResearch/Hermes-3-Llama-3.1-8B --disable-log-requests
94 +vllm serve NousResearch/Hermes-3-Llama-3.1-8B
95 ```
96
97 Then run the benchmarking script
@@ -146,7 +146,7 @@ If the dataset you want to benchmark is not supported yet in vLLM, even then you
146
147 ```bash
148 # start server
149 -VLLM_USE_V1=1 vllm serve meta-llama/Llama-3.1-8B-Instruct --disable-log-requests
149 +VLLM_USE_V1=1 vllm serve meta-llama/Llama-3.1-8B-Instruct
150 ```
151
152 ```bash
@@ -171,7 +171,7 @@ You can skip applying chat template if your data already has it by using `--cust
171
172 ```bash
173 # need a model with vision capability here
174 -vllm serve Qwen/Qwen2-VL-7B-Instruct --disable-log-requests
174 +vllm serve Qwen/Qwen2-VL-7B-Instruct
175 ```
176
177 ```bash
@@ -205,7 +205,7 @@ vllm bench serve \
205 ### Other HuggingFaceDataset Examples
206
207 ```bash
208 -vllm serve Qwen/Qwen2-VL-7B-Instruct --disable-log-requests
208 +vllm serve Qwen/Qwen2-VL-7B-Instruct
209 ```
210
211 `lmms-lab/LLaVA-OneVision-Data`:
@@ -430,7 +430,7 @@ Benchmark the performance of structured output generation (JSON, grammar, regex)
430 ### Server Setup
431
432 ```bash
433 -vllm serve NousResearch/Hermes-3-Llama-3.1-8B --disable-log-requests
433 +vllm serve NousResearch/Hermes-3-Llama-3.1-8B
434 ```
435
436 ### JSON Schema Benchmark
benchmarks/auto_tune/auto_tune.sh
-1
@@ -60,7 +60,6 @@ start_server() {
60 pkill -f vllm
61
62 VLLM_USE_V1=1 VLLM_SERVER_DEV_MODE=1 VLLM_TORCH_PROFILER_DIR=$profile_dir vllm serve $MODEL \
63 - --disable-log-requests \
63 --port 8004 \
64 --gpu-memory-utilization $gpu_memory_utilization \
65 --max-num-seqs $max_num_seqs \
benchmarks/benchmark_serving.py
+1 -2
@@ -5,8 +5,7 @@ r"""Benchmark online serving throughput.
5 On the server side, run one of the following commands:
6 vLLM OpenAI API server
7 vllm serve <your_model> \
8 - --swap-space 16 \
9 - --disable-log-requests
8 + --swap-space 16
9
10 On the client side, run:
11 python benchmarks/benchmark_serving.py \
benchmarks/benchmark_serving_structured_output.py
+1 -1
@@ -4,7 +4,7 @@ r"""Benchmark online serving throughput with structured outputs.
