You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
I'm currently using TinyNN to convert a PyTorch model to TFLite format. Currently this model is able to get converted to ONNX completely fine, but when trying to convert it to TFLite I run into the error:
assert len(filtered_dims) == 1, "Multiple indices for aten::index is not supported"
^^^^^^^^^^^^^^^^^^^^^^^
AssertionError: Multiple indices for aten::index is not supported
This is linked, I'm assuming, to 2 nodes of type aten::index.Tensor in the machine learning model (Currently working with MobileSAMv2) where, in PyTorch format, take three inputs and output a single value. Of the three inputs, two of them are indices. I'm currently unsure whether this is caused by an issue with the TinyNN converter or whether there is something I myself am missing. Is this something TinyNN could support or is there a workaround?
This is the conversion code that's been written so far:
I'm currently using TinyNN to convert a PyTorch model to TFLite format. Currently this model is able to get converted to ONNX completely fine, but when trying to convert it to TFLite I run into the error:
This is linked, I'm assuming, to 2 nodes of type aten::index.Tensor in the machine learning model (Currently working with MobileSAMv2) where, in PyTorch format, take three inputs and output a single value. Of the three inputs, two of them are indices. I'm currently unsure whether this is caused by an issue with the TinyNN converter or whether there is something I myself am missing. Is this something TinyNN could support or is there a workaround?
This is the conversion code that's been written so far:
The text was updated successfully, but these errors were encountered: