minke.yu/sglang
1
0
Fork 0
Code Issues Pull Requests Actions 12 Packages Projects Releases Wiki Activity

[Config] Retire get_global_server_args, and clear the deprecated flags that have a replacement (#38375)

Browse Source
This commit is contained in:
Cheng Wan
2026-09-08 16:42:12 -07:00
committed by GitHub
parent 5177a3ec08
commit db272201a2
213 changed files with 625 additions and 892 deletions
Download Patch File Download Diff File Expand all files Collapse all files
docs/src/snippets/configs/Qwen/qwen3.8-27b.jsx
+6 -6
View File
@@ -589,7 +589,7 @@ export const config = {
// RTX 5090 32GB. NVFP4 is the only checkpoint that fits (FP8 does not
// boot — total_rest_memory negative at every mem-fraction, measured —
// and BF16 does not fit). Published operating point is ONE request in
// flight; --cuda-graph-max-bs 1 also protects the token pool (default
// flight; --cuda-graph-max-bs-decode 1 also protects the token pool (default
// capture set costs 39,247 -> 37,347 and K 8 -> 7). The `warn` below
// carries the user-facing guidance for raising concurrency.
match: { hw: "rtx5090", variant: "default", quant: "nvfp4-bf16-head", nodes: "single" },
@@ -598,7 +598,7 @@ export const config = {
// multi-user deployment unaware.
warn:
"This recipe serves ONE request at a time: --max-running-requests 1 " +
"and --cuda-graph-max-bs 1 pin it to the validated single-stream " +
"and --cuda-graph-max-bs-decode 1 pin it to the validated single-stream " +
"envelope. To handle more concurrent requests, raise both flags " +
"together and re-derive --mamba-full-memory-ratio (and mem-fraction) " +
"with the [Mamba ratio calculator](#mamba-ratio-calculator) — on this " +
@@ -611,7 +611,7 @@ export const config = {
"--mem-fraction-static 0.9",
"--attention-backend flashinfer",
"--max-running-requests 1",
"--cuda-graph-max-bs 1",
"--cuda-graph-max-bs-decode 1",
"--reasoning-parser qwen3",
"--tool-call-parser qwen3_coder",
"--host {{HOST_IP}}",
@@ -624,7 +624,7 @@ export const config = {
// RTX 5090 32GB. NVFP4 is the only checkpoint that fits (FP8 does not
// boot — total_rest_memory negative at every mem-fraction, measured —
// and BF16 does not fit). Published operating point is ONE request in
// flight; --cuda-graph-max-bs 1 also protects the token pool (default
// flight; --cuda-graph-max-bs-decode 1 also protects the token pool (default
// capture set costs 39,247 -> 37,347 and K 8 -> 7). The `warn` below
// carries the user-facing guidance for raising concurrency.
match: { hw: "rtx5090", variant: "default", quant: "nvfp4-fp4-head", nodes: "single" },
@@ -633,7 +633,7 @@ export const config = {
// multi-user deployment unaware.
warn:
"This recipe serves ONE request at a time: --max-running-requests 1 " +
"and --cuda-graph-max-bs 1 pin it to the validated single-stream " +
"and --cuda-graph-max-bs-decode 1 pin it to the validated single-stream " +
"envelope. To handle more concurrent requests, raise both flags " +
"together and re-derive --mamba-full-memory-ratio (and mem-fraction) " +
"with the [Mamba ratio calculator](#mamba-ratio-calculator) — on this " +
@@ -646,7 +646,7 @@ export const config = {
"--mem-fraction-static 0.9",
"--attention-backend flashinfer",
"--max-running-requests 1",
"--cuda-graph-max-bs 1",
"--cuda-graph-max-bs-decode 1",
"--reasoning-parser qwen3",
"--tool-call-parser qwen3_coder",
"--host {{HOST_IP}}",
docs/src/snippets/configs/Qwen/qwen3.8.jsx
+2 -2
View File
@@ -233,8 +233,8 @@ export const config = {
// ----- Card: "PD Disaggregation" -----
// Role flags follow the P/D bundle's own prefill and decode workers. Two
// flags those recipes carry are deliberately not emitted:
// --prefill-round-robin-balance is a DeprecatedAction on current SGLang and
// does nothing, and --mamba-track-interval is context-dependent (the source
// --prefill-round-robin-balance no longer exists on current SGLang (it was
// a deprecated no-op), and --mamba-track-interval is context-dependent (the source
// recipes set it equal to their context cap) so a fixed value here would be
// wrong for cells serving the native window.
