[Bug Report] Severe VRAM Allocation Instability in PyTorch after llama-cpp-python is Imported
Maintainer antworten meist innerhalb von 1 Tag
Dieses Issue hat noch niemand übernommen.
Bewertung
- Schwierigkeit
- 5/5
- Geschätzter Aufwand
- Über eine Woche
- Anfängerfreundlichkeit
- 30/100
- Issue-Typ
- Bug
- Klarheit
- Größtenteils klar
- Aktivitätsstatus
- Veraltet
- Bereich
- ai-infra-agents
Rechercherichtung
Das Issue beschreibt einen Low-Level-Konflikt zwischen CUDA-Kontexten von llama-cpp-python und PyTorch. Beginne damit, den CUDA-Initialisierungscode in den Bindings von llama-cpp-python zu untersuchen, insbesondere zur Importzeit. Suche nach der Erstellung von CUDA-Kontexten oder Einstellungen für den Speicher-Allocator. Reproduziere das Problem in einer minimalen Umgebung, die llama-cpp-python importiert und anschließend ein großes PyTorch-Modell lädt, und überwache dabei die Muster der VRAM-Belegung. „Done“ wäre, den spezifischen Aufruf oder die Konfiguration zu identifizieren, der bzw. die den Konflikt verursacht, und einen Fix vorzuschlagen.
Vom Indexierungsmodell aus dem Issue-Text verfasst.
Beschreibung
Prerequisites
- I am running the latest code.
- I carefully followed the README.md.
- I searched using keywords relevant to my issue to make sure that I am creating a new issue that is not already open (or closed).
- I reviewed the Discussions, and have a new bug to share.
Expected Behavior
When loading a large PyTorch-based model (e.g., an SDXL checkpoint >10GB) within a host application (ComfyUI), I expect the GPU's dedicated memory (VRAM) usage to increase smoothly and stabilize at the required level for the model. This process should be fast and efficient.
Current Behavior
The VRAM allocation is extremely erratic and inefficient. The memory usage graph shows a severe sawtooth/spiky pattern—it spikes up, drops slightly, spikes again, and repeats this cycle until the model is finally loaded. This process is significantly slower than normal and suggests a fundamental conflict in memory management at the CUDA level.
This behavior only occurs when a component that imports llama-cpp-python is loaded by the host application at startup. The issue manifests before any llama-cpp-python functions are actively called; the mere act of importing the library is enough to trigger this conflict with subsequent PyTorch operations.
Environment and Context
-
Hardware:
- GPU: NVIDIA RTX PRO 6000 Blackwell Workstation Edition (94.5 GB VRAM)
-
Operating System:
- Windows 11
-
SDK and Software Versions:
- Python Version:
3.12.11 - Compiler: MSVC / Visual Studio
2022used to compile for CUDA - CUDA Toolkit: 12.8 (locally compiled)
- NVIDIA Driver Version:
580.97 - Host Application: ComfyUI (Release
2025-08-27or newer)
- Python Version:
-
Key Python Packages:
- llama-cpp-python:
0.3.16 - torch:
2.8.0+cu128 - diskcache: 5.6.3
- llama-cpp-python:
Failure Information (for bugs)
This report details a deep-seated conflict between llama-cpp-python and torch regarding CUDA context and/or VRAM management.
Investigation Summary & Key Findings
We conducted an extensive, multi-day investigation to isolate this issue, and the results were conclusive:
-
The problem is triggered by the
importofllama-cpp-python. Any host application that loads this library at startup will cause subsequent, unrelatedtorchVRAM allocations to become unstable. Disabling the components that importllama-cpp-pythoncompletely resolves the issue. -
The
diskcachedependency is a Red Herring. We initially suspected thediskcachelibrary was the cause. To test this, we performed a definitive experiment: we hid the realdiskcachelibrary and replaced it with a completely inert, non-functional fake Python package that only contained an emptyCacheclass.- Result: The VRAM instability problem persisted even with the fake library.
- Log Confirmation: The application log confirmed our fake library was being used:
--- 警告:正在使用一個偽造的 (FAKE) diskcache 函式庫 --- Set up system temp directory: C:\Users\Win\AppData\Local\Temp\latentsync_16880d62 - Conclusion: This proves that
diskcache's code is not the active cause. The problem isllama-cpp-pythonitself. The initial correlation was due todiskcachebeing a mandatory dependency; uninstalling it simply preventedllama-cpp-pythonfrom being imported successfully.
Final Hypothesis
The evidence overwhelmingly suggests a low-level conflict between how llama-cpp-python (especially a version compiled against a modern CUDA 12.8 toolkit) initializes its CUDA context and how PyTorch expects to manage the VRAM. The initialization of llama-cpp-python appears to "pollute" or "fragment" the VRAM state, making subsequent large-scale allocations by PyTorch highly inefficient.
