{ "schema_version": 1, "builder_plan": { "scenario": "Profiling shows attention computation is a bottleneck in your self-hosted service.", "change": "Check the supported GPU, tensor shapes, and precision, then replace only the relevant kernel.", "baseline": "The existing kernel inside the same complete serving pipeline.", "steps": [ "Locate the paper's implementation and check its license, model support, data needs, and hardware requirements. Stop here if you cannot obtain the required artifacts.", "Save a small set of representative product tasks and expected outcomes. Keep evaluation tasks out of any training or tuning, and choose quality and cost limits before running.", "Run the baseline and the proposed change on identical inputs. Keep other settings fixed and repeat paired runs to expose variation.", "Save per-task outputs, failures, timings, and all usage costs. Review regressions before expanding the test." ], "measure": "Numerical differences, end-to-end latency, throughput, and memory at realistic shapes.", "decision": "Keep the change only if it meets your preselected quality and latency limits and improves the metric you care about after setup and operating costs. Otherwise retain the baseline. This is an exploration plan, not a validated recommendation." }, "builder_review": null, "arxiv_id": "2609.15627", "title": "DeepSeek-V4-Flash on AMD gfx90a: Correctness Recovery and Inference Performance Engineering", "summary": "The work replaces a default SGLang execution path with a hardware-specific stack combining FP4 routed-expert computation, FP8 dense projections, CDNA2 MFMA and dot-product kernels, HIP graphs, and a load-time repair of a numerically incorrect routed-expert W2 weight layout, validated by fixed-token and hash-based correctness checks. The cost is substantial porting and validation effort tied to AMD gfx90a hardware, plus the ongoing burden of maintaining packed FP4 formats, topology-aware kernel tuning, and custom correctness harnesses. It can fail silently through weight-layout or execution-format mismatches that produce fast but numerically wrong output, and through low-M utilization and per-layer synchronization overhead that erode the expected gains.", "rationale": "The techniques require four MI250 GCDs and low-level CDNA2 kernel engineering that the reader's 24 GB GPU and small team cannot replicate, though the load-time layout verification practice is worth noting as a general principle.", "technique": "FP4 MoE inference enablement on AMD CDNA2", "family": "kernels_e_atencao", "recommendation": "observar", "published": "2026-09-14", "as_of": "2026-10-18", "editorial_status": "indexed", "source_url": "https://arxiv.org/abs/2609.15627", "independent_implementations": 0, "total_implementations": 0, "stars": 0, "citations": 0, "claims": [ { "claim_id": "claim-01", "statement": "The work replaces a default SGLang execution path with a hardware-specific stack combining FP4 routed-expert computation, FP8 dense projections, CDNA2 MFMA and dot-product kernels, HIP graphs, and a load-time repair of a numerically incorrect routed-expert W2 weight layout, validated by fixed-token and hash-based correctness checks. The cost is substantial porting and validation effort tied to AMD gfx90a hardware, plus the ongoing burden of maintaining packed FP4 formats, topology-aware kernel tuning, and custom correctness harnesses. It can fail silently through weight-layout or execution-format mismatches that produce fast but numerically wrong output, and through low-M utilization and per-layer synchronization overhead that erode the expected gains.", "basis": "inferred", "result": "", "baseline": "", "conditions": "", "source_url": "https://arxiv.org/abs/2609.15627", "source_page": null, "source_excerpt": "" }, { "claim_id": "claim-02", "statement": "On four MI250 GCDs, TP4/EP1 decode reaches approximately 74.5 tok/s and a 4,604-token prompt achieves approximately 2,234 input tok/s prefill throughput; no baseline comparison factor is reported.", "basis": "inferred", "result": "", "baseline": "", "conditions": "", "source_url": "https://arxiv.org/abs/2609.15627", "source_page": null, "source_excerpt": "" } ], "exposure_map": [ { "dimension": "quality", "basis": "not_evaluated", "finding": "", "interpretation": "", "limitation": "", "next_check": "", "source_url": "", "source_page": null, "source_excerpt": "" }, { "dimension": "compute", "basis": "not_evaluated", "finding": "", "interpretation": "", "limitation": "", "next_check": "", "source_url": "", "source_page": null, "source_excerpt": "" }, { "dimension": "latency", "basis": "not_evaluated", "finding": "", "interpretation": "", "limitation": "", "next_check": "", "source_url": "", "source_page": null, "source_excerpt": "" }, { "dimension": "operations", "basis": "not_evaluated", "finding": "", "interpretation": "", "limitation": "", "next_check": "", "source_url": "", "source_page": null, "source_excerpt": "" }, { "dimension": "compatibility", "basis": "not_evaluated", "finding": "", "interpretation": "", "limitation": "", "next_check": "", "source_url": "", "source_page": null, "source_excerpt": "" }, { "dimension": "security", "basis": "not_evaluated", "finding": "", "interpretation": "", "limitation": "", "next_check": "", "source_url": "", "source_page": null, "source_excerpt": "" }, { "dimension": "data_and_training", "basis": "not_evaluated", "finding": "", "interpretation": "", "limitation": "", "next_check": "", "source_url": "", "source_page": null, "source_excerpt": "" }, { "dimension": "reproducibility", "basis": "not_evaluated", "finding": "", "interpretation": "", "limitation": "", "next_check": "", "source_url": "", "source_page": null, "source_excerpt": "" } ], "risks": [], "minimum_test": [], "open_questions": [], "independent_tests": [], "lab_reproductions": [], "report_available": false, "validation_tier": "unknown", "evidence_tier": "unknown", "infrastructure_basis": "unknown", "equations": [], "equations_status": "not_formula", "core_kind": "system", "equations_fetched_at": "2026-10-08", "equations_source_url": "", "equations_source_sha256": "", "exposures_reviewed_at": "", "exposures_source_sha256": "" }