Intel One DNN

[ste-q]

Hi,

I am trying to run the sentence-transformers/paraphrase-multilingual-MiniLM-L12-v2 model in ONNX Runtime Java using Intel's DNNL execution provider. I have exported the model to ONNX and also generated the build with the --use_dnnl flag.

However, when I attempt to infer with the model, I am encountering the following error:

 [W:onnxruntime:, session_state.cc:1166 VerifyEachNodeIsAssignedToAnEp] Some nodes were not assigned to the preferred execution providers which may or may not have an negative impact on performance. e.g. ORT explicitly assigns shape related ops to CPU to improve perf.
 [W:onnxruntime:, session_state.cc:1168 VerifyEachNodeIsAssignedToAnEp] Rerunning with verbose output on a non-minimal build will show node assignments.
 [E:onnxruntime:, sequential_executor.cc:516 ExecuteKernel] Non-zero status code returned while running DNNL_17874086050993188016_0 node. Name:'DnnlExecutionProvider_DNNL_17874086050993188016_0_0' Status Message: /onnxruntime/onnxruntime/core/providers/dnnl/subgraph/dnnl_dequantizelinear.cc:191 void onnxruntime::ort_dnnl::DnnlDequantizeLinear::ValidateDims(onnxruntime::ort_dnnl::DnnlSubgraphPrimitive&, onnxruntime::ort_dnnl::DnnlNode&) x_scale and x_zero_point dimensions does not match

Here is the snippet for loading the model:

env = OrtEnvironment.getEnvironment();
OrtSession.SessionOptions options = new OrtSession.SessionOptions();
options.addDnnl(true);
options.setOptimizationLevel(OrtSession.SessionOptions.OptLevel.ALL_OPT);
session = env.createSession(Paths.get(model_home_path,"model.onnx").toString(), options); 

The same model works fine in ONNX CPU executive provider.
Am I missing anything here? Please assist.