pdDisagg: {
docs/src/snippets/configs/inclusionAI/ling-3.0-flash.jsx
+5 -5
View File
@@ -114,7 +114,7 @@ sgl-eval run gsm8k \\
hicache: {
defaultBackend: "mooncake",
requiredFlags: [
"--mamba-scheduler-strategy extra_buffer",
"--mamba-radix-cache-strategy extra_buffer",
"--enable-cache-report",
],
backends: [
@@ -618,7 +618,7 @@ sgl-eval run gsm8k \\
"--hicache-storage-backend mooncake",
"--hicache-io-backend direct",
"--hicache-mem-layout page_first_direct",
"--mamba-scheduler-strategy extra_buffer",
"--mamba-radix-cache-strategy extra_buffer",
"--enable-cache-report",
"--hicache-storage-prefetch-policy wait_complete",
"--hicache-storage-backend-extra-config '{\"hicache_storage_pass_prefix_keys\":true}'",
@@ -647,7 +647,7 @@ sgl-eval run gsm8k \\
"--hicache-storage-backend mooncake",
"--hicache-io-backend direct",
"--hicache-mem-layout page_first_direct",
"--mamba-scheduler-strategy extra_buffer",
"--mamba-radix-cache-strategy extra_buffer",
"--enable-cache-report",
"--hicache-storage-prefetch-policy wait_complete",
"--hicache-storage-backend-extra-config '{\"hicache_storage_pass_prefix_keys\":true}'",
@@ -675,7 +675,7 @@ sgl-eval run gsm8k \\
"--hicache-storage-backend mooncake",
"--hicache-io-backend direct",
"--hicache-mem-layout page_first_direct",
"--mamba-scheduler-strategy extra_buffer",
"--mamba-radix-cache-strategy extra_buffer",
"--enable-cache-report",
"--hicache-storage-prefetch-policy wait_complete",
"--hicache-storage-backend-extra-config '{\"hicache_storage_pass_prefix_keys\":true}'",
@@ -704,7 +704,7 @@ sgl-eval run gsm8k \\
"--hicache-storage-backend mooncake",
"--hicache-io-backend direct",
"--hicache-mem-layout page_first_direct",
"--mamba-scheduler-strategy extra_buffer",
"--mamba-radix-cache-strategy extra_buffer",
"--enable-cache-report",
"--hicache-storage-prefetch-policy wait_complete",
"--hicache-storage-backend-extra-config '{\"hicache_storage_pass_prefix_keys\":true}'",
docs/src/snippets/configs/meituan-longcat/longcat-2.0.jsx
+4 -4
View File
@@ -134,7 +134,7 @@ sgl-eval run gsm8k \\
"--max-running-requests 64",
"--mem-fraction-static 0.92",
"--chunked-prefill-size 2048",
"--nsa-prefill-backend fa3",
"--dsa-prefill-backend fa3",
"--kv-cache-dtype bfloat16",
"--model-loader-extra-config '{\"enable_multithread_load\":true,\"num_threads\":12}'",
"--host {{HOST_IP}}",
@@ -154,7 +154,7 @@ sgl-eval run gsm8k \\
"--max-running-requests 64",
"--mem-fraction-static 0.92",
"--chunked-prefill-size 2048",
"--nsa-prefill-backend fa3",
"--dsa-prefill-backend fa3",
"--kv-cache-dtype bfloat16",
"--model-loader-extra-config '{\"enable_multithread_load\":true,\"num_threads\":12}'",
