search

openvinotoolkit / openvino_xai

OpenVINO™ Explainable AI (XAI) Toolkit: Visual Explanation for OpenVINO Models
https://openvinotoolkit.github.io/openvino_xai/
Apache License 2.0
29 stars 9 forks source link

Runtime error from ONNX / OpenVINO IR models while conversion or inference for XAI #29

Open goodsong81 opened 5 months ago

goodsong81 commented 5 months ago

Following models are not working from the ONNX / OpenVINO IR conversion or inference stage.

NOT_SUPPORTED_BY_BB_MODELS = {
    # "_nfnet_": "RuntimeError: Exception from src/inference/src/cpp/core.cpp:90: Training mode of Batc# hNormalization is not supported.",
    # "convnext_xxlarge": "RuntimeError: The serialized model is larger than the 2GiB limit imposed by the protobuf library.",
    # "convnextv2_huge": "RuntimeError: The serialized model is larger than the 2GiB limit imposed by the protobuf library.",
    # "deit3_huge": "RuntimeError: The serialized model is larger than the 2GiB limit imposed by the protobuf library.",
    # "dm_nfnet": "openvino._pyopenvino.GeneralFailure: Check 'false' failed at src/frontends/onnx/frontend/src/frontend.cpp:144",
    # "eca_nfnet": "openvino._pyopenvino.GeneralFailure: Check 'false' failed at src/frontends/onnx/frontend/src/frontend.cpp:144",
    # "eva_giant": "RuntimeError: The serialized model is larger than the 2GiB limit imposed by the protobuf library.",
    # "halo": "torch.onnx.errors.SymbolicValueError: Unsupported: ONNX export of operator Unfold, input size not accessible.",
    # "nf_regnet": "RuntimeError: Exception from src/inference/src/cpp/core.cpp:90: Training mode of BatchNormalization is not supported.",
    # "nf_resnet": "RuntimeError: Exception from src/inference/src/cpp/core.cpp:90: Training mode of BatchNormalization is not supported.",
    # "nfnet_l0": "RuntimeError: Exception from src/inference/src/cpp/core.cpp:90: Training mode of BatchNormalization is not supported.",
    # "regnety_1280": "RuntimeError: The serialized model is larger than the 2GiB limit imposed by the protobuf library.",
    # "regnety_2560": "RuntimeError: The serialized model is larger than the 2GiB limit imposed by the protobuf library.",
    "repvit": "urllib.error.HTTPError: HTTP Error 404: Not Found",
    # "resnetv2": "RuntimeError: Exception from src/inference/src/cpp/core.cpp:90: Training mode of BatchNormalization is not supported.",
    "tf_efficientnet_cc": "torch.onnx.errors.SymbolicValueError: Unsupported: ONNX export of convolution for kernel of unknown shape.",
    "vit_base_r50_s16_224.orig_in21k": "RuntimeError: Error(s) in loading state_dict for VisionTransformer",
    "vit_gigantic_patch16_224_ijepa.in22k": "RuntimeError: shape '[1, 13, 13, -1]' is invalid for input of size 274560",
    "vit_huge_patch14_224.orig_in21k": "RuntimeError: Error(s) in loading state_dict for VisionTransformer",
    "vit_large_patch32_224.orig_in21k": "RuntimeError: Error(s) in loading state_dict for VisionTransformer",
    # "vit_large_r50_s32": "RuntimeError: Exception from src/inference/src/cpp/core.cpp:90: Training mode of BatchNormalization is not supported.",
    # "vit_small_r26_s32": "RuntimeError: Exception from src/inference/src/cpp/core.cpp:90: Training mode of BatchNormalization is not supported.",
    # "vit_tiny_r_s16": "RuntimeError: Exception from src/inference/src/cpp/core.cpp:90: Training mode of BatchNormalization is not supported.",
    "volo_": "torch.onnx.errors.UnsupportedOperatorError: Exporting the operator 'aten::col2im' to ONNX opset version 14 is not supported.",
}
goodsong81 commented 3 months ago

Partially resolved by