You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
{{ message }}
Repository navigation
Replay Phase 2 and Phase 6 T-C5 live 8 GiB RTX 3080 telemetry on geminus #4
Phase 2 hardware discovery and Phase 6 T-C5 serving-budget measurements in Proven-Commons/mcproof PR 3 executed on a local 64 GiB Apple M1 Max (llama-server 0.5.0, $0 cloud) because the geminus 8 GiB RTX 3080 host (ASUS UX582HS) had Tailscale SSH and Linux workbench ports closed (linux_rtx3080_online = false).
This issue tracks replaying live nvidia-smi peak VRAM and tok/s telemetry on geminus under an active Tailscale SSH session.
Analytical Baseline Already Verified
From compile/artifacts/phase2_hardware_checkpoint_report.json and verification/VERIFICATION_RECEIPT.json under siliconcordat/mde issue 4:
Confirm Tailscale SSH connectivity to geminus and verify nvidia-smi + llama-server0.5.0+ (--lora-scaled, --lora-init-without-apply, --ctx-size 8192).
Stage Qwen3.5-4B-Q8_0.gguf (10cc391b...) and Qwen3.5-4B-Q4_K_M.gguf (00fe7986...) alongside the three F16 rank-8 LoRAs in compile/artifacts/composition/ (corpus, weft-eido, formal-verification).
Run compile/src/mcproof_compile/hardware_probe.py and MCPROOF_LIVE_LLAMA_TESTS=1 python3 -m pytest stack/tests/test_stack_suite.py -v.
Record live nvidia-smi peak allocated VRAM (<= 7.2 GiB requirement), 8-slot 8192-token KV cache residency, and single-slot / 8-slot tok/s overhead (<= 15.0% vs pure W0) in stack/results/final/phase6_receipt.json and docs/CONTINUITY.md.
Automated posting by agentic team with human oversight.
Live Phase 6 T-C5 8 GiB RTX 3080 telemetry replay on the ASUS UX582HS workbench (Windows 11 Build 26200, NVIDIA GeForce RTX 3080 Laptop GPU8192 MiB, driver 566.07, CUDA 12.7, llama-server b12549) and 3-OS cross-ISA (arm64 macOS, x86_64 Linux, and AMD64 Windows 11) CPU float64 re-derivation via weft.cross_isa (siliconcordat/weft PR 4, commit 7176902419ec11b9454225934a75736b62dd3b47) are both complete.
1. Live 8 GiB RTX 3080 T-C5 Replay (45528746556e4236dcb4a8d10d6fed5b26673c57)
Both Qwen3.5-4B checkpoints were launched with -c 65536 -np 8 --cache-type-k f16 --cache-type-v f16 --lora-init-without-apply and all three F16 rank-8 LoRAs (corpusd81bab6b..., weft-eido71bedfee..., formal-verification728272c0...), polling nvidia-smi every 100 ms across 5 interleaved W0 vs triple-LoRA trials plus an 8-slot concurrent 7486-token/slot KV residency stress test:
Checkpoint (SHA-256)
Idle WDDM VRAM
Peak nvidia-smi VRAM (<= 7.2 GiB)
8-Slot 8192 KV Residency
1-Slot W0 -> Stack (<= 15.0%)
8-Slot W0 -> Stack (<= 15.0%)
Qwen3.5-4B-Q8_0.gguf (10cc391b...)
0 MiB
6.8525 GiB (7017 MiB) PASS
8 x 8192 (7486/slot live) PASS
64.16 -> 63.91 tok/s (+0.38%) PASS
282.78 -> 284.65 tok/s (-0.66%) PASS
Qwen3.5-4B-Q4_K_M.gguf (00fe7986...)
0 MiB
5.2314 GiB (5357 MiB) PASS
8 x 8192 (7486/slot live) PASS
86.27 -> 85.86 tok/s (+0.47%) PASS
346.55 -> 344.89 tok/s (+0.48%) PASS
Live measurement also isolated two offsetting modeling assumptions in hardware_probe.py: NVIDIA Optimus drives both UX582HS panels on the Intel iGPU so idle Windows WDDM dGPU usage is 0 MiB (rather than the +1.20 GiB allowance), while -np 8 -c 65536 (kv_unified=false) allocates 8 x 8192 F16 KV slots (~2.0 GiB plus 8x recurrent state) rather than 1 x 8192 (0.25 GiB), leaving 0.35 GiB of Q8_0 headroom below the 7.2 GiB gate and returning to 0 MiB after shutdown.
2. Three-OS Cross-ISA (arm64 vs x86_64) BetaEdit Re-Derivation
Re-derivation across arm64 macOS (Apple Accelerate, Python 3.12.15, NumPy 2.5.5), x86_64 Linux (Netlib CBLAS/LAPACK 3.12.0, Python 3.14.8, NumPy 2.5.4), and AMD64 Windows 11 (Python 3.12.10, NumPy 2.5.5) pinned the legacy Weft cross-ISA drift boundary and confirmed 0-byte parity under weft.cross_isa:
Adapter / Suite
Legacy Weft (arm64 vs x86_64)
weft.cross_isa (macOS arm64 / Linux x86_64 / Win11 AMD64)
FAIL (8bed53d4... vs 0b01806c..., 9898c603... vs 17995dd0...)