4
5 On the server side, run one of the following commands:
6 (vLLM OpenAI API server)
7 - vllm serve <your_model> --disable-log-requests
7 + vllm serve <your_model>
8
9 On the client side, run:
10 python benchmarks/benchmark_serving_structured_output.py \
docs/design/p2p_nccl_connector.md
-8
@@ -109,7 +109,6 @@ python3 disagg_proxy_p2p_nccl_xpyd.py &
109 --max-num-seqs 256 \
110 --trust-remote-code \
111 --gpu-memory-utilization 0.9 \
112 - --disable-log-request \
112 --kv-transfer-config \
113 '{"kv_connector":"P2pNcclConnector","kv_role":"kv_producer","kv_buffer_size":"1e1","kv_port":"21001","kv_connector_extra_config":{"proxy_ip":"10.0.1.1","proxy_port":"30001","http_port":"20001"}}' > /var/vllm.log 2>&1 &
114 ```
@@ -131,7 +130,6 @@ python3 disagg_proxy_p2p_nccl_xpyd.py &
130 --max-num-seqs 256 \
131 --trust-remote-code \
132 --gpu-memory-utilization 0.7 \
134 - --disable-log-request \
133 --kv-transfer-config \
134 '{"kv_connector":"P2pNcclConnector","kv_role":"kv_consumer","kv_buffer_size":"8e9","kv_port":"22001","kv_connector_extra_config":{"proxy_ip":"10.0.1.1","proxy_port":"30001","http_port":"20002"}}' > /var/vllm.log 2>&1 &
135 ```
@@ -153,7 +151,6 @@ python3 disagg_proxy_p2p_nccl_xpyd.py &
151 --max-num-seqs 256 \
152 --trust-remote-code \
153 --gpu-memory-utilization 0.7 \
156 - --disable-log-request \
154 --kv-transfer-config \
155 '{"kv_connector":"P2pNcclConnector","kv_role":"kv_consumer","kv_buffer_size":"8e9","kv_port":"23001","kv_connector_extra_config":{"proxy_ip":"10.0.1.1","proxy_port":"30001","http_port":"20003"}}' > /var/vllm.log 2>&1 &
156 ```
@@ -175,7 +172,6 @@ python3 disagg_proxy_p2p_nccl_xpyd.py &
172 --max-num-seqs 256 \
173 --trust-remote-code \
174 --gpu-memory-utilization 0.7 \
178 - --disable-log-request \
175 --kv-transfer-config \
176 '{"kv_connector":"P2pNcclConnector","kv_role":"kv_consumer","kv_buffer_size":"8e9","kv_port":"24001","kv_connector_extra_config":{"proxy_ip":"10.0.1.1","proxy_port":"30001","http_port":"20004"}}' > /var/vllm.log 2>&1 &
177 ```
@@ -206,7 +202,6 @@ python3 disagg_proxy_p2p_nccl_xpyd.py &
202 --max-num-seqs 256 \
203 --trust-remote-code \
204 --gpu-memory-utilization 0.9 \
209 - --disable-log-request \
205 --kv-transfer-config \
206 '{"kv_connector":"P2pNcclConnector","kv_role":"kv_producer","kv_buffer_size":"1e1","kv_port":"21001","kv_connector_extra_config":{"proxy_ip":"10.0.1.1","proxy_port":"30001","http_port":"20001"}}' > /var/vllm.log 2>&1 &
207 ```
@@ -228,7 +223,6 @@ python3 disagg_proxy_p2p_nccl_xpyd.py &
223 --max-num-seqs 256 \
224 --trust-remote-code \
225 --gpu-memory-utilization 0.9 \
231 - --disable-log-request \
226 --kv-transfer-config \
227 '{"kv_connector":"P2pNcclConnector","kv_role":"kv_producer","kv_buffer_size":"1e1","kv_port":"22001","kv_connector_extra_config":{"proxy_ip":"10.0.1.1","proxy_port":"30001","http_port":"20002"}}' > /var/vllm.log 2>&1 &
228 ```
@@ -250,7 +244,6 @@ python3 disagg_proxy_p2p_nccl_xpyd.py &