"--host {{HOST_IP}}",
@@ -173,7 +173,7 @@ sgl-eval run gsm8k \\
"--max-running-requests 64",
"--mem-fraction-static 0.92",
"--chunked-prefill-size 2048",
"--nsa-prefill-backend fa3",
"--dsa-prefill-backend fa3",
"--kv-cache-dtype bfloat16",
"--model-loader-extra-config '{\"enable_multithread_load\":true,\"num_threads\":12}'",
"--host {{HOST_IP}}",
@@ -192,7 +192,7 @@ sgl-eval run gsm8k \\
"--max-running-requests 64",
"--mem-fraction-static 0.92",
"--chunked-prefill-size 2048",
"--nsa-prefill-backend fa3",
"--dsa-prefill-backend fa3",
"--kv-cache-dtype bfloat16",
"--model-loader-extra-config '{\"enable_multithread_load\":true,\"num_threads\":12}'",
"--host {{HOST_IP}}",
docs/src/snippets/configs/moonshotai/kimi-k3.jsx
+5 -5
View File
@@ -1183,7 +1183,7 @@ export const config = {
"--kv-cache-dtype fp8_e4m3",
"--dtype bfloat16",
"--mem-fraction-static 0.85",
"--cuda-graph-max-bs 256",
"--cuda-graph-max-bs-decode 256",
"--reasoning-parser kimi_k3",
"--tool-call-parser kimi_k3",
"--host {{HOST_IP}}",
@@ -1210,7 +1210,7 @@ export const config = {
"--kv-cache-dtype fp8_e4m3",
"--dtype bfloat16",
"--mem-fraction-static 0.85",
"--cuda-graph-max-bs 256",
"--cuda-graph-max-bs-decode 256",
"--reasoning-parser kimi_k3",
"--tool-call-parser kimi_k3",
"--host {{HOST_IP}}",
@@ -1959,7 +1959,7 @@ export const config = {
"--kv-cache-dtype fp8_e4m3",
"--dtype bfloat16",
"--mem-fraction-static 0.85",
"--cuda-graph-max-bs 256",
"--cuda-graph-max-bs-decode 256",
"--reasoning-parser kimi_k3",
"--tool-call-parser kimi_k3",
"--disaggregation-mode decode",
@@ -1987,7 +1987,7 @@ export const config = {
"--kv-cache-dtype fp8_e4m3",
"--dtype bfloat16",
"--mem-fraction-static 0.85",
"--cuda-graph-max-bs 256",
"--cuda-graph-max-bs-decode 256",
"--reasoning-parser kimi_k3",
"--tool-call-parser kimi_k3",
"--disaggregation-mode decode",
@@ -2379,7 +2379,7 @@ export const config = {
"--enable-dp-lm-head",
"--mem-fraction-static 0.78",
"--chunked-prefill-size 16384",
"--cuda-graph-bs 2 4 8 16",
"--cuda-graph-bs-decode 2 4 8 16",
"--max-running-requests 64",
"--max-mamba-cache-size 64",
"--moe-a2a-backend deepep",
docs/src/snippets/configs/openbmb/minicpm5-2b-benchmarks.jsx
+1 -1
View File
@@ -4,7 +4,7 @@
// RTX 5090 32GB (`zijiexia-5090-1gpu`, `lmsysorg/sglang:dev` @ 30705c004c) with
// `python3 -m sglang.bench_serving`, random 1024/1024, against the exact cell
// command on this page — including `--mem-fraction-static 0.75
// --cuda-graph-max-bs 128`, which is what keeps concurrency 64 and 128
// --cuda-graph-max-bs-decode 128`, which is what keeps concurrency 64 and 128
// graph-backed (see the model page's Configuration Tips).
