{ "checkpoint": { "audit_verdict": "pass", "output_indexed_payload_bytes": 21921428072, "output_shard_count": 3, "output_tensor_count": 2194, "source_repository": "Qwen/Qwen3.8-27B", "source_revision": "e13a4f0e35203116364e3b3f3f0c82f6ef1afd3c" }, "evaluation": { "accuracy": 0.9727065959059894, "correct": 1283, "dataset": "gsm8k", "empty_generations": 0, "examples": 1319, "finish_reasons": { "stop": 1319 }, "harness_revision": "6690895609dcbc5df1e7b00dd57c9502b868ec4d", "incorrect": 36, "latency_seconds": 144.55181782902218, "request_errors": 0, "stop_rate": 1.0, "truncations": 0 }, "evidence_sha256": { "conversion_manifest": "c71e938cadabd25b2d6ec6b1bd15afecb618674cd8004cf1ecf04e28badbbf55", "evaluation_log": "63c86c3c2aa44765b1ed0c779fc5f2e26bfe091a4cb6ae0858d5e8d92eb6f90d", "evaluation_metrics": "caa4c373713f7cad8e54c2825694d649fa3d77afcf93e990f047e2098fdf44f6", "evaluation_predictions": "e100169dbff801908c3e952ce14e13c133814c55fc4a25ee8d970ce08380eb3e", "production_tensor_audit": "8a7801e2b46298432a129689879c9e4f8c69444e0b71b4c971470c2747794679", "qualification_criteria": "8e3b57367245354acb040a133e63df8eb1f4aef787e036156b944dfda217352a", "runtime_metrics": "99142766caa9138d864b21c241deebb8a1505d15b702d108033f10cf8aff60ac", "server_log": "bfdbfb5d95aaae1caeb9ed5e454562f6896d4fcc9f4a6f437f719f76527901d1" }, "fp8_kv_cache": { "dtype": "fp8_e4m3", "scales_calibrated": false, "warning": "Using FP8 KV cache but no scaling factors provided. Defaulting to scaling factors of 1.0. This may lead to less accurate results!" }, "gates": { "maximum_empty_generations": 0, "maximum_request_errors": 0, "maximum_truncations": 0, "minimum_accuracy": 0.965, "minimum_mtp_acceptance_length": 1.0, "required_stop_rate": 1.0, "target_oom_kills": 0 }, "mtp": { "final_acceptance_length": 2.775, "final_acceptance_rate": 0.5916666666666667, "smoke_acceptance_length": 3.725, "smoke_exact_answer": "5", "smoke_finish_reason": "stop" }, "oom_kills": { "after_evaluation": 0, "after_load": 0, "after_shutdown": 0, "before_startup": 0 }, "runtime": { "gpu_count": 4, "gpu_model": "NVIDIA GB300", "intentional_shutdown": "SIGTERM", "sglang_revision": "29481685462732237d80d86076d6563e1f658102", "speculative_algorithm_resolved": "EAGLE", "tensor_parallel_size": 4 }, "sampling": { "min_p": 0.0, "presence_penalty": 0.0, "repetition_penalty": 1.0, "temperature": 1.0, "thinking": true, "top_k": 20, "top_p": 0.95 }, "verdict": "pass" }