#2324 Strata v0.1.44 Fails to run Quadro P6000, Linux 6.8.0-146-generic; GPU 580.178.04; CUDA: 13.0

Beschreibung

Strata was running in all others lasts commits, but this last version: 0.1.44 is with errors.

## Hardware and software

- GPU and VRAM: Quadro P6000, 24 GiB VRAM
- CPU: Intel(R) Xeon(R) CPU           X5670  @ 2.93GHz (12 cores, 24 threads); installed RAM: 94.4 GiB; storage: not measured
- OS: Linux 6.8.0-146-generic; GPU driver: 580.178.04; CUDA: 13.0
- Strata: project 0.1.44; commit ed4773e, engine 0.1.44
- Background workloads and power limits: not recorded

## Errors in Start
Starting qwen3.8-flash-next-q2_0: it loads about 38 GB into RAM and locks part of it for the GPU.
  While it does, YOUR PC CAN BE SLOW OR STOP RESPONDING FOR 1-3 MINUTES (longer the first time after a
  restart). That is normal: please wait and don't close this window - the browser opens when it is ready.
  Later, closing this window stops the model.
  ----------------------------------------------------------------------------------------------------
  Settings (strata-q2_0.json): --expert-cache auto --prefill auto --spec 4 --max-context 131072 --kv
    int8 --kv-resident 32768 --vision --vram-reserve-mib 700 --control-vector-layer-range 4 --cvec-mode
    project --cvec-dir per-layer --pcie-frac 0.90 --spec-min-p 0.70; server 0.0.0.0:11434, gpu 1
loading the vision encoder ...
loading the model (the first start takes a minute or two) ...
[strata] starting the engine: reading the model's weights ...
[strata] the engine stopped while the GPU runtime was starting (most often right after another engine exited, whose GPU queues were still being released): starting it again in 2 s (try 2 of 4)
[strata] starting the engine: reading the model's weights ...
[strata] the engine stopped while the GPU runtime was starting (most often right after another engine exited, whose GPU queues were still being released): starting it again in 4 s (try 3 of 4)
[strata] starting the engine: reading the model's weights ...
[strata] the engine stopped while the GPU runtime was starting (most often right after another engine exited, whose GPU queues were still being released): starting it again in 8 s (try 4 of 4)
[strata] starting the engine: reading the model's weights ...
Traceback (most recent call last):
  File "/home/user/models-ssd/Strata/serve/server.py", line 6869, in <module>
    sys.exit(main())
             ^^^^^^
  File "/home/user/models-ssd/Strata/serve/server.py", line 6646, in main
    engine = StrataEngine(exe, engine_args(cfg) + (effort_end or []), cwd=cfg.get("cwd"), log=cfg.get("log"),
             ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/home/user/models-ssd/Strata/serve/server.py", line 781, in __init__
    StrataEngine.__init__(self, exe, args, cwd, log, env, lazy, cpus, _attempt + 1)
  File "/home/user/models-ssd/Strata/serve/server.py", line 781, in __init__
    StrataEngine.__init__(self, exe, args, cwd, log, env, lazy, cpus, _attempt + 1)
  File "/home/user/models-ssd/Strata/serve/server.py", line 781, in __init__
    StrataEngine.__init__(self, exe, args, cwd, log, env, lazy, cpus, _attempt + 1)
  File "/home/user/models-ssd/Strata/serve/server.py", line 787, in __init__
    raise EngineDied(why + (f" (see {log})" if log else "") +
EngineDied: the engine exited before it was ready (see /home/user/models-ssd/Strata/strata-q2_0.log)
the engine log's last lines:
  [strata] 2026-10-13 17:54:30 engine started: /home/user/models-ssd/Strata/engine-cuda12/strata --serve --pack /home/user/models-ssd/qflash/Strata-data/packs/q2_0 --native /home/user/models-ssd/qflash/Strata-data/models/Q2_0/Qwen3.8-Flash-Next-GSQ-RCO-Q2_0-00001-of-00002.gguf --ple-gguf /home/user/models-ssd/qflash/Strata-data/models/Q2_0/Qwen3.8-Flash-Next-GSQ-RCO-Q2_0-00002-of-00002.gguf --expert-profile /home/user/models-ssd/Strata/data/expert-profile.bin --expert-cache auto --prefill auto --spec 4 --mtp /home/user/models-ssd/qflash/Strata-data/mtp/rt --max-context 131072 --kv int8 --kv-resident 32768 --vision --vram-reserve-mib 700 --control-vector-scaled /home/user/models-ssd/Strata/data/experimental-speed-projection/Qwen3.8-Flash-Next-experimental-speed-projection.gguf:1.0 --control-vector-layer-range 4 44 --cvec-mode project --cvec-dir per-layer --pcie-frac 0.90 --spec-min-p 0.70