244 --max-num-seqs 256 \
245 --trust-remote-code \
246 --gpu-memory-utilization 0.9 \
253 - --disable-log-request \
247 --kv-transfer-config \
248 '{"kv_connector":"P2pNcclConnector","kv_role":"kv_producer","kv_buffer_size":"1e1","kv_port":"23001","kv_connector_extra_config":{"proxy_ip":"10.0.1.1","proxy_port":"30001","http_port":"20003"}}' > /var/vllm.log 2>&1 &
249 ```
@@ -272,7 +265,6 @@ python3 disagg_proxy_p2p_nccl_xpyd.py &
265 --max-num-seqs 256 \
266 --trust-remote-code \
267 --gpu-memory-utilization 0.7 \
275 - --disable-log-request \
268 --kv-transfer-config \
269 '{"kv_connector":"P2pNcclConnector","kv_role":"kv_consumer","kv_buffer_size":"8e9","kv_port":"24001","kv_connector_extra_config":{"proxy_ip":"10.0.1.1","proxy_port":"30001","http_port":"20004"}}' > /var/vllm.log 2>&1 &
270 ```
docs/models/supported_models.md
+1 -1
@@ -255,7 +255,7 @@ export https_proxy=http://your.proxy.server:port
255 https_proxy=http://your.proxy.server:port huggingface-cli download <model_name>
256
257 # or use vllm cmd directly
258 -https_proxy=http://your.proxy.server:port vllm serve <model_name> --disable-log-requests
258 +https_proxy=http://your.proxy.server:port vllm serve <model_name>
259 ```
260
261 - Set the proxy in Python interpreter:
examples/online_serving/disaggregated_serving_p2p_nccl_xpyd/disagg_example_p2p_nccl_xpyd.sh
-2
@@ -178,7 +178,6 @@ main() {
178 --max-num-seqs 256 \
179 --trust-remote-code \
180 --gpu-memory-utilization 0.9 \
181 - --disable-log-request \
181 --kv-transfer-config \
182 "{\"kv_connector\":\"P2pNcclConnector\",\"kv_role\":\"kv_producer\",\"kv_buffer_size\":\"1e1\",\"kv_port\":\"$kv_port\",\"kv_connector_extra_config\":{\"proxy_ip\":\"0.0.0.0\",\"proxy_port\":\"$PROXY_PORT\",\"http_port\":\"$port\",\"send_type\":\"PUT_ASYNC\",\"nccl_num_channels\":\"16\"}}" > prefill$((i+1)).log 2>&1 &
183 PIDS+=($!)
@@ -207,7 +206,6 @@ main() {
206 --max-num-seqs 256 \
207 --trust-remote-code \
208 --gpu-memory-utilization 0.7 \
210 - --disable-log-request \
209 --kv-transfer-config \
210 "{\"kv_connector\":\"P2pNcclConnector\",\"kv_role\":\"kv_consumer\",\"kv_buffer_size\":\"8e9\",\"kv_port\":\"$kv_port\",\"kv_connector_extra_config\":{\"proxy_ip\":\"0.0.0.0\",\"proxy_port\":\"$PROXY_PORT\",\"http_port\":\"$port\",\"send_type\":\"PUT_ASYNC\",\"nccl_num_channels\":\"16\"}}" > decode$((i+1)).log 2>&1 &
211 PIDS+=($!)
examples/online_serving/prometheus_grafana/README.md
+1 -2
@@ -13,8 +13,7 @@ Prometheus metric logging is enabled by default in the OpenAI-compatible server.
13
14 ```bash
15 vllm serve mistralai/Mistral-7B-v0.1 \
16 - --max-model-len 2048 \
17 - --disable-log-requests
16 + --max-model-len 2048
17 ```
18
19 Launch Prometheus and Grafana servers with `docker compose`:
examples/others/lmcache/disagg_prefill_lmcache_v1/disagg_vllm_launcher.sh
-2
@@ -28,7 +28,6 @@ if [[ $1 == "prefiller" ]]; then
28 CUDA_VISIBLE_DEVICES=0 \
29 vllm serve $MODEL \
30 --port 8100 \