Steps to Reproduce
- Set up a ComfyUI environment with PyTorch and the system components listed above.
- Compile and install
llama-cpp-pythonwith CUDA 12.8 support. - Install any ComfyUI custom node that imports
llama-cpp-pythonat startup (e.g.,ComfyUI-JoyCaptionorComfyUI-Searge-LLM-Node). - Launch ComfyUI. The server will start, and
llama-cpp-pythonwill be imported into the process. - Load a standard workflow designed to load a large PyTorch-based checkpoint model (e.g., an SDXL checkpoint, >10GB).
- Observe the "Dedicated GPU Memory" graph in Windows Task Manager. You will see the erratic, spiky allocation pattern instead of a smooth ramp-up.
Note on llama.cpp comparison: This issue is not about the performance or functionality of llama-cpp-python's inference, but rather a resource conflict it creates with another major library (torch). Therefore, comparing behavior against llama.cpp's ./main executable is not applicable, as the issue requires both runtimes to be active in the same Python process.
- Vorherrschende Sprache
- Python
- Sterne
- 10.7k
- Forks
- 1.5k
- Ø Merge
- 1 Std. 40 Min.
- Gemergte PRs (30 T.)
- 5
Entwicklungsumgebung
- Kein Dockerfile und keine Docker-Compose-Datei
- Keine Pull-Request-Vorlage
- Beitragsleitfaden lesen
Erste Schritte
- Lesen Sie das ganze Issue und danach den Beitragsleitfaden des Projekts.
- Schreiben Sie ins Issue, dass Sie es übernehmen — das erspart doppelte Arbeit.
- Forken Sie das Repository und arbeiten Sie in einem Branch.
- Öffnen Sie einen Pull Request, der die Issue-Nummer nennt.
Mehr aus abetlen/llama-cpp-python
-
Schwierigkeit 2/5 1-3 Stunden Anfängerfreundlichkeit 78/100
abetlen/llama-cpp-python#2424 ·
Maintainer antworten meist innerhalb von 1 Tag
-
Seven llama_sampler_init_* bindings admit keyword arguments that the ctypes function object silently dropsEvtl. vergeben @Belal0066 hat das vor 59 Tagen übernommen. Offen
Schwierigkeit 2/5 1-3 Stunden Anfängerfreundlichkeit 88/100
abetlen/llama-cpp-python#2371 ·
Maintainer antworten meist innerhalb von 1 Tag
-
uv add llama-cpp-python wheels fails for versions above 0.3.30Evtl. vergeben Ein verknüpfter Pull Request ist offen oder bereits gemergt. Offen
Schwierigkeit 2/5 1-3 Stunden Anfängerfreundlichkeit 65/100
abetlen/llama-cpp-python#2352 · 1 Kommentar · 3 Reaktionen ·
Maintainer antworten meist innerhalb von 1 Tag
-
Schwierigkeit 2/5 1-3 Stunden Anfängerfreundlichkeit 75/100
abetlen/llama-cpp-python#2314 ·
Maintainer antworten meist innerhalb von 1 Tag
-
Llama.embed() calls LlamaBatch.add_sequence with old 3-arg signature; missing logits_arrayEvtl. wieder frei @lxcxjxhx hat das vor 136 Tagen übernommen, und es ist kein Pull Request offen. Offen
Schwierigkeit 2/5 1-3 Stunden Anfängerfreundlichkeit 65/100
abetlen/llama-cpp-python#2211 · 2 Kommentare ·
Maintainer antworten meist innerhalb von 1 Tag
Alle Issues in abetlen/llama-cpp-python
Ähnliche Issues
-
Bug IronPython Python 3
Schwierigkeit 2/5 1-3 Stunden Anfängerfreundlichkeit 72/100
pyrevitlabs/pyRevit#4184 ·
Maintainer antworten meist innerhalb von 1 Tag
-
Schwierigkeit 1/5 Unter einer Stunde Anfängerfreundlichkeit 74/100
jqueryscript/awesome-claude-code#1009 ·
-
tests
Schwierigkeit 2/5 1-3 Stunden Anfängerfreundlichkeit 66/100
Maintainer antworten meist innerhalb von 3 Tagen
-
Schwierigkeit 2/5 1-3 Stunden Anfängerfreundlichkeit 72/100
Maintainer antworten meist innerhalb von 1 Tag
-
[Plugin] 群聊记忆 · 人格养成Offen
Schwierigkeit 1/5 1-3 Stunden Anfängerfreundlichkeit 74/100