-
Notifications
You must be signed in to change notification settings - Fork 6.8k
Converting mxnet checkpoint to Tensorflow model #4882
Comments
Is the tool backend relevant? Or it just rely on Keras, which means it uses the Keras output, the graph definition and params. If so, https://github.com/dmlc/keras might do the trick. |
Hi. It just needs the model to be loadable in Keras, and then it can be exported & converted for Keras.js to use (as in Keras.js README). Regarding https://github.com/dmlc/keras , is it a project to use mxnet as a backend for Keras? Thanks. |
Yes, but I haven't participate in this project. @piiswrong |
@zihaolucky Hi it seems Keras.js requires the model to be a Keras model, so that it be exported using save_weights and to_json, as written in https://github.com/transcranial/keras-js . Therefore other models need to be converted to Keras models first. It would be great if such a tool can be offered. |
This issue is closed due to lack of activity in the last 90 days. Feel free to reopen if this is still an active issue. Thanks! |
I am looking to do the same. Did someone find a way out for this? |
@apache/mxnet-committers: This issue has been inactive for the past 90 days. It has no label and needs triage. For general "how-to" questions, our user forum (and Chinese version) is a good place to get help. |
Currently the best method to run CNN in browser might be Keras.js (with Tensorflow backend):
https://github.com/transcranial/keras-js
On the other hand, mxnet is probably the fastest in training models.
So I wonder if there's any tools for converting trained mxnet checkpoints (.json & .params) to Tensorflow models. Thanks.
The text was updated successfully, but these errors were encountered: