minke.yu/sglang
1
0
Fork 0
Code Issues Pull Requests Actions 10 Packages Projects Releases Wiki Activity

[AMD] Update v4 amd cookbook 0830 (#37092)

Browse Source
This commit is contained in:
Thomas Wang
2026-08-29 23:55:28 -07:00
committed by GitHub
parent 512df615de
commit 7399c2b558
2 changed files with 34 additions and 16 deletions
Download Patch File Download Diff File Expand all files Collapse all files
docs/cookbook/autoregressive/DeepSeek/DeepSeek-V4.mdx
+4 -4
View File
@@ -52,11 +52,11 @@ docker run --gpus all \
AMD uses the daily-updated `lmsysorg/sglang-rocm` images. You can find the latest images on [Docker Hub](https://hub.docker.com/r/lmsysorg/sglang-rocm/tags). We recommend the ROCm 7.2 version.
For example:
- **MI355X** → `lmsysorg/sglang-rocm:v0.5.18-rocm720-mi35x-20260828`
- **MI300X** → `lmsysorg/sglang-rocm:v0.5.18-rocm720-mi30x-20260828`
- **MI355X** → `lmsysorg/sglang-rocm:v0.5.18-rocm720-mi35x-20260829`
- **MI300X** → `lmsysorg/sglang-rocm:v0.5.18-rocm720-mi30x-20260829`
```bash Command
docker pull lmsysorg/sglang-rocm:v0.5.18-rocm720-{mi35x,mi30x}-20260828
docker pull lmsysorg/sglang-rocm:v0.5.18-rocm720-{mi35x,mi30x}-20260829
docker run \
--device=/dev/kfd --device=/dev/dri \
@@ -66,7 +66,7 @@ docker run \
-p 30000:30000 \
-v ~/.cache/huggingface:/root/.cache/huggingface \
--env "HF_TOKEN=<your-hf-token>" \
lmsysorg/sglang-rocm:v0.5.18-rocm720-{mi35x,mi30x}-20260828 \
lmsysorg/sglang-rocm:v0.5.18-rocm720-{mi35x,mi30x}-20260829 \
sglang serve <use args below>
```
docs/src/snippets/configs/deepseek-ai/deepseek-v4.jsx
+30 -12
View File
@@ -183,8 +183,8 @@ sgl-eval run aime25 \\
gb300: "lmsysorg/sglang:latest",
// AMD daily-updated lmsysorg/sglang-rocm images. Bump the dated tag when you
// re-verify on a newer build.
mi300x: "lmsysorg/sglang-rocm:v0.5.18-rocm720-mi30x-20260828",
mi355x: "lmsysorg/sglang-rocm:v0.5.18-rocm720-mi35x-20260828",
mi300x: "lmsysorg/sglang-rocm:v0.5.18-rocm720-mi30x-20260829",
mi355x: "lmsysorg/sglang-rocm:v0.5.18-rocm720-mi35x-20260829",
},
// Pre-selects the issue template's `model` dropdown on "Submit verified cell".
@@ -1643,7 +1643,7 @@ sgl-eval run aime25 \\
// DSpark requires CUDA; EAGLE binds a head that accepts nothing on 0813 -> target-only.
match: { hw: "mi355x", variant: "pro-official", quant: "fp4", strategy: "low-latency", nodes: "single" },
verified: false,
env: ["SGLANG_USE_ROCM700A=0", "SGLANG_HACK_FLASHMLA_BACKEND=unified_kv_triton", "AITER_BF16_FP8_MOE_BOUND=0", "SGLANG_OPT_USE_AITER_BATCHED_GEMM=true"],
env: ["SGLANG_USE_ROCM700A=0", "TORCH_BLAS_PREFER_HIPBLASLT=1", "SGLANG_HACK_FLASHMLA_BACKEND=unified_kv_triton", "AITER_BF16_FP8_MOE_BOUND=0", "SGLANG_OPT_USE_AITER_BATCHED_GEMM=true"],
flags: [
"--trust-remote-code",
"--model-path {{MODEL_NAME}}",
@@ -1654,7 +1654,7 @@ sgl-eval run aime25 \\
"--swa-full-tokens-ratio 0.15",
"--enforce-shared-experts-fusion",
"--kv-cache-dtype fp8_e4m3",
"--chunked-prefill-size 8192",
"--chunked-prefill-size 16384",
"--host {{HOST_IP}}",
"--port {{PORT}}",
],
@@ -1663,7 +1663,7 @@ sgl-eval run aime25 \\
// DSpark requires CUDA; EAGLE binds a head that accepts nothing on 0813 -> target-only.
