Skip to content

Conversation

@rthadur
Copy link
Contributor

@rthadur rthadur commented Jan 25, 2020

Fixes #2698


This change is Reviewable

@Nithanaroy
Copy link

Wow! So fast. Thank you.
Do we need a similar mapping for other types supported by Python Tensorflow (e.g float64)? https://www.tensorflow.org/api_docs/python/tf/dtypes/DType

@kangyizhang
Copy link
Contributor

@rthadur thanks for the PR. But this will not work because int64 is not supported in TensorFlow.js (same as other unmentioned datatypes). What we need to do is to mention this in documentation. Can you add a PR to list supported datatypes in the jsdoc (https://github.com/tensorflow/tfjs/blob/master/tfjs-node/src/saved_model.ts#L357), and also update the error message saying that users can modify their model in python to convert the datatype.

@Nithanaroy
Copy link

Hi Kangiyi, just for my knowledge why doesn't tfjs support 64 bit computation?
Is it because of no native support in JS (saw from https://developer.mozilla.org/en-US/docs/Mozilla/js-ctypes/js-ctypes_reference/Int64)? Are there plans to in near future?

@rthadur
Copy link
Contributor Author

rthadur commented Jan 27, 2020

@kangyizhang will submit a new PR soon. Thank you

@rthadur
Copy link
Contributor Author

rthadur commented Jan 27, 2020

@kangyizhang I have update this PR , can you please review

@kangyizhang kangyizhang merged commit ac52f1c into master Jan 29, 2020
@rthadur rthadur deleted the rthadur-patch-3 branch January 29, 2020 04:35
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

Projects

None yet

Development

Successfully merging this pull request may close these issues.

Error: Unsupported tensor DataType: DT_INT64

4 participants