strata generate: PCIe probe: 28.7 GB/s host->device -> pcie_frac 0.55 (default 0.55)
strata generate: this CPU has no AVX-512: the expert kernels run on AVX-2 (multi-token for the i-quant gate/up rows)
strata generate: native pack: /data/strata-bench/packs/iq3_s experts (largest blob 2.66 MB), token embedding type 23 in mapped host memory (322 MiB)
strata generate: 1466 MiB of weights loaded from /data/strata-bench/packs/iq3_s (302 canonical tensors skipped: served natively)
strata generate: 300 native projection matrices, 2018.88 MiB of weights
strata generate: PLE on, table 320001536 rows of /data/strata-bench/models/ISTA-ed59f920/IQ3_S/Qwen3.8-Flash-Next-GSQ-RCO-IQ3_S-00002-of-00002.gguf
strata mtp: draft layer loaded, 950 MiB of VRAM (experts 675, dense 111), files read in 0.17 s (4643 MiB/s)
strata generate: expert arena: cudaHostRegister PORTABLE ok; MAP_HUGETLB unavailable (no hugetlb pool configured?); using 4 KB pages
strata generate: loaded 46.84 GiB at 5.65 GiB/s
strata generate: profile /data/strata-bench/src/strata-8fc40dde/data/expert-profile.bin: 24576 ranked pairs, built for 24576 slots
strata generate: experimental native Q5_K head, 521472000 bytes
strata generate: expert cache auto: 16.08 GiB free, 700 MiB reserved -> 6210 slots
strata generate: expert cache 8120 slots, 15.40 GiB of VRAM; policy is
strata generate: the GPU computes the experts in the cache; it rounds differently from the CPU,
                 so a reply can differ slightly from a run without the cache (same quality:
                 bench/results/2026-09-27-cache-parity).
                 PROFILE, ranked by routing frequency, no eviction.
strata generate: pre-filled 8120 of 8120 slots from the profile; slot 0 verified
strata generate: R4 hit path ON - resident experts are computed on the GPU
strata generate: 23 expert-pool workers + the host thread
strata generate: session is up (engine 0.1.25)
strata generate: token graph hit path: 8120 resident experts, decided on the device
strata serve: prompt chunk auto: 8192 tokens
strata serve: the prompt path borrows 2349 cache slots (4.43 GiB)
strata verify: window up to 6 tokens, 74.7 MiB of device buffers
strata mtp: draft head over 40525 tokens (81.2 MiB)
strata serve: 233 MiB of VRAM free with everything loaded - LOW: requests may stall; add --vram-reserve-mib 979 to the config's args (or lower --max-context)
strata verify: captured the 6-token window (upload no error, sync no error)
strata verify: captured the 1-token window (upload no error, sync no error)
strata verify: captured the 4-token window (upload no error, sync no error)
strata verify: captured the 3-token window (upload no error, sync no error)
strata verify: captured the 2-token window (upload no error, sync no error)
strata serve: prompt 4038 tokens = 0 reused + 4038 read in 2128 ms (1897.3 tok/s), 256 generated in 3298 ms (77.6 tok/s), drafts accepted 159 of 247, 1 checkpoints
strata serve: decode expert cache hit rate: 87.3% (133644 hits / 153006 lookups)
strata serve: suffix drafts: 3 windows, 8 of 15 drafts accepted
strata serve: prompt 4038 tokens = 0 reused + 4038 read in 1844 ms (2190.2 tok/s), 256 generated in 2810 ms (91.1 tok/s), drafts accepted 154 of 217, 1 checkpoints
strata serve: decode expert cache hit rate: 91.8% (133745 hits / 145639 lookups)
strata serve: suffix drafts: 1 windows, 2 of 3 drafts accepted
strata serve: prompt 4038 tokens = 0 reused + 4038 read in 1838 ms (2196.6 tok/s), 256 generated in 2759 ms (92.8 tok/s), drafts accepted 161 of 222, 1 checkpoints
strata serve: decode expert cache hit rate: 92.1% (133487 hits / 144912 lookups)
strata serve: suffix drafts: 1 windows, 2 of 2 drafts accepted
strata serve: prompt 32038 tokens = 0 reused + 32038 read in 11889 ms (2694.8 tok/s), 256 generated in 2842 ms (90.1 tok/s), drafts accepted 154 of 212, 2 checkpoints
strata serve: decode expert cache hit rate: 92.6% (134001 hits / 144710 lookups)
strata serve: prompt 32038 tokens = 0 reused + 32038 read in 11800 ms (2715.0 tok/s), 256 generated in 2685 ms (95.3 tok/s), drafts accepted 159 of 212, 2 checkpoints
strata serve: decode expert cache hit rate: 93.1% (132900 hits / 142747 lookups)
strata serve: suffix drafts: 1 windows, 1 of 2 drafts accepted
strata serve: prompt 32038 tokens = 0 reused + 32038 read in 11848 ms (2704.0 tok/s), 256 generated in 2771 ms (92.4 tok/s), drafts accepted 155 of 211, 2 checkpoints
strata serve: decode expert cache hit rate: 93.3% (135157 hits / 144898 lookups)
strata serve: suffix drafts: 1 windows, 1 of 2 drafts accepted
strata serve: prompt 128037 tokens = 0 reused + 128037 read in 46733 ms (2739.8 tok/s), 256 generated in 3146 ms (81.4 tok/s), drafts accepted 154 of 223, 6 checkpoints
strata serve: decode expert cache hit rate: 90.6% (134085 hits / 148069 lookups)
strata serve: suffix drafts: 2 windows, 5 of 5 drafts accepted