match: { hw: "mi355x", variant: "pro-official", quant: "fp4", strategy: "balanced", nodes: "single" },
verified: false,
env: ["SGLANG_USE_ROCM700A=0", "SGLANG_SHARED_EXPERT_TP1=1", "SGLANG_DP_SHARED_EXPERT_LOCAL=1", "SGLANG_DP_USE_GATHERV=1", "SGLANG_DP_USE_REDUCE_SCATTER=1", "SGLANG_HACK_FLASHMLA_BACKEND=unified_kv_triton", "AITER_BF16_FP8_MOE_BOUND=0", "SGLANG_OPT_USE_AITER_BATCHED_GEMM=true"],
env: ["SGLANG_USE_ROCM700A=0", "TORCH_BLAS_PREFER_HIPBLASLT=1", "SGLANG_SHARED_EXPERT_TP1=1", "SGLANG_DP_SHARED_EXPERT_LOCAL=1", "SGLANG_DP_USE_GATHERV=1", "SGLANG_DP_USE_REDUCE_SCATTER=1", "SGLANG_HACK_FLASHMLA_BACKEND=unified_kv_triton", "AITER_BF16_FP8_MOE_BOUND=0", "SGLANG_OPT_USE_AITER_BATCHED_GEMM=true"],
flags: [
"--trust-remote-code",
"--model-path {{MODEL_NAME}}",
@@ -1686,7 +1686,7 @@ sgl-eval run aime25 \\
// DSpark requires CUDA; EAGLE binds a head that accepts nothing on 0813 -> target-only.
match: { hw: "mi355x", variant: "pro-official", quant: "fp4", strategy: "high-throughput", nodes: "single" },
verified: false,
env: ["SGLANG_USE_ROCM700A=0", "SGLANG_SHARED_EXPERT_TP1=1", "SGLANG_DP_SHARED_EXPERT_LOCAL=1", "SGLANG_DP_USE_GATHERV=1", "SGLANG_DP_USE_REDUCE_SCATTER=1", "SGLANG_HACK_FLASHMLA_BACKEND=unified_kv_triton", "AITER_BF16_FP8_MOE_BOUND=0", "SGLANG_OPT_USE_AITER_BATCHED_GEMM=true"],
env: ["SGLANG_USE_ROCM700A=0", "TORCH_BLAS_PREFER_HIPBLASLT=1", "SGLANG_SHARED_EXPERT_TP1=1", "SGLANG_DP_SHARED_EXPERT_LOCAL=1", "SGLANG_DP_USE_GATHERV=1", "SGLANG_DP_USE_REDUCE_SCATTER=1", "SGLANG_HACK_FLASHMLA_BACKEND=unified_kv_triton", "AITER_BF16_FP8_MOE_BOUND=0", "SGLANG_OPT_USE_AITER_BATCHED_GEMM=true"],
flags: [
"--trust-remote-code",
"--model-path {{MODEL_NAME}}",
@@ -2236,6 +2236,7 @@ sgl-eval run aime25 \\
verified: true,
env: [
"SGLANG_USE_ROCM700A=0",
"TORCH_BLAS_PREFER_HIPBLASLT=1",
"SGLANG_HACK_FLASHMLA_BACKEND=unified_kv_triton",
"AITER_BF16_FP8_MOE_BOUND=0",
],
@@ -2249,7 +2250,7 @@ sgl-eval run aime25 \\
"--swa-full-tokens-ratio 0.1",
"--disable-shared-experts-fusion",
"--kv-cache-dtype fp8_e4m3",
"--chunked-prefill-size 8192",
"--chunked-prefill-size 16384",
"--speculative-algorithm EAGLE",
"--speculative-num-steps 3",
"--speculative-eagle-topk 1",
@@ -2263,6 +2264,7 @@ sgl-eval run aime25 \\
verified: true,
env: [
"SGLANG_USE_ROCM700A=0",
"TORCH_BLAS_PREFER_HIPBLASLT=1",
"SGLANG_DP_USE_GATHERV=1",
"SGLANG_HACK_FLASHMLA_BACKEND=unified_kv_triton",
"AITER_BF16_FP8_MOE_BOUND=0",
@@ -2295,6 +2297,7 @@ sgl-eval run aime25 \\
verified: true,
env: [
