IPAdapter model not found: unified presets and manual loaders use different rules

Symptoms and scope

A downloaded IPAdapter file can exist on disk without satisfying a unified loader's preset matching. Identify the actual failing node before renaming or relocating files.

Error fragments for search; IDs, values and filenames may vary:

IPAdapter model not found.
IPAdapterUnifiedLoader

Source-supported context

The author's README requires the documented names for unified loading, while the legacy model loader exposes an ipadapter_file dropdown. In pinned source, get_ipadapter_file(preset, is_sdxl) reads registered ipadapter filenames and chooses a pattern based on the preset and whether the model is SDXL-family. If there is no matching candidate, the unified loader raises IPAdapter model not found. It searches for and loads CLIP Vision first; a vision discovery failure has its own earlier message. A matched filename does not prove that the file contains compatible weights. Author README Pinned resolver Pinned loaders

Cases to distinguish

1. The running instance has no candidate in its ipadapter file list, perhaps because the model path is missing or a download kept a generic filename.

2. The selected preset and SD1.5/SDXL model family lead to a different filename pattern or an explicit unsupported-combination error.

3. A filename matches, but the file identity, contents, vision encoder or required FaceID LoRA are wrong; loading or application then fails at a later stage.

Diagnostic sequence

This is an editorial procedure based on the cited source, not a diagnosis of your machine.

Step 1. Record the exact node class, exception, selected preset and base-model family. Separate ClipVision model not found., IPAdapter model not found., LoRA model not found. and later load/shape errors.

Step 2. Inspect the running server's registered ipadapter filenames and configured model paths. Compare the installed get_ipadapter_file pattern for that preset and family with the publisher's artifact name and identity; do not rename unrelated weights merely to satisfy a pattern.

Step 3. Check clip_vision and loras as separate categories when the preset needs them. If multiple filenames match, identify which file the installed resolver selects first.

Step 4. In a workflow copy, test one reference image with a publisher-supported adapter, vision encoder and base model. The manual IPAdapterModelLoader permits explicit filename selection but leaves compatible component pairing to the workflow author.

Completion check

The running instance selects and loads the intended file and the minimal paired workflow reaches sampling. A filename match alone is insufficient.

Limitations and cautions

The author marks this repository maintenance-only. The pinned source documents one implementation version; check the installed node's behavior before applying its patterns. Publisher links are not proof that this site downloaded or verified weights.

Original sources

• IPAdapter Plus README — checked 2026-09-26.
• Pinned IPAdapter filename resolver — checked 2026-09-26.
• Pinned unified and manual loaders — checked 2026-09-26.

Source and English-language review: 2026-09-26. No model was downloaded, loaded or run by this site; no GPU test or compatibility guarantee is claimed.

Related troubleshooting and guides

• CLIP Vision missing or incompatible: it is not the prompt's text encoder
• InsightFace installation or initialization fails: separate packages from FaceID assets
• IPAdapter Plus reference: presets, vision encoders and FaceID dependencies