Models won't load after v3.x update

Hi, I've used LM Studio in the past on my W11 machine just fine, but after updating, it refuses to load any model using the CUDA runtime. Be noted that Vulkan and CPU do work, but are much slower.

I've updated the GPU driver to it's latest version using GeForce Experience and even installed the CUDA Toolkit. I ran SFC and DISM, which didn't find anything wrong.

As soon as I try loading ANY ("Llama 3.1" or any other) model, it loads for half a second, and then throws:

(Exit code: 0). Some model operation failed. Try a different model and/or config.

Ocasionally I get "Error -1073740791", but that's all.

GPU:

runtime: CUDA llama.cpp (Windows) v1.2.0
{
  "result": {
    "code": "Success",
    "message": ""
  },
  "gpuInfo": [
    {
      "name": "NVIDIA GeForce RTX 2080",
      "deviceId": 0,
      "totalMemoryCapacityBytes": 8589606912,
      "dedicatedMemoryCapacityBytes": 0,
      "integrationType": "Discrete",
      "detectionPlatform": "CUDA",
      "detectionPlatformVersion": "",
      "otherInfo": {}
    }
  ]
}

EDIT: Copying the gguf model into 'oobabooga/text-generation-webui' loads it correctly, and also works perfectly fine. What gives?

lmstudio-ai / lmstudio-bug-tracker

Models won't load after v3.x update #205