"SGLANG_USE_ROCM700A=0",
"TORCH_BLAS_PREFER_HIPBLASLT=1",
"SGLANG_DP_USE_GATHERV=1",
"SGLANG_HACK_FLASHMLA_BACKEND=unified_kv_triton",
"AITER_BF16_FP8_MOE_BOUND=0",
@@ -2329,6 +2332,7 @@ sgl-eval run aime25 \\
verified: false,
env: [
"SGLANG_USE_ROCM700A=0",
"TORCH_BLAS_PREFER_HIPBLASLT=1",
"SGLANG_HACK_FLASHMLA_BACKEND=unified_kv_triton",
"AITER_BF16_FP8_MOE_BOUND=0",
"SGLANG_OPT_USE_AITER_BATCHED_GEMM=true",
@@ -2343,7 +2347,7 @@ sgl-eval run aime25 \\
"--swa-full-tokens-ratio 0.15",
"--enforce-shared-experts-fusion",
"--kv-cache-dtype fp8_e4m3",
"--chunked-prefill-size 8192",
"--chunked-prefill-size 16384",
"--host {{HOST_IP}}",
"--port {{PORT}}",
],
@@ -2353,6 +2357,7 @@ sgl-eval run aime25 \\
verified: true,
env: [
"SGLANG_USE_ROCM700A=0",
"TORCH_BLAS_PREFER_HIPBLASLT=1",
"SGLANG_HACK_FLASHMLA_BACKEND=unified_kv_triton",
"AITER_BF16_FP8_MOE_BOUND=0",
"SGLANG_OPT_USE_AITER_BATCHED_GEMM=true",
@@ -2367,7 +2372,7 @@ sgl-eval run aime25 \\
"--swa-full-tokens-ratio 0.15",
"--enforce-shared-experts-fusion",
"--kv-cache-dtype fp8_e4m3",
"--chunked-prefill-size 8192",
"--chunked-prefill-size 16384",
"--speculative-algorithm EAGLE",
"--speculative-num-steps 3",
"--speculative-eagle-topk 1",
@@ -2381,6 +2386,7 @@ sgl-eval run aime25 \\
verified: false,
env: [
"SGLANG_USE_ROCM700A=0",
"TORCH_BLAS_PREFER_HIPBLASLT=1",
"SGLANG_SHARED_EXPERT_TP1=1",
"SGLANG_DP_SHARED_EXPERT_LOCAL=1",
"SGLANG_DP_USE_GATHERV=1",
@@ -2412,6 +2418,7 @@ sgl-eval run aime25 \\
verified: true,
env: [
"SGLANG_USE_ROCM700A=0",
"TORCH_BLAS_PREFER_HIPBLASLT=1",
"SGLANG_SHARED_EXPERT_TP1=1",
"SGLANG_DP_SHARED_EXPERT_LOCAL=1",
"SGLANG_DP_USE_GATHERV=1",
@@ -2447,6 +2454,7 @@ sgl-eval run aime25 \\
verified: false,
env: [
"SGLANG_USE_ROCM700A=0",
"TORCH_BLAS_PREFER_HIPBLASLT=1",
"SGLANG_SHARED_EXPERT_TP1=1",
"SGLANG_DP_SHARED_EXPERT_LOCAL=1",
"SGLANG_DP_USE_GATHERV=1",
@@ -2478,6 +2486,7 @@ sgl-eval run aime25 \\
verified: true,
env: [
"SGLANG_USE_ROCM700A=0",
"TORCH_BLAS_PREFER_HIPBLASLT=1",
"SGLANG_SHARED_EXPERT_TP1=1",
"SGLANG_DP_SHARED_EXPERT_LOCAL=1",
"SGLANG_DP_USE_GATHERV=1",
@@ -2515,6 +2524,7 @@ sgl-eval run aime25 \\
verified: true,
env: [
"SGLANG_USE_ROCM700A=0",
"TORCH_BLAS_PREFER_HIPBLASLT=1",
"SGLANG_HACK_FLASHMLA_BACKEND=unified_kv_triton",
"AITER_BF16_FP8_MOE_BOUND=0",
"SGLANG_OPT_USE_AITER_BATCHED_GEMM=true",
@@ -2529,7 +2539,7 @@ sgl-eval run aime25 \\
"--swa-full-tokens-ratio 0.15",
"--enforce-shared-experts-fusion",
"--kv-cache-dtype fp8_e4m3",
"--chunked-prefill-size 8192",
