Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Models not supported by white box XAI methods #30

Open
goodsong81 opened this issue Jun 25, 2024 · 1 comment
Open

Models not supported by white box XAI methods #30

goodsong81 opened this issue Jun 25, 2024 · 1 comment

Comments

@goodsong81
Copy link
Contributor

goodsong81 commented Jun 25, 2024

Following models are not yet supported in addition to NOT_SUPPORTED_BY_BB_MODELS mentioned in #29 due to various reasons.

SUPPORTED_BUT_FAILED_BY_WB_MODELS = {
    # "convformer": "Cannot find output backbone_node in auto mode, please provide target_layer.",
    "swin": "Only two outputs of the between block Add node supported, but got 1. Try to use black-box.",
    "vit_base_patch16_rpn_224": "Number of normalization outputs > 1",
    "vit_relpos_medium_patch16_rpn_224": "ValueError in openvino_xai/methods/white_box/recipro_cam.py:215",
}

NOT_SUPPORTED_BY_WB_MODELS = {
    **NOT_SUPPORTED_BY_BB_MODELS,
    # Killed on WB
    "beit_large_patch16_512": "Failed to allocate 94652825600 bytes of memory",
    "convmixer_1536_20": "OOM Killed",
    "eva_large_patch14_336": "OOM Killed",
    "eva02_base_patch14_448": "OOM Killed",
    "eva02_large_patch14_448": "OOM Killed",
    "mobilevit_": "Segmentation fault",
    "mobilevit_xxs": "Segmentation fault",
    "mvitv2_base.fb_in1k": "Segmentation fault",
    "mvitv2_large": "OOM Killed",
    "mvitv2_small": "Segmentation fault",
    "mvitv2_tiny": "Segmentation fault",
    "pit_": "Segmentation fault",
    "pvt_": "Segmentation fault",
    "tf_efficientnet_l2.ns_jft_in1k": "OOM Killed",
    "xcit_large": "Failed to allocate 81581875200 bytes of memory",
    "xcit_medium_24_p8_384": "OOM Killed",
    "xcit_small_12_p8_384": "OOM Killed",
    "xcit_small_24_p8_384": "OOM Killed",
    # Not expected to work for now
    # "botnet26t_256": "Only two outputs of the between block Add node supported, but got 1",
    # "caformer": "One (and only one) of the nodes has to be Add type. But got MVN and Multiply.",
    "cait_": "Cannot create an empty Constant. Please provide valid data.",
    "coat_": "Only two outputs of the between block Add node supported, but got 1.",
    # "coatn": "Cannot find output backbone_node in auto mode, please provide target_layer.",
    "convmixer": "Cannot find output backbone_node in auto mode, please provide target_layer.",
    "crossvit": "One (and only one) of the nodes has to be Add type. But got StridedSlice and StridedSlice.",
    # "davit": "Only two outputs of the between block Add node supported, but got 1.",
    # "edgenext": "Only two outputs of the between block Add node supported, but got 1",
    # "efficientformer": "Cannot find output backbone_node in auto mode.",
    # "focalnet": "Cannot find output backbone_node in auto mode, please provide target_layer.",
    # "gcvit": "Cannot find output backbone_node in auto mode, please provide target_layer.",
    "levit_": "Cannot find output backbone_node in auto mode, please provide target_layer.",
    # "maxvit": "Cannot find output backbone_node in auto mode, please provide target_layer.",
    # "maxxvit": "Cannot find output backbone_node in auto mode, please provide target_layer.",
    # "mobilevitv2": "Cannot find output backbone_node in auto mode, please provide target_layer.",
    # "nest_": "Cannot find output backbone_node in auto mode, please provide target_layer.",
    # "poolformer": "Cannot find output backbone_node in auto mode, please provide target_layer.",
    # "sebotnet": "Only two outputs of the between block Add node supported, but got 1.",
    "sequencer2d": "Cannot find output backbone_node in auto mode, please provide target_layer.",
    "tnt_s_patch16_224": "Only two outputs of the between block Add node supported, but got 1.",
    # "tresnet": "Batch shape of the output should be dynamic, but it is static.",
    "twins": "One (and only one) of the nodes has to be Add type. But got ShapeOf and Transpose.",
    # "visformer": "Cannot find output backbone_node in auto mode, please provide target_layer",
}
@goodsong81
Copy link
Contributor Author

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
None yet
Projects
None yet
Development

No branches or pull requests

1 participant