diff --git a/configs/nvidia-master.yaml b/configs/nvidia-master.yaml index 9eeeffd50..9db3aa21d 100644 --- a/configs/nvidia-master.yaml +++ b/configs/nvidia-master.yaml @@ -860,7 +860,7 @@ dsr1-fp8-b300-dynamo-trt: ep: 8 dp-attn: true dsr1-fp4-b200-sglang: - image: lmsysorg/sglang:v0.5.16-cu130 + image: lmsysorg/sglang:v0.5.19-cu130 model: nvidia/DeepSeek-R1-0528-FP4-V2 model-prefix: dsr1 runner: cluster:b200-nscale diff --git a/perf-changelog.yaml b/perf-changelog.yaml index f85ee11f6..4bc0ff86c 100644 --- a/perf-changelog.yaml +++ b/perf-changelog.yaml @@ -7162,3 +7162,10 @@ description: - "Use the pinned SGLang nightly-dev-cu13-20260901-07c8f729 image with FlashInfer 0.6.18, which includes the BF16 TRTLLM MoE allocation fix for small-batch Blackwell execution. Model, TP8, HiCache, MTP settings and concurrency grid are unchanged." pr-link: https://github.com/SemiAnalysisAI/InferenceX/pull/2829 + +- config-keys: + - dsr1-fp4-b200-sglang + description: + - "Update SGLang image from lmsysorg/sglang:v0.5.16-cu130 (build commit sgl-project/sglang@fdebc938f7f4d16fe6b9f55dcd9a767cf0899ea1, CUDA 13.0.1, FlashInfer 0.6.14, sgl-kernel 0.4.5) to the v0.5.19 release image lmsysorg/sglang:v0.5.19-cu130 (digest sha256:d6e7288627be8b02be88e4bba38e73f6d50e2826869f753c13a4c4385ab3eda9, build commit sgl-project/sglang@0bcd822377da7b5718e674eaf9c870d349424dd1, CUDA 13.0.3, FlashInfer 0.6.18, sgl-kernel 0.4.6.post1)." + - "benchmarks/single_node/fixed_seq_len/dsr1_fp4_b200.sh is unchanged: nvidia/DeepSeek-R1-0528-FP4-V2 with modelopt_fp4, trtllm_mla attention, flashinfer_trtllm MoE, fp8_e4m3 KV cache, TP4/EP1 concurrency 1-32 and TP4/EP4 DP-attention concurrency 64-256 on the 8k1k workload." + pr-link: https://github.com/SemiAnalysisAI/InferenceX/pull/2989