Error when using VLM pipeline with docker

### Bug
When trying to use the VLM pipeline in Docling Serve UI to extract content from a PDF, the job fails with the following error:

ERROR: 2025-09-17 09:51:45,283 - docling_jobkit.orchestrators.local.worker - Worker 0 failed to process job d6fbed88-b675-460e-8488-c49f11bc162c: Unrecognized processing class in /opt/app-root/src/.cache/docling/models. Can't instantiate a processor, a tokenizer, an image processor or a feature extractor for this model. Make sure the repository contains the files of at least one of those processing classes.

### Steps to reproduce

1. Run Docling in Docker with GPU support.
2. Configure Open WebUI to use Docling as the content extraction engine.
3. Enable the VLM pipeline and attempt to process a PDF.
4. Observe the error above in the container logs.

### Docker compose content

services:
docling:
image: quay.io/docling-project/docling-serve-cu128
container_name: docling-serve
restart: unless-stopped
ports:
- "5001:5001"
gpus: all
environment:
DOCLING_SERVE_ENABLE_UI: "1"
DOCLING_SERVE_HOST: "0.0.0.0"
DOCLING_SERVE_PORT: "5001"
DOCLING_SERVE_ENABLE_REMOTE_SERVICES: "true"
UVICORN_WORKERS: "2"
DOCLING_SERVE_ENG_LOC_NUM_WORKERS: "4"
HF_HOME: "/models/hf"
HF_HUB_CACHE: "/models/hf/hub"
volumes:
- /opt/docling/models:/models
- /opt/docling/cache:/root/.cache

### Docling version
2.51.0

### Python version
3.13.5