//
// `tokens_per_sec_per_gpu` is total (input+output) per GPU = the measured
docs/src/snippets/configs/openbmb/minicpm5-2b.jsx
+2 -2
View File
@@ -206,7 +206,7 @@ export const config = {
"--reasoning-parser qwen3",
"--tool-call-parser minicpm5",
"--mem-fraction-static 0.75",
"--cuda-graph-max-bs 128",
"--cuda-graph-max-bs-decode 128",
"--host {{HOST_IP}}",
"--port {{PORT}}",
],
@@ -225,7 +225,7 @@ export const config = {
"--reasoning-parser qwen3",
"--tool-call-parser minicpm5",
"--mem-fraction-static 0.30",
"--cuda-graph-max-bs 128",
"--cuda-graph-max-bs-decode 128",
"--host {{HOST_IP}}",
"--port {{PORT}}",
],
docs/src/snippets/configs/zai-org/glm-5.2.jsx
+15 -15
View File
@@ -772,7 +772,7 @@ sgl-eval run aime25 \\
"--chunked-prefill-size 8192",
"--mem-fraction-static 0.85",
"--max-running-requests 16",
"--cuda-graph-max-bs 16",
"--cuda-graph-max-bs-decode 16",
"--max-prefill-tokens 8192",
"--host {{HOST_IP}}",
"--port {{PORT}}",
@@ -837,7 +837,7 @@ sgl-eval run aime25 \\
"--chunked-prefill-size 8192",
"--mem-fraction-static 0.85",
"--max-running-requests 16",
"--cuda-graph-max-bs 16",
"--cuda-graph-max-bs-decode 16",
"--max-prefill-tokens 8192",
"--host {{HOST_IP}}",
"--port {{PORT}}",
@@ -927,7 +927,7 @@ sgl-eval run aime25 \\
"--dsa-decode-backend tilelang",
"--chunked-prefill-size 32768",
"--mem-fraction-static 0.85",
"--cuda-graph-max-bs 128",
"--cuda-graph-max-bs-decode 128",
"--max-running-requests 80",
"--watchdog-timeout 1200",
"--host {{HOST_IP}}",
@@ -944,7 +944,7 @@ sgl-eval run aime25 \\
"--dsa-prefill-backend tilelang",
"--dsa-decode-backend tilelang",
"--mem-fraction-static 0.85",
"--cuda-graph-max-bs 256",
"--cuda-graph-max-bs-decode 256",
"--max-running-requests 256",
"--watchdog-timeout 1200",
"--host {{HOST_IP}}",
@@ -978,7 +978,7 @@ sgl-eval run aime25 \\
"--dsa-decode-backend tilelang",
"--chunked-prefill-size 32768",
"--mem-fraction-static 0.85",
"--cuda-graph-max-bs 128",
"--cuda-graph-max-bs-decode 128",
"--max-running-requests 80",
"--watchdog-timeout 1200",
"--host {{HOST_IP}}",
@@ -995,7 +995,7 @@ sgl-eval run aime25 \\
"--dsa-prefill-backend tilelang",
"--dsa-decode-backend tilelang",
"--mem-fraction-static 0.85",
"--cuda-graph-max-bs 256",
"--cuda-graph-max-bs-decode 256",
"--max-running-requests 256",
"--watchdog-timeout 1200",
"--host {{HOST_IP}}",
@@ -1043,7 +1043,7 @@ sgl-eval run aime25 \\
"--dsa-decode-backend tilelang",
"--chunked-prefill-size 32768",
"--mem-fraction-static 0.85",
"--cuda-graph-max-bs 128",
"--cuda-graph-max-bs-decode 128",
"--max-running-requests 80",
"--watchdog-timeout 1200",
"--host {{HOST_IP}}",
@@ -1062,7 +1062,7 @@ sgl-eval run aime25 \\
"--dsa-prefill-backend tilelang",
"--dsa-decode-backend tilelang",
"--mem-fraction-static 0.85",
"--cuda-graph-max-bs 256",
"--cuda-graph-max-bs-decode 256",
"--max-running-requests 256",
"--watchdog-timeout 1200",
"--host {{HOST_IP}}",
@@ -1094,7 +1094,7 @@ sgl-eval run aime25 \\
"--speculative-num-draft-tokens 4",
"--chunked-prefill-size 131072",
"--mem-fraction-static 0.80",
"--cuda-graph-max-bs 160",
"--cuda-graph-max-bs-decode 160",