31 - --disable-log-requests \
31 --enforce-eager \
32 --kv-transfer-config \
33 '{"kv_connector":"LMCacheConnectorV1","kv_role":"kv_producer","kv_connector_extra_config": {"discard_partial_chunks": false, "lmcache_rpc_port": "producer1"}}'
@@ -46,7 +45,6 @@ elif [[ $1 == "decoder" ]]; then
45 CUDA_VISIBLE_DEVICES=1 \
46 vllm serve $MODEL \
47 --port 8200 \
49 - --disable-log-requests \
48 --enforce-eager \
49 --kv-transfer-config \
50 '{"kv_connector":"LMCacheConnectorV1","kv_role":"kv_consumer","kv_connector_extra_config": {"discard_partial_chunks": false, "lmcache_rpc_port": "consumer1"}}'
tests/entrypoints/openai/correctness/test_lmeval.py
+1 -1
@@ -22,7 +22,7 @@ TASK = "gsm8k"
22 FILTER = "exact_match,strict-match"
23 RTOL = 0.03
24 EXPECTED_VALUE = 0.54
25 -DEFAULT_ARGS = ["--max-model-len", "4096", "--disable-log-requests"]
25 +DEFAULT_ARGS = ["--max-model-len", "4096"]
26 MORE_ARGS_LIST = [
27 [], # Default
28 ["--enable-chunked-prefill"], # Chunked
tests/entrypoints/openai/test_chunked_prompt.py
-2
@@ -26,8 +26,6 @@ def server():
26 "--enable-chunked-prefill",
27 "--max-num-batched-tokens",
28 "1000",
29 - # large prompts create a lot of output
30 - "--disable-log-requests",
29 ]
30
31 with RemoteOpenAIServer(MODEL_NAME, args) as remote_server:
tests/models/quantization/test_bitsandbytes.py
-1
@@ -102,7 +102,6 @@ def test_load_tp_4bit_bnb_model(hf_runner, vllm_runner, example_prompts,
102 def test_load_pp_4bit_bnb_model(model_name, description) -> None:
103 common_args = [
104 "--disable-log-stats",
105 - "--disable-log-requests",
105 "--dtype",
106 "bfloat16",
107 "--enable-prefix-caching",
tests/v1/kv_connector/nixl_integration/run_accuracy_test.sh
-2
@@ -88,7 +88,6 @@ run_tests_for_model() {
88 BASE_CMD="CUDA_VISIBLE_DEVICES=$GPU_ID VLLM_NIXL_SIDE_CHANNEL_PORT=$SIDE_CHANNEL_PORT vllm serve $model_name \
89 --port $PORT \
90 --enforce-eager \
91 - --disable-log-requests \
91 --gpu-memory-utilization 0.2 \
92 --tensor-parallel-size $PREFILLER_TP_SIZE \
93 --kv-transfer-config '{\"kv_connector\":\"NixlConnector\",\"kv_role\":\"kv_both\"}'"
@@ -121,7 +120,6 @@ run_tests_for_model() {
120 BASE_CMD="CUDA_VISIBLE_DEVICES=$GPU_ID VLLM_NIXL_SIDE_CHANNEL_PORT=$SIDE_CHANNEL_PORT vllm serve $model_name \
121 --port $PORT \
122 --enforce-eager \
124 - --disable-log-requests \
123 --gpu-memory-utilization 0.2 \
124 --tensor-parallel-size $DECODER_TP_SIZE \
125 --kv-transfer-config '{\"kv_connector\":\"NixlConnector\",\"kv_role\":\"kv_both\"}'"
tests/v1/kv_connector/nixl_integration/run_edge_case_test.sh
-2
@@ -57,7 +57,6 @@ run_tests_for_model() {
57 BASE_CMD="CUDA_VISIBLE_DEVICES=0 VLLM_NIXL_SIDE_CHANNEL_PORT=5559 vllm serve $model_name \
58 --port $PREFILL_PORT \
59 --enforce-eager \
60 - --disable-log-requests \
60 --gpu-memory-utilization 0.2 \
61 --kv-transfer-config '{\"kv_connector\":\"NixlConnector\",\"kv_role\":\"kv_both\"}'"
62
@@ -76,7 +75,6 @@ run_tests_for_model() {
