CoolFace
Modelpublic

anthonymikinka/Qwen3-0.6B-w_uint4_per_group_asym-awq

sourceHugging Faceupdated 1y agoView on Hugging Face
0likes
Model Card

Modified Microsoft/Olive, Microsoft/Onnxgenruntime-ai, and AMD Quark to accept and configure qwen3 for AMD NPU support. This is still in the works, as one of the layers was left out due to complexity.

RyzenAI 1.6.0 dropped 3 days ago and it supports Qwen3. Maybe this would run on there. I am on RyzenAI 1.5.1, I am encoutering the following issue a layer not being a registered function/op.

Loading model from: Qwen3-0.6B-wuint4pergroupasym-awq-FINAL\\model

onnxruntime.capi.onnxruntimepybind11state.Fail: [ONNXRuntimeError] : 1 : FAIL : Load model from Qwen3-0.6B-wuint4pergroupasym-awq-FINAL\\model\model.onnx failed:Fatal error: com.amd:AMDSimplifiedLayerNormalization(-1) is not a registered function/op