"--max-running-requests 160",
"--watchdog-timeout 1800",
"--host {{HOST_IP}}",
@@ -1128,7 +1128,7 @@ sgl-eval run aime25 \\
"--dsa-decode-backend tilelang",
"--chunked-prefill-size 32768",
"--mem-fraction-static 0.85",
"--cuda-graph-max-bs 128",
"--cuda-graph-max-bs-decode 128",
"--max-running-requests 80",
"--watchdog-timeout 1200",
"--host {{HOST_IP}}",
@@ -1145,7 +1145,7 @@ sgl-eval run aime25 \\
"--dsa-prefill-backend tilelang",
"--dsa-decode-backend tilelang",
"--mem-fraction-static 0.85",
"--cuda-graph-max-bs 256",
"--cuda-graph-max-bs-decode 256",
"--max-running-requests 256",
"--watchdog-timeout 1200",
"--host {{HOST_IP}}",
@@ -1179,7 +1179,7 @@ sgl-eval run aime25 \\
"--dsa-decode-backend tilelang",
"--chunked-prefill-size 32768",
"--mem-fraction-static 0.85",
"--cuda-graph-max-bs 128",
"--cuda-graph-max-bs-decode 128",
"--max-running-requests 80",
"--watchdog-timeout 1200",
"--host {{HOST_IP}}",
@@ -1196,7 +1196,7 @@ sgl-eval run aime25 \\
"--dsa-prefill-backend tilelang",
"--dsa-decode-backend tilelang",
"--mem-fraction-static 0.85",
"--cuda-graph-max-bs 256",
"--cuda-graph-max-bs-decode 256",
"--max-running-requests 256",
"--watchdog-timeout 1200",
"--host {{HOST_IP}}",
@@ -1230,7 +1230,7 @@ sgl-eval run aime25 \\
"--dsa-decode-backend tilelang",
"--chunked-prefill-size 32768",
"--mem-fraction-static 0.85",
"--cuda-graph-max-bs 128",
"--cuda-graph-max-bs-decode 128",
"--max-running-requests 80",
"--watchdog-timeout 1200",
"--host {{HOST_IP}}",
@@ -1247,7 +1247,7 @@ sgl-eval run aime25 \\
"--dsa-prefill-backend tilelang",
"--dsa-decode-backend tilelang",
"--mem-fraction-static 0.85",
"--cuda-graph-max-bs 256",
"--cuda-graph-max-bs-decode 256",
"--max-running-requests 256",
"--watchdog-timeout 1200",
"--host {{HOST_IP}}",
docs/src/snippets/configs/zai-org/glm-5.3-flash.jsx
+4 -4
View File
@@ -416,7 +416,7 @@ sgl-eval run gsm8k \\
"--reasoning-parser glm45",
"--tool-call-parser glm47",
"--mem-fraction-static 0.85",
"--cuda-graph-max-bs 32",
"--cuda-graph-max-bs-decode 32",
"--host {{HOST_IP}}",
"--port {{PORT}}",
],
@@ -472,7 +472,7 @@ sgl-eval run gsm8k \\
"--reasoning-parser glm45",
"--tool-call-parser glm47",
"--mem-fraction-static 0.85",
"--cuda-graph-max-bs 32",
"--cuda-graph-max-bs-decode 32",
"--host {{HOST_IP}}",
"--port {{PORT}}",
],
@@ -518,7 +518,7 @@ sgl-eval run gsm8k \\
"--reasoning-parser glm45",
"--tool-call-parser glm47",
"--mem-fraction-static 0.85",
"--cuda-graph-max-bs 32",
"--cuda-graph-max-bs-decode 32",
"--host {{HOST_IP}}",
"--port {{PORT}}",
],
@@ -564,7 +564,7 @@ sgl-eval run gsm8k \\
"--reasoning-parser glm45",
"--tool-call-parser glm47",
"--mem-fraction-static 0.85",
"--cuda-graph-max-bs 32",
"--cuda-graph-max-bs-decode 32",
"--host {{HOST_IP}}",
"--port {{PORT}}",
],
docs/src/snippets/configs/zai-org/glm-5.3.jsx
+12 -12
View File
@@ -716,7 +716,7 @@ sgl-eval run aime25 \\
"--chunked-prefill-size 8192",
"--mem-fraction-static 0.85",
"--max-running-requests 16",
"--cuda-graph-max-bs 16",
"--cuda-graph-max-bs-decode 16",
"--max-prefill-tokens 8192",
"--host {{HOST_IP}}",