"--chunked-prefill-size 16384",
"--speculative-algorithm EAGLE",
"--speculative-num-steps 3",
"--speculative-eagle-topk 1",
@@ -2543,6 +2553,7 @@ sgl-eval run aime25 \\
verified: true,
env: [
"SGLANG_USE_ROCM700A=0",
"TORCH_BLAS_PREFER_HIPBLASLT=1",
"SGLANG_SHARED_EXPERT_TP1=1",
"SGLANG_DP_SHARED_EXPERT_LOCAL=1",
"SGLANG_DP_USE_GATHERV=1",
@@ -2578,6 +2589,7 @@ sgl-eval run aime25 \\
verified: true,
env: [
"SGLANG_USE_ROCM700A=0",
"TORCH_BLAS_PREFER_HIPBLASLT=1",
"SGLANG_SHARED_EXPERT_TP1=1",
"SGLANG_DP_SHARED_EXPERT_LOCAL=1",
"SGLANG_DP_USE_GATHERV=1",
@@ -2615,6 +2627,7 @@ sgl-eval run aime25 \\
verified: true,
env: [
"SGLANG_USE_ROCM700A=0",
"TORCH_BLAS_PREFER_HIPBLASLT=1",
"SGLANG_HACK_FLASHMLA_BACKEND=unified_kv_triton",
"AITER_BF16_FP8_MOE_BOUND=0",
"SGLANG_OPT_USE_AITER_BATCHED_GEMM=true",
@@ -2629,7 +2642,7 @@ sgl-eval run aime25 \\
"--swa-full-tokens-ratio 0.15",
"--enforce-shared-experts-fusion",
"--kv-cache-dtype fp8_e4m3",
"--chunked-prefill-size 8192",
"--chunked-prefill-size 16384",
"--speculative-algorithm EAGLE",
"--speculative-num-steps 3",
"--speculative-eagle-topk 1",
@@ -2643,6 +2656,7 @@ sgl-eval run aime25 \\
verified: true,
env: [
"SGLANG_USE_ROCM700A=0",
"TORCH_BLAS_PREFER_HIPBLASLT=1",
"SGLANG_SHARED_EXPERT_TP1=1",
"SGLANG_DP_SHARED_EXPERT_LOCAL=1",
"SGLANG_DP_USE_GATHERV=1",
@@ -2678,6 +2692,7 @@ sgl-eval run aime25 \\
verified: true,
env: [
"SGLANG_USE_ROCM700A=0",
"TORCH_BLAS_PREFER_HIPBLASLT=1",
"SGLANG_SHARED_EXPERT_TP1=1",
"SGLANG_DP_SHARED_EXPERT_LOCAL=1",
"SGLANG_DP_USE_GATHERV=1",
@@ -2715,6 +2730,7 @@ sgl-eval run aime25 \\
verified: true,
env: [
"SGLANG_USE_ROCM700A=0",
"TORCH_BLAS_PREFER_HIPBLASLT=1",
"SGLANG_HACK_FLASHMLA_BACKEND=unified_kv_triton",
"AITER_BF16_FP8_MOE_BOUND=0",
"SGLANG_OPT_USE_AITER_BATCHED_GEMM=true",
@@ -2729,7 +2745,7 @@ sgl-eval run aime25 \\
"--swa-full-tokens-ratio 0.15",
"--enforce-shared-experts-fusion",
"--kv-cache-dtype fp8_e4m3",
"--chunked-prefill-size 8192",
"--chunked-prefill-size 16384",
"--speculative-algorithm EAGLE",
"--speculative-num-steps 3",
"--speculative-eagle-topk 1",
@@ -2743,6 +2759,7 @@ sgl-eval run aime25 \\
verified: true,
env: [
"SGLANG_USE_ROCM700A=0",
"TORCH_BLAS_PREFER_HIPBLASLT=1",
"SGLANG_SHARED_EXPERT_TP1=1",
"SGLANG_DP_SHARED_EXPERT_LOCAL=1",
"SGLANG_DP_USE_GATHERV=1",
@@ -2778,6 +2795,7 @@ sgl-eval run aime25 \\
verified: true,
env: [
"SGLANG_USE_ROCM700A=0",
"TORCH_BLAS_PREFER_HIPBLASLT=1",
"SGLANG_SHARED_EXPERT_TP1=1",
"SGLANG_DP_SHARED_EXPERT_LOCAL=1",
"SGLANG_DP_USE_GATHERV=1",