LoRAX: No local weights error when loading an adapter from a local directory

TL;DR

For "GenerationError: Request failed during generation: Server error: No local weights...": mount the adapter directory into the container with -v and pass the container-side path like /data/lora\models/unsighing with adapter\source="local".

GenerationError: Request failed during generation: Server error: No local weights...

Steps

1. The No local weights error almost always means the adapter path you passed is not visible inside the LoRAX container. The path in adapter_id is resolved by the server, so mount your adapter directory with -v and use the container-side path such as /data/lora_models/unsighing, with adapter_source="local". Double check the directory actually contains adapter_config.json, since the loader needs the config metadata too. The reporter confirmed this fixed it.

When to use

You are seeing this: Loading a local LoRA adapter failed with GenerationError: Request failed during generation: Server error: No local weights... even though the directory contained adapterconfig.json and adaptermodel.bin. Use this skill when you run into "LoRAX: No local weights error when loading an adapter from a local directory".

When not to use

If your error message or symptom does not match what is described above, this is probably not your fix. Search for your exact error text instead of forcing this one to fit.

Versions

No specific versions are mentioned in the source material, so treat the fix as generally applicable and check the examples against whatever you have installed.

Why this happens

The original report does not dig into a root cause. It documents the symptom and the fix that resolved it.

Asked in the thread (https://vectle.com/threads/thrDXJHFSBpeYPmuQFI9QOPA): Loading a local LoRA adapter failed with GenerationError: Request failed during generation: Server error: No local weights... even though the directory contained adapterconfig.json and adaptermodel.bin. Maintainer tgaddair spotted the real problem: the reporter was running the client locally but the server runs inside Docker, so the adapter path has to be the path inside the container, not the host path. Mount the adapter directory into the container with -v and pass the container-side path like /data/lora_models/unsighing with adapter_source="local". The reporter confirmed it worked after that. The verified answer: The No local weights error almost always means the adapter path you passed is not visible inside the LoRAX container. The path in adapter_id is resolved by the server, so mount your adapter directory with -v and use the container-side path such as /data/lora_models/unsighing, with adapter_source="local". Double check the directory actually contains adapterconfig.json, since the loader needs the config metadata too. The reporter confirmed this fixed it. Source: https://github.com/predibase/lorax/issues/51 Thread: https://vectle.com/threads/thrDXJHF_SBpeYPmuQFI9QOPA