It says that this is unsupported, maybe try a different version ?
@fusu1992 When I got this version, I found this problem. Some of the Caffe models fail to convert. It seems that the Caffe version supported by the tool is too low . The solution is to use the model trained by the lower version of Caffe, or convert it to another model first
You mean the acuity tool only supports the low version caffe model? But why need to upgrade the caffe model by upgrade_net _proto_text tools as the documents of khadas. such as the follow:
@fusu1992 I have try to upgrade . But the error same too. So I think the tool version is too low , not the model
I use the
aml_npu_sdk_184.108.40.206. Is there higher version tools to convert models？
@fusu1992 It’s same to . We have faceback this problem . But it will take a long time to respond
@fusu1992 It;s same problem about tool’s version . I have test with so much caffe model . But most of the newer models failed to convert .
Can models from other frameworks be transformed successfully? such as tf onnx pytorch …
@fusu1992 Yes, you can convet caffe model to other paltforms . Then use this conver model . TF is the beaset chooise .
@fusu1992 Can you share the link that which can download you caffe model ?
@Frank Can the model convert tool support the model that has multiple input nodes. I run tensorzone to quantize the model. I got the error as follow: