{
  "engine": "GGZ14 vllm-mxfp4 / Radiance",
  "runtime": {
    "gpu_count": 1,
    "name": "AMD Radeon AI PRO R9700",
    "torch": "2.11.0+rocm7.14",
    "hip": "7.14.60850",
    "vllm": "0.27.1",
    "triton": "3.6.0+git7c56a5e4"
  },
  "environment": [
    "ROCR_VISIBLE_DEVICES=<R9700-selector>",
    "HIP_VISIBLE_DEVICES=0",
    "HF_HUB_OFFLINE=1",
    "VLLM_LOGGING_LEVEL=INFO",
    "VLLM_NO_USAGE_STATS=1",
    "VLLM_ROCM_USE_AITER=1",
    "VLLM_ROCM_USE_AITER_UNIFIED_ATTENTION=1",
    "VLLM_ROCM_USE_AITER_MHA=0",
    "VLLM_ROCM_USE_AITER_MLA=0",
    "VLLM_ROCM_USE_AITER_MOE=0",
    "VLLM_ROCM_USE_AITER_LINEAR=0",
    "VLLM_ROCM_USE_AITER_FP8BMM=0",
    "VLLM_ROCM_USE_AITER_FP4BMM=0",
    "VLLM_ROCM_USE_AITER_RMSNORM=0",
    "NCCL_PROTO=Simple",
    "RADIANCE_USE_R4D=1",
    "RADIANCE_USE_R4D_AR=1",
    "RADIANCE_USE_R4D_AR_QUANT=1",
    "RADIANCE_R4D_REPORT=1",
    "RADIANCE_AR_MAX_KB=29696",
    "RADIANCE_AR_QBITS=6",
    "RADIANCE_PRESHUFFLE=1",
    "RADIANCE_FUSE_RMS_QUANT=1",
    "RADIANCE_MXFP4=1",
    "RADIANCE_MXFP4_W4A8=1",
    "RADIANCE_MXFP4_W4A8_MIN_M=0",
    "RADIANCE_FAST_DRAFT=1",
    "RADIANCE_DRAFT_TAU=0.20",
    "RADIANCE_DRAFT_RERANK=80",
    "RADIANCE_DFLASH_SELECTOR_TOPK=",
    "RADIANCE_VERIFY_HEAD=1",
    "RADIANCE_VERIFY_HEAD_MAX_M=32",
    "RADIANCE_MXFP4_DEBUG=0",
    "RADIANCE_MXFP4_PUREQUANT=0",
    "RADIANCE_MXFP4_SYNC=0",
    "RADIANCE_MXFP4_CLONE=0",
    "RADIANCE_MXFP4_CHECKX=0",
    "RADIANCE_MXFP4_PADOUT=0",
    "RADIANCE_MXFP4_TN4_MIN_M=2048",
    "RADIANCE_MXFP4_DECODE_MAX_M=64",
    "RADIANCE_MXFP4_DECODE_NT=1",
    "RADIANCE_MXFP4_A_TILED_MIN_M=513",
    "RADIANCE_MXFP4_WPERM=1",
    "RADIANCE_TP_PAD=0",
    "RADIANCE_TP_PAD_INTERMEDIATE=",
    "RADIANCE_TP_PAD_DRAFTER=1",
    "RADIANCE_TP_PAD_STRICT=1",
    "RADIANCE_GDN_MERGE_INPROJ=1",
    "RADIANCE_FP8_STREAM_TP1=1",
    "RADIANCE_GDN_NORM_QUANT=1",
    "RADIANCE_GDN_STRIDED_GATES=0",
    "RADIANCE_GDN_EMPTY_OUT=0",
    "R4D_ATTN_FP8=3",
    "RADIANCE_AR_OVERLAP=0",
    "RADIANCE_GDN_FUSED_UPDATE=1",
    "RADIANCE_GDN_FUSED_MAX_ITEMS=48",
    "RADIANCE_GDN_TRACE_SIDX=0",
    "RADIANCE_GDN_LAZY=0",
    "RADIANCE_DYNAMIC_WIDTH=1",
    "RADIANCE_DYNW_ALPHA=0.35",
    "RADIANCE_DYNW_MARGIN=2",
    "RADIANCE_DYNW_MIN=2",
    "RADIANCE_DYNW_MIN_BATCH=3",
    "RADIANCE_AR_QNB=96",
    "RADIANCE_AR_QNT=1024",
    "MXFP4_CUMODE=0",
    "RADIANCE_STEP_TRACE=0",
    "RADIANCE_AR_OVERLAP_MIN_M=2048",
    "RADIANCE_AR_OVERLAP_SLICES=4",
    "RADIANCE_MXFP4_EPIFAST=1",
    "RADIANCE_MXFP4_R4D_DECODE_MAX_M=0",
    "RADIANCE_TOPK_TRITON_MIN_ROWS=1",
    "RADIANCE_SKINNY_GEMM=1",
    "RADIANCE_DFLASH_CALIB=",
    "RADIANCE_DFLASH_CALIB_TOKENS=200000",
    "RADIANCE_MXFP4_HOIST_QUANT=1",
    "RADIANCE_MXFP4_TRACED_QUANT=1",
    "RADIANCE_FP8_STREAM=1",
    "RADIANCE_RMS_QUANT_FUSION=1",
    "RADIANCE_MXFP4_SHADOW=",
    "RADIANCE_MXFP4_SANITIZE=0",
    "RADIANCE_NVFP4_MXFP4=0",
    "RADIANCE_NVFP4_EXP=mse",
    "RADIANCE_NVFP4_FP8_LAYERS=mxfp4",
    "RADIANCE_NVFP4_BF16_LAYERS=in_proj_ba",
    "RADIANCE_NVFP4_LMHEAD=bf16",
    "RADIANCE_GDN_PATHS=both",
    "RADIANCE_GDN_NANTRACE=0",
    "RADIANCE_MXFP4_KERNEL_N=",
    "RADIANCE_MXFP4_KERNEL_NK=",
    "RADIANCE_MXFP4_CHECKALL=",
    "RADIANCE_MXFP4_MHIST=0",
    "RADIANCE_MXFP4_DECODE_KS=",
    "RADIANCE_MXFP4_DECODE_BK=",
    "RADIANCE_MXFP4_CHECK_MAX_M=128",
    "RADIANCE_MXFP4_PERBLOCK_NK=",
    "RADIANCE_MXFP4_REFLINEAR=0",
    "VLLM_CACHE_ROOT=/cache/vllm",
    "TORCHINDUCTOR_CACHE_DIR=/cache/inductor",
    "TRITON_CACHE_DIR=/cache/triton",
    "AITER_ROOT_DIR=/cache/aiter",
    "TRITON_CACHE_AUTOTUNING=1",
    "R4D_SO=~/coding-local/sLLM/.local-artifacts/vllm-mxfp4/libr4d/b9e42ab-rx9"
  ],
  "serve_argv": [
    "vllm",
    "serve",
    "/models/target-mtpfp8",
    "--served-model-name",
    "Qwen3.8",
    "Qwen3.6",
    "Qwen3.8-MXFP4",
    "--host",
    "0.0.0.0",
    "--port",
    "18184",
    "--kv-cache-dtype",
    "fp8",
    "--tensor-parallel-size",
    "1",
    "--gpu-memory-utilization",
    "0.98",
    "--max-model-len",
    "40960",
    "--max-num-seqs",
    "3",
    "--max-num-batched-tokens",
    "2560",
    "--attention-backend",
    "R4D",
    "--speculative-config",
    "{\"method\":\"dflash\",\"model\":\"/models/draft\",\"num_speculative_tokens\":7,\"attention_backend\":\"TRITON_ATTN\",\"disable_padded_drafter_batch\":true,\"draft_sample_method\":\"greedy\"}",
    "--no-async-scheduling",
    "--mamba-cache-dtype",
    "bfloat16",
    "--mamba-ssm-cache-dtype",
    "float16",
    "--compilation-config",
    "{\"cudagraph_capture_sizes\":[1,2,4,8,16,24],\"pass_config\":{\"fuse_norm_quant\":true,\"fuse_act_quant\":true}}",
    "--enable-prefix-caching",
    "--mamba-cache-mode",
    "align",
    "--enable-auto-tool-choice",
    "--tool-call-parser",
    "qwen3_coder",
    "--reasoning-parser",
    "qwen3",
    "--override-generation-config",
    "{\"temperature\":0.7,\"top_p\":0.95,\"top_k\":20}",
    "--chat-template",
    "/patches/qwen-fixed-v22.3.jinja",
    "--language-model-only"
  ],
  "bootstrap": "Use the pinned GGZ14 serve-mxfp4.sh; full bootstrap/logs remain local",
  "source_commit": "31b9a94a7f74eeb3f59e66d16b1b27dfafcd0663"
}