oobabooga/textgen · error · ServiceUnavailableError

No image model loaded. Load a model via the UI first.

Error message

No image model loaded. Load a model via the UI first.

What it means

The image generation endpoint (/v1/images/generations style generations() in modules/api/images.py) requires an image/diffusion model to already be loaded into shared.image_model. The API only loads LLMs; diffusion models must be selected and loaded through the UI's image generation tab first. Absence raises ServiceUnavailableError (503).

Source

Thrown at modules/api/images.py:24

import io
import json
import time

from PIL.PngImagePlugin import PngInfo

from .errors import ServiceUnavailableError
from modules import shared


def generations(request):
    """
    Generate images using the loaded diffusion model.
    Returns dict with 'created' timestamp and 'data' list of images.
    """
    from modules.ui_image_generation import build_generation_metadata, generate

    if shared.image_model is None:
        raise ServiceUnavailableError("No image model loaded. Load a model via the UI first.")

    width, height = request.get_width_height()

    # Build state dict: GenerationOptions fields + image-specific keys
    state = request.model_dump()
    state.update({
        'image_model_menu': shared.image_model_name,
        'image_prompt': request.prompt,
        'image_neg_prompt': request.negative_prompt,
        'image_width': width,
        'image_height': height,
        'image_steps': request.steps,
        'image_seed': request.image_seed,
        'image_batch_size': request.batch_size,
        'image_batch_count': request.batch_count,
        'image_cfg_scale': request.cfg_scale,
        'image_llm_variations': False,
    })

View on GitHub (pinned to ed888c71f2)

Solutions

1. Open the UI image generation tab, select a diffusion model, and wait for it to load before calling the API.
2. For headless use, load the model once via the UI (or programmatic session) and keep the process alive.
3. Verify readiness by checking the loaded image model via the models/status endpoints before calling generations.
4. Treat 503 from this endpoint as 'warm-up required', not as a retryable transient.

import requests

def image_model_loaded(base_url: str) -> bool:
    # probe: a 503 with 'No image model loaded' means warm-up needed
    r = requests.post(f'{base_url}/v1/images/generations', json={'prompt': 'probe', 'steps': 1}, timeout=30)
    return r.status_code == 200 or 'No image model loaded' not in r.text

try:
    img = client.images.generate(model='x', prompt='a cat')
except openai.APIStatusError as e:
    if e.status_code == 503 and 'No image model loaded' in str(e):
        raise RuntimeError('Load a diffusion model in the UI image-generation tab before calling this API')
    raise

When it happens

Trigger: POST /v1/images/generations (or equivalent) against a headless server started with --api where no diffusion model was loaded via the UI image-generation tab; after a model load failure or UI session reset; before visiting the image tab at all.

Common situations: Headless/API-only deployments assuming image models autoload; calling image generation right after server start; the diffusion model load previously failed leaving shared.image_model None.

Related errors

• Error: Failed to load embedding model: {model}
• Image generation failed or produced no images.
• functions is not supported.
• function_call is not supported.
• messages is required

AI-assisted analysis of oobabooga/textgen@ed888c71f2 (2026-08-15). Data as JSON: /api/errors/9076d38b7a1a6e2a. Report an issue: GitHub.