Coder Social home page Coder Social logo

Comments (10)

kitstar avatar kitstar commented on April 27, 2024

Hi @coolrishi2005,
Thanks for helping test the converter. Tensorflow converter is in very initial state, any question and issue is welcome.

Fix some bugs and seems it works. Tensorflow -> IR command (step 1) is

python -m mmdnn.conversion._script.convertToIR -node add -f tensorflow -d IRModel_3_5 -n /tmp/model.ckpt.meta -w /tmp/model.ckpt

Note: Please specify the output node of tensorflow model, which can help the converter trims useless nodes. like -node add in above command.

For this model, we can get the output node from codes:

code

or from tensorboard

tensorboard

Thanks.

from mmdnn.

coolrishi2005 avatar coolrishi2005 commented on April 27, 2024

Thanks kitstar for the response.
The model has been converted successfully now from tf -> cntk.
The issue was due to those useless nodes only, which got trimmed while generating IR Code from tf.

from mmdnn.

mhajiaghayi avatar mhajiaghayi commented on April 27, 2024

hi there, i'm trying to convert a transformer model that is trained with tf to cntk model. I get the following error.

image

from mmdnn.

mhajiaghayi avatar mhajiaghayi commented on April 27, 2024

it seems it's not recognizing the attention layer. is there any way to get around that?

from mmdnn.

namizzz avatar namizzz commented on April 27, 2024

Hi @mhajiaghayi , could you please provide your model files?

from mmdnn.

mhajiaghayi avatar mhajiaghayi commented on April 27, 2024

hi namizz, how would you like me to send it to you? it's 800meg.

from mmdnn.

namizzz avatar namizzz commented on April 27, 2024

Hi @mhajiaghayi , any free cloud storage is ok.

from mmdnn.

mhajiaghayi avatar mhajiaghayi commented on April 27, 2024

hi @namizzz , you can find the model here https://1drv.ms/f/s!AqysHcfYW1hSgc57psU_Bh8MDND74A and I used the following script to convert it

mmconvert -sf tensorflow -in checkpoints\model-0.meta -iw checkpoints\model-0 --dstNodeName output/fwside1/fwside1/fully_connected/Relu -df cntk -om cntk-model

from mmdnn.

namizzz avatar namizzz commented on April 27, 2024

Hi @mhajiaghayi , MMdnn doesn't support RNN network now. Sorry,thanks!

from mmdnn.

mhajiaghayi avatar mhajiaghayi commented on April 27, 2024

hi @namizzz, would you please look at this model located in https://1drv.ms/f/s!AqysHcfYW1hSgc8KG_BnQRrwT7vCJA? . This model doesn't use any RNN or attention layer.
mmconvert -sf tensorflow -in checkpoints\model-0.meta -iw checkpoints\model-0 --dstNodeName output/fwside1/fwside1/fully_connected/Relu -df cntk -om cntk-model

from mmdnn.

Related Issues (20)

Recommend Projects

  • React photo React

    A declarative, efficient, and flexible JavaScript library for building user interfaces.

  • Vue.js photo Vue.js

    🖖 Vue.js is a progressive, incrementally-adoptable JavaScript framework for building UI on the web.

  • Typescript photo Typescript

    TypeScript is a superset of JavaScript that compiles to clean JavaScript output.

  • TensorFlow photo TensorFlow

    An Open Source Machine Learning Framework for Everyone

  • Django photo Django

    The Web framework for perfectionists with deadlines.

  • D3 photo D3

    Bring data to life with SVG, Canvas and HTML. 📊📈🎉

Recommend Topics

  • javascript

    JavaScript (JS) is a lightweight interpreted programming language with first-class functions.

  • web

    Some thing interesting about web. New door for the world.

  • server

    A server is a program made to process requests and deliver data to clients.

  • Machine learning

    Machine learning is a way of modeling and interpreting data that allows a piece of software to respond intelligently.

  • Game

    Some thing interesting about game, make everyone happy.

Recommend Org

  • Facebook photo Facebook

    We are working to build community through open source technology. NB: members must have two-factor auth.

  • Microsoft photo Microsoft

    Open source projects and samples from Microsoft.

  • Google photo Google

    Google ❤️ Open Source for everyone.

  • D3 photo D3

    Data-Driven Documents codes.