PASS (0-byte match: e0dbf794... / 12970dcc...)
PCG64(20261119) 3-OS portable suite
Diverges across BLAS/LAPACK backends
PASS (0-byte match across all 3 OS/ISA hosts in < 1 s)
What Is Not Proved Here
T-C5 ran via a standalone stdlib replica of the phase6_composition.pyT-C5 harness (python -m unittest tests.test_stack_suite: 10 passed, 3 lab-only skipped). Running the full sibling-repo T-C1..T-C4pytest suite (mde, exo, warp_modeld, weft) natively on UX582HS awaits WSL2/Linux toolchain bring-up.
Activity
ahrazzle commentedon Nov 19, 2026
Live Phase 6
T-C58 GiB RTX 3080 telemetry replay on theASUS UX582HSworkbench (Windows 11 Build 26200,NVIDIA GeForce RTX 3080 Laptop GPU8192 MiB, driver566.07,CUDA 12.7,llama-server b12549) and 3-OS cross-ISA (arm64macOS,x86_64Linux, andAMD64Windows 11) CPUfloat64re-derivation viaweft.cross_isa(siliconcordat/weft PR 4, commit7176902419ec11b9454225934a75736b62dd3b47) are both complete.1. Live 8 GiB RTX 3080
T-C5Replay (45528746556e4236dcb4a8d10d6fed5b26673c57)Both
Qwen3.5-4Bcheckpoints were launched with-c 65536 -np 8 --cache-type-k f16 --cache-type-v f16 --lora-init-without-applyand all three F16 rank-8 LoRAs (corpusd81bab6b...,weft-eido71bedfee...,formal-verification728272c0...), pollingnvidia-smievery100 msacross 5 interleavedW0vs triple-LoRA trials plus an 8-slot concurrent7486-token/slot KV residency stress test:SHA-256)nvidia-smiVRAM (<= 7.2 GiB)8192KV ResidencyW0-> Stack (<= 15.0%)W0-> Stack (<= 15.0%)Qwen3.5-4B-Q8_0.gguf(10cc391b...)0 MiB6.8525 GiB(7017 MiB) PASS8 x 8192(7486/slot live) PASS64.16 -> 63.91 tok/s(+0.38%) PASS282.78 -> 284.65 tok/s(-0.66%) PASSQwen3.5-4B-Q4_K_M.gguf(00fe7986...)0 MiB5.2314 GiB(5357 MiB) PASS8 x 8192(7486/slot live) PASS86.27 -> 85.86 tok/s(+0.47%) PASS346.55 -> 344.89 tok/s(+0.48%) PASSLive measurement also isolated two offsetting modeling assumptions in
hardware_probe.py: NVIDIA Optimus drives bothUX582HSpanels on the Intel iGPU so idle Windows WDDM dGPU usage is0 MiB(rather than the+1.20 GiBallowance), while-np 8 -c 65536(kv_unified=false) allocates8 x 8192F16 KV slots (~2.0 GiBplus8xrecurrent state) rather than1 x 8192(0.25 GiB), leaving0.35 GiBofQ8_0headroom below the7.2 GiBgate and returning to0 MiBafter shutdown.2. Three-OS Cross-ISA (
arm64vsx86_64)BetaEditRe-DerivationRe-derivation across
arm64macOS (Apple Accelerate,Python 3.12.15,NumPy 2.5.5),x86_64Linux (Netlib CBLAS/LAPACK3.12.0,Python 3.14.8,NumPy 2.5.4), andAMD64Windows 11 (Python 3.12.10,NumPy 2.5.5) pinned the legacyWeftcross-ISA drift boundary and confirmed0-byte parity underweft.cross_isa:Weft(arm64vsx86_64)weft.cross_isa(macOS arm64/Linux x86_64/Win11 AMD64)formal-verificationP2-B(f16/int4)f167ffa8780..., cert56820652...)0-byte match:f16e2caa16f..., certcb571814...)formal-verificationP2-B(float64/f32)float64diff5.816e-13,5byte diffs inf32)0-byte match:f32dd04281f..., cert498bbd38...)weft-eido(eido_soft/eido_hardf16)8bed53d4...vs0b01806c...,9898c603...vs17995dd0...)0-byte match:e0dbf794.../12970dcc...)PCG64(20261119)3-OS portable suite0-byte match across all 3 OS/ISA hosts in< 1 s)What Is Not Proved Here
T-C5ran via a standalone stdlib replica of thephase6_composition.pyT-C5harness (python -m unittest tests.test_stack_suite:10passed,3lab-only skipped). Running the full sibling-repoT-C1..T-C4pytestsuite (mde,exo,warp_modeld,weft) natively onUX582HSawaits WSL2/Linux toolchain bring-up.