"--port {{PORT}}",
@@ -758,7 +758,7 @@ sgl-eval run aime25 \\
"--chunked-prefill-size 8192",
"--mem-fraction-static 0.85",
"--max-running-requests 16",
"--cuda-graph-max-bs 16",
"--cuda-graph-max-bs-decode 16",
"--max-prefill-tokens 8192",
"--host {{HOST_IP}}",
"--port {{PORT}}",
@@ -825,7 +825,7 @@ sgl-eval run aime25 \\
"--dsa-decode-backend tilelang",
"--chunked-prefill-size 32768",
"--mem-fraction-static 0.85",
"--cuda-graph-max-bs 128",
"--cuda-graph-max-bs-decode 128",
"--max-running-requests 80",
"--watchdog-timeout 1200",
"--host {{HOST_IP}}",
@@ -842,7 +842,7 @@ sgl-eval run aime25 \\
"--dsa-prefill-backend tilelang",
"--dsa-decode-backend tilelang",
"--mem-fraction-static 0.85",
"--cuda-graph-max-bs 256",
"--cuda-graph-max-bs-decode 256",
"--max-running-requests 256",
"--watchdog-timeout 1200",
"--host {{HOST_IP}}",
@@ -876,7 +876,7 @@ sgl-eval run aime25 \\
"--dsa-decode-backend tilelang",
"--chunked-prefill-size 32768",
"--mem-fraction-static 0.85",
"--cuda-graph-max-bs 128",
"--cuda-graph-max-bs-decode 128",
"--max-running-requests 80",
"--watchdog-timeout 1200",
"--host {{HOST_IP}}",
@@ -893,7 +893,7 @@ sgl-eval run aime25 \\
"--dsa-prefill-backend tilelang",
"--dsa-decode-backend tilelang",
"--mem-fraction-static 0.85",
"--cuda-graph-max-bs 256",
"--cuda-graph-max-bs-decode 256",
"--max-running-requests 256",
"--watchdog-timeout 1200",
"--host {{HOST_IP}}",
@@ -927,7 +927,7 @@ sgl-eval run aime25 \\
"--dsa-decode-backend tilelang",
"--chunked-prefill-size 32768",
"--mem-fraction-static 0.85",
"--cuda-graph-max-bs 128",
"--cuda-graph-max-bs-decode 128",
"--max-running-requests 80",
"--watchdog-timeout 1200",
"--host {{HOST_IP}}",
@@ -944,7 +944,7 @@ sgl-eval run aime25 \\
"--dsa-prefill-backend tilelang",
"--dsa-decode-backend tilelang",
"--mem-fraction-static 0.85",
"--cuda-graph-max-bs 256",
"--cuda-graph-max-bs-decode 256",
"--max-running-requests 256",
"--watchdog-timeout 1200",
"--host {{HOST_IP}}",
@@ -978,7 +978,7 @@ sgl-eval run aime25 \\
"--dsa-decode-backend tilelang",
"--chunked-prefill-size 32768",
"--mem-fraction-static 0.85",
"--cuda-graph-max-bs 128",
"--cuda-graph-max-bs-decode 128",
"--max-running-requests 80",
"--watchdog-timeout 1200",
"--host {{HOST_IP}}",
@@ -995,7 +995,7 @@ sgl-eval run aime25 \\
"--dsa-prefill-backend tilelang",
"--dsa-decode-backend tilelang",
"--mem-fraction-static 0.85",
"--cuda-graph-max-bs 256",
"--cuda-graph-max-bs-decode 256",
"--max-running-requests 256",
"--watchdog-timeout 1200",
"--host {{HOST_IP}}",
@@ -1029,7 +1029,7 @@ sgl-eval run aime25 \\
"--dsa-decode-backend tilelang",
"--chunked-prefill-size 32768",
"--mem-fraction-static 0.85",
"--cuda-graph-max-bs 128",
"--cuda-graph-max-bs-decode 128",
"--max-running-requests 80",
"--watchdog-timeout 1200",
"--host {{HOST_IP}}",
@@ -1046,7 +1046,7 @@ sgl-eval run aime25 \\
"--dsa-prefill-backend tilelang",
"--dsa-decode-backend tilelang",
"--mem-fraction-static 0.85",
"--cuda-graph-max-bs 256",
"--cuda-graph-max-bs-decode 256",
"--max-running-requests 256",
"--watchdog-timeout 1200",
"--host {{HOST_IP}}",