- Mark as New
- Bookmark
- Subscribe
- Mute
- Subscribe to RSS Feed
- Permalink
- Report Inappropriate Content
I use mo.py to convert BERT model, but error occur.
I use tf.train.Example().SerializeToString() as input Node, I think the error is about this, but I can not fix it.
Link Copied
- Mark as New
- Bookmark
- Subscribe
- Mute
- Subscribe to RSS Feed
- Permalink
- Report Inappropriate Content
Dear lin, xiaobo,
OpenVino is primarily a Computer Vision (CNN) tool. However some aspects of RNN such as LSTM and GRU (used by NLP applications) are definitely supported.
BERT apparently uses the Transformer which is LSTM-based Encoder/Decoder topology which avoids RNN. I will tell you that BERT is not one of OpenVino's tested and validated models.
The train.* error you are getting though makes me think that you are not freezing the BERT Tensorflow model before feeding it to OpenVino Model Optimizer. If the model is properly frozen, then you will not see errors related to training. Please review Freezing Tensorflow Models for more information about freezing a Tensorflow model.
Hope it helps,
Thanks,
Shubha
- Subscribe to RSS Feed
- Mark Topic as New
- Mark Topic as Read
- Float this Topic for Current User
- Bookmark
- Subscribe
- Printer Friendly Page