75 BASE_CMD="CUDA_VISIBLE_DEVICES=1 VLLM_NIXL_SIDE_CHANNEL_PORT=6000 vllm serve $model_name \
76 --port $DECODE_PORT \
77 --enforce-eager \
79 - --disable-log-requests \
78 --gpu-memory-utilization 0.2 \
79 --kv-transfer-config '{\"kv_connector\":\"NixlConnector\",\"kv_role\":\"kv_both\"}'"
80
tests/v1/kv_connector/nixl_integration/run_tpu_disagg_accuracy_test.sh
-3
@@ -63,7 +63,6 @@ launch_baseline() {
63 --seed 42 \
64 --block-size ${BLOCK_SIZE} \
65 --gpu-memory-utilization 0.5 \
66 - --disable-log-requests \
66 --enforce-eager"
67 echo ${BASELINE_BASE_CMD}
68 ssh -tt ${BASELINE_HOST} "${BASELINE_BASE_CMD}" &
@@ -87,7 +86,6 @@ launch_pd() {
86 --block-size ${BLOCK_SIZE} \
87 --enforce-eager \
88 --gpu-memory-utilization 0.5 \
90 - --disable-log-requests \
89 --kv-transfer-config '{\"kv_connector\":\"NixlConnector\",\"kv_role\":\"kv_both\",\"kv_buffer_device\":\"cpu\"}'"
90
91
@@ -106,7 +104,6 @@ launch_pd() {
104 --block-size ${BLOCK_SIZE} \
105 --enforce-eager \
106 --gpu-memory-utilization 0.5 \
109 - --disable-log-requests \
107 --kv-transfer-config '{\"kv_connector\":\"NixlConnector\",\"kv_role\":\"kv_both\",\"kv_buffer_device\":\"cpu\"}'"
108
109 echo ${PREFILL_BASE_CMD}
tests/v1/kv_connector/nixl_integration/run_tpu_edge_case_test.sh
-2
@@ -68,7 +68,6 @@ launch_pd() {
68 --block-size ${BLOCK_SIZE} \
69 --enforce-eager \
70 --gpu-memory-utilization 0.5 \
71 - --disable-log-requests \
71 --kv-transfer-config '{\"kv_connector\":\"NixlConnector\",\"kv_role\":\"kv_both\",\"kv_buffer_device\":\"cpu\"}'"
72
73
@@ -87,7 +86,6 @@ launch_pd() {
86 --block-size ${BLOCK_SIZE} \
87 --enforce-eager \
88 --gpu-memory-utilization 0.5 \
90 - --disable-log-requests \
89 --kv-transfer-config '{\"kv_connector\":\"NixlConnector\",\"kv_role\":\"kv_both\",\"kv_buffer_device\":\"cpu\"}'"
90
91 echo ${PREFILL_BASE_CMD}
tests/v1/sample/test_logprobs_e2e.py
+1 -1
@@ -15,7 +15,7 @@ EXPECTED_VALUE = 0.62
15 MODEL = "meta-llama/Llama-3.2-1B-Instruct"
16 MODEL_ARGS = f"pretrained={MODEL},enforce_eager=True,enable_prefix_caching=False,gpu_memory_utilization=0.8" # noqa: E501
17 SERVER_ARGS = [
18 - "--enforce_eager", "--no_enable_prefix_caching", "--disable-log-requests",
18 + "--enforce_eager", "--no_enable_prefix_caching",
19 "--gpu-memory-utilization=0.8"
20 ]
21 NUM_CONCURRENT = 100
vllm/utils/__init__.py
+3 -2
@@ -1673,8 +1673,9 @@ class FlexibleArgumentParser(ArgumentParser):
1673 # Special case warning because the warning below won't trigger
1674 # if –-disable-log-requests because its value is default.
1675 logger.warning_once(
1676 - "argument '--disable-log-requests' is deprecated. This "
1677 - "will be removed in v0.12.0.")
1676 + "argument '--disable-log-requests' is deprecated and "
1677 + "replaced with '--enable-log-requests'. This will be "
1678 + "removed in v0.12.0.")
1679 namespace, args = super().parse_known_args(args, namespace)
1680 for action in FlexibleArgumentParser._deprecated:
1681 if (hasattr(namespace, dest := action.dest)