  
## Strata LOGS
Local -> /home/user/models-ssd/Strata/strata-q2_0.log
  
  strata-vision: model files loaded in 6.4 s
strata-vision: warmed up at 1024 image tokens in 1.6 s
[strata] 2026-10-13 17:49:15 engine started: /home/user/models-ssd/Strata/engine-cuda12/strata --serve --pack /home/user/models-ssd/qflash/Strata-data/packs/q2_0 --native /home/user/models-ssd/qflash/Strata-data/models/Q2_0/Qwen3.8-Flash-Next-GSQ-RCO-Q2_0-00001-of-00002.gguf --ple-gguf /home/user/models-ssd/qflash/Strata-data/models/Q2_0/Qwen3.8-Flash-Next-GSQ-RCO-Q2_0-00002-of-00002.gguf --expert-profile /home/user/models-ssd/Strata/data/expert-profile.bin --expert-cache auto --prefill auto --spec 4 --mtp /home/user/models-ssd/qflash/Strata-data/mtp/rt --max-context 131072 --kv int8 --kv-resident 32768 --vision --vram-reserve-mib 700 --control-vector-scaled /home/user/models-ssd/Strata/data/experimental-speed-projection/Qwen3.8-Flash-Next-experimental-speed-projection.gguf:1.0 --control-vector-layer-range 4 44 --cvec-mode project --cvec-dir per-layer --pcie-frac 0.90 --spec-min-p 0.70
[strata] 2026-10-13 17:49:21 engine started: /home/user/models-ssd/Strata/engine-cuda12/strata --serve --pack /home/user/models-ssd/qflash/Strata-data/packs/q2_0 --native /home/user/models-ssd/qflash/Strata-data/models/Q2_0/Qwen3.8-Flash-Next-GSQ-RCO-Q2_0-00001-of-00002.gguf --ple-gguf /home/user/models-ssd/qflash/Strata-data/models/Q2_0/Qwen3.8-Flash-Next-GSQ-RCO-Q2_0-00002-of-00002.gguf --expert-profile /home/user/models-ssd/Strata/data/expert-profile.bin --expert-cache auto --prefill auto --spec 4 --mtp /home/user/models-ssd/qflash/Strata-data/mtp/rt --max-context 131072 --kv int8 --kv-resident 32768 --vision --vram-reserve-mib 700 --control-vector-scaled /home/user/models-ssd/Strata/data/experimental-speed-projection/Qwen3.8-Flash-Next-experimental-speed-projection.gguf:1.0 --control-vector-layer-range 4 44 --cvec-mode project --cvec-dir per-layer --pcie-frac 0.90 --spec-min-p 0.70
[strata] 2026-10-13 17:49:28 engine started: /home/user/models-ssd/Strata/engine-cuda12/strata --serve --pack /home/user/models-ssd/qflash/Strata-data/packs/q2_0 --native /home/user/models-ssd/qflash/Strata-data/models/Q2_0/Qwen3.8-Flash-Next-GSQ-RCO-Q2_0-00001-of-00002.gguf --ple-gguf /home/user/models-ssd/qflash/Strata-data/models/Q2_0/Qwen3.8-Flash-Next-GSQ-RCO-Q2_0-00002-of-00002.gguf --expert-profile /home/user/models-ssd/Strata/data/expert-profile.bin --expert-cache auto --prefill auto --spec 4 --mtp /home/user/models-ssd/qflash/Strata-data/mtp/rt --max-context 131072 --kv int8 --kv-resident 32768 --vision --vram-reserve-mib 700 --control-vector-scaled /home/user/models-ssd/Strata/data/experimental-speed-projection/Qwen3.8-Flash-Next-experimental-speed-projection.gguf:1.0 --control-vector-layer-range 4 44 --cvec-mode project --cvec-dir per-layer --pcie-frac 0.90 --spec-min-p 0.70
[strata] 2026-10-13 17:49:39 engine started: /home/user/models-ssd/Strata/engine-cuda12/strata --serve --pack /home/user/models-ssd/qflash/Strata-data/packs/q2_0 --native /home/user/models-ssd/qflash/Strata-data/models/Q2_0/Qwen3.8-Flash-Next-GSQ-RCO-Q2_0-00001-of-00002.gguf --ple-gguf /home/user/models-ssd/qflash/Strata-data/models/Q2_0/Qwen3.8-Flash-Next-GSQ-RCO-Q2_0-00002-of-00002.gguf --expert-profile /home/user/models-ssd/Strata/data/expert-profile.bin --expert-cache auto --prefill auto --spec 4 --mtp /home/user/models-ssd/qflash/Strata-data/mtp/rt --max-context 131072 --kv int8 --kv-resident 32768 --vision --vram-reserve-mib 700 --control-vector-scaled /home/user/models-ssd/Strata/data/experimental-speed-projection/Qwen3.8-Flash-Next-experimental-speed-projection.gguf:1.0 --control-vector-layer-range 4 44 --cvec-mode project --cvec-dir per-layer --pcie-frac 0.90 --spec-min-p 0.70
load_hparams: Qwen-VL models require at minimum 1024 image tokens to function correctly on grounding tasks
load_hparams: if you encounter problems with accuracy, try adding --image-min-tokens 1024
load_hparams: more info: https://github.com/ggml-org/llama.cpp/issues/16842

strata-vision: model files loaded in 1.5 s
strata-vision: warmed up at 1024 image tokens in 1.5 s
[strata] 2026-10-13 17:49:48 engine started: /home/user/models-ssd/Strata/engine-cuda12/strata --serve --pack /home/user/models-ssd/qflash/Strata-data/packs/q2_0 --native /home/user/models-ssd/qflash/Strata-data/models/Q2_0/Qwen3.8-Flash-Next-GSQ-RCO-Q2_0-00001-of-00002.gguf --ple-gguf /home/user/models-ssd/qflash/Strata-data/models/Q2_0/Qwen3.8-Flash-Next-GSQ-RCO-Q2_0-00002-of-00002.gguf --expert-profile /home/user/models-ssd/Strata/data/expert-profile.bin --expert-cache auto --prefill auto --spec 4 --mtp /home/user/models-ssd/qflash/Strata-data/mtp/rt --max-context 131072 --kv int8 --kv-resident 32768 --vision --vram-reserve-mib 700 --control-vector-scaled /home/user/models-ssd/Strata/data/experimental-speed-projection/Qwen3.8-Flash-Next-experimental-speed-projection.gguf:1.0 --control-vector-layer-range 4 44 --cvec-mode project --cvec-dir per-layer --pcie-frac 0.90 --spec-min-p 0.70
[strata] 2026-10-13 17:49:54 engine started: /home/user/models-ssd/Strata/engine-cuda12/strata --serve --pack /home/user/models-ssd/qflash/Strata-data/packs/q2_0 --native /home/user/models-ssd/qflash/Strata-data/models/Q2_0/Qwen3.8-Flash-Next-GSQ-RCO-Q2_0-00001-of-00002.gguf --ple-gguf /home/user/models-ssd/qflash/Strata-data/models/Q2_0/Qwen3.8-Flash-Next-GSQ-RCO-Q2_0-00002-of-00002.gguf --expert-profile /home/user/models-ssd/Strata/data/expert-profile.bin --expert-cache auto --prefill auto --spec 4 --mtp /home/user/models-ssd/qflash/Strata-data/mtp/rt --max-context 131072 --kv int8 --kv-resident 32768 --vision --vram-reserve-mib 700 --control-vector-scaled /home/user/models-ssd/Strata/data/experimental-speed-projection/Qwen3.8-Flash-Next-experimental-speed-projection.gguf:1.0 --control-vector-layer-range 4 44 --cvec-mode project --cvec-dir per-layer --pcie-frac 0.90 --spec-min-p 0.70
[strata] 2026-10-13 17:50:01 engine started: /home/user/models-ssd/Strata/engine-cuda12/strata --serve --pack /home/user/models-ssd/qflash/Strata-data/packs/q2_0 --native /home/user/models-ssd/qflash/Strata-data/models/Q2_0/Qwen3.8-Flash-Next-GSQ-RCO-Q2_0-00001-of-00002.gguf --ple-gguf /home/user/models-ssd/qflash/Strata-data/models/Q2_0/Qwen3.8-Flash-Next-GSQ-RCO-Q2_0-00002-of-00002.gguf --expert-profile /home/user/models-ssd/Strata/data/expert-profile.bin --expert-cache auto --prefill auto --spec 4 --mtp /home/user/models-ssd/qflash/Strata-data/mtp/rt --max-context 131072 --kv int8 --kv-resident 32768 --vision --vram-reserve-mib 700 --control-vector-scaled /home/user/models-ssd/Strata/data/experimental-speed-projection/Qwen3.8-Flash-Next-experimental-speed-projection.gguf:1.0 --control-vector-layer-range 4 44 --cvec-mode project --cvec-dir per-layer --pcie-frac 0.90 --spec-min-p 0.70
[strata] 2026-10-13 17:50:12 engine started: /home/user/models-ssd/Strata/engine-cuda12/strata --serve --pack /home/user/models-ssd/qflash/Strata-data/packs/q2_0 --native /home/user/models-ssd/qflash/Strata-data/models/Q2_0/Qwen3.8-Flash-Next-GSQ-RCO-Q2_0-00001-of-00002.gguf --ple-gguf /home/user/models-ssd/qflash/Strata-data/models/Q2_0/Qwen3.8-Flash-Next-GSQ-RCO-Q2_0-00002-of-00002.gguf --expert-profile /home/user/models-ssd/Strata/data/expert-profile.bin --expert-cache auto --prefill auto --spec 4 --mtp /home/user/models-ssd/qflash/Strata-data/mtp/rt --max-context 131072 --kv int8 --kv-resident 32768 --vision --vram-reserve-mib 700 --control-vector-scaled /home/user/models-ssd/Strata/data/experimental-speed-projection/Qwen3.8-Flash-Next-experimental-speed-projection.gguf:1.0 --control-vector-layer-range 4 44 --cvec-mode project --cvec-dir per-layer --pcie-frac 0.90 --spec-min-p 0