Where communities thrive


  • Join over 1.5M+ people
  • Join over 100K+ communities
  • Free without limits
  • Create your own community
People
Activity
    iuhiyuh
    @iuhiyuh
    Does tensorrt support 3D convolution and 3D pooling
    Anyone knows?
    Chin Huang
    @chinhuang007
    Sorry, guys, Got the AM/PM messed up. Let's reschedule the Converters SIG meeting to Aug 14, 2019 at 9-10 am pacific time, web meeting: https://ibm.webex.com/join/chhuang
    call-in: 1-844-531-0958, passcode 927 169 571.
    Shivansh Mundra
    @Shivanshmundra
    Hey Everyone, I wanted metadata of both tensorflow graphs and converted ONNX graph to visualize difference made during graph tranformation. While native functionalities in both tf and onnx provide some tools to extract pbtxt files, it doesn't contain shapes of every node in that. Can someone help me in this regard?
    Wei-Sheng Chin
    @wschin
    Yaman Umuroğlu
    @maltanar_twitter
    Hi all! I'm considering using ONNX as an IR for one of our tools, and I want to do graph transformations in Python. I know that there's C++ infrastructure for writing graph optimization passes and numerous passes implemented in onnx already, but I was wondering if a pure Python version of this also exists.
    If there's a pure Python backend (not for high-performance inference, for educational/reference purposes) that shows how to walk an ONNX graph to execute it node by node, this would also be super helpful. Thanks in advance!
    Chin Huang
    @chinhuang007
    The next converters SIG meeting is 5-6 pm Pacific Time on Thursday, Sept 5, 2019. web meeting: https://ibm.webex.com/join/chhuang, call-in: 1-844-531-0958, passcode 927 169 571. Please let me know if you want to get a meeting invitation on your calendar. Thanks!
    Chin Huang
    @chinhuang007
    The next converters SIG meeting is 9-10 am pacific time on Thursday, Sept 19, 2019, web meeting: https://ibm.webex.com/join/chhuang, call-in: 1-844-531-0958, passcode 927 169 571.
    Svetlana Levitan
    @sveta-levitan
    Next meeting of ONNX Training working group will be in two weeks. We invited several leaders of SIGs to our next WG meeting where we will review the current proposal for ONNX training representation. Anyone interested is also welcome to attend: October 1, 2019, at 10:30 am US Pacific time, join by Webex: https://ibm.webex.com/join/slevitan
    Prasanth Pulavarthi
    @prasanthpul
    Hey everyone - we are updating the ONNX.ai website and will be adding a section for services that output ONNX models. If you have a service that creates ONNX models on behalf of users that you'd like to be considered for the website, please let me know by Friday 9/20! (an example is the Azure Custom Vision service which allows people to upload images and it will train a classifier model for you to download in ONNX format)
    Svetlana Levitan
    @sveta-levitan
    Is there any update on the next in-person ONNX meeting? Thank you.
    Svetlana Levitan
    @sveta-levitan
    Has anybody looked into detailed comparisons of ONNX, including ONNX-ML, with PMML and PFA for tradiional ML? I would like to start this work, wondering if some info already exists.
    Prasanth Pulavarthi
    @prasanthpul
    The next ONNX workshop will be on November 18 in Shanghai. Invites should be sent out in the coming weeks.
    @sveta-levitan PMML and PFA were examined/compared when originally designing ONNX-ML. would love to hear any analysis you have of how to improve traditional ML support.
    Svetlana Levitan
    @sveta-levitan
    Thank you, Prasanth. Do you know the address in Shanghai?
    Any reason why it is on a Monday rather than the usual Friday?
    Svetlana Levitan
    @sveta-levitan
    Prasanth, when your team was doing the examination/comparison of PMML and PFA, did they produce any document that you could publicly share? It would be extremely helpful. I am thinking of creating a big table trying to map all PMML features to ONNX, but that's a lot of work, so if there already is some start, that would help a lot.
    Prasanth Pulavarthi
    @prasanthpul
    @sveta-levitan an operator mapping table was not created previously.
    The invite will have the address in Shanghai. The venue was available on Monday
    Svetlana Levitan
    @sveta-levitan
    OK, thank you, Prasanth.
    Ryan Loney
    @ryan-nervana

    This is an announcement from the ONNX Foundation Working Group. We have created a poll to collect preferences from the ONNX community on what non-profit foundation we should join.

    Please make your voice heard by voting here: https://forms.gle/z4zEs2MGcQiV6tar9

    The options are Linux Foundation, Apache or Eclipse. Our selection and interview process is shared in the link above.

    Thank you,

    Jim Spohrer (IBM) and Ryan Loney (Intel)
    Co-chairs, ONNX Foundation WG

    Chin Huang
    @chinhuang007
    The next Converters SIG meeting is 10-11 am pacific time on Thursday Oct 3, 2019, web meeting: https://ibm.webex.com/join/chhuang, call-in: 1-844-531-0958, passcode 927 169 571.
    Zhipeng Huang
    @hannibalhuang
    @prasanthpul plz loop me in for the Nov Workshop
    Wei-Sheng Chin
    @wschin

    Dear all,
    With the efforts from the ONNX community, ONNX-1.6 is released and available at pypi site.
    This version includes many new features (new operators, new types, more documentations, and so on) and fixes.
    The release details can be found at https://github.com/onnx/onnx/releases/tag/v1.6.0 and its python packages can be installed by using "pip install onnx==1.6.0".
    If you have any comments, please don't hesitate to share them with the community (for example, opening Github issues, posting to Gitter channels, etc).

    Cheers,
    Wei-Sheng

    Zhipeng Huang
    @hannibalhuang
    Prasanth Pulavarthi
    @prasanthpul
    The next ONNX Community Workshop will be held on November 18 in Shanghai! If you are using ONNX in your services and applications, building software or hardware that supports ONNX, or contributing to ONNX, you should attend! This is a great opportunity to meet with and hear from people working with ONNX from many companies. You’ll also have opportunities to participate in technical breakout sessions. Due to limited space, please submit a proposal for a short talk if you would like to attend.
    Ryan Loney
    @ryan-nervana
    The ONNX Foundation WG put together a whitepaper for members of the ONNX community who would like to better understand the reasons for establishing a non-profit, vendor-neutral legal entity. You can read it here: https://github.com/onnx/working-groups/blob/master/foundation/whitepaper.md - any feedback is welcome. We are still collecting input through our poll and gitter.im/foundation channel.
    Chin Huang
    @chinhuang007
    The next Converters SIG meeting is 10-11 am pacific time on Thursday Oct 24, 2019, web meeting: https://ibm.webex.com/join/chhuang, call-in: 1-844-531-0958, passcode 927 169 571.
    matejBandik
    @matejBandik
    Hi, please is possible add some additional information about inputs to onnx model ?
    description or some kind of custom metadata for each inputs
    Prasanth Pulavarthi
    @prasanthpul
    We identified a bug in 1.6 release that impacts production models. The fix has been merged as part of onnx/onnx#2402 and we plan to do a patch release 1.6.1. FYI arch/infra and operators folks: @houseroad @linkerzhang @ebarsoum @postrational
    Svetlana Levitan
    @sveta-levitan
    Is there a schedule of events for the Shanghai meeting somewhere? Thank you in advance.
    yyfhaku
    @yyfhaku
    ImportError: libprotobuf.so.20: cannot open shared object file: No such file or directory
    when I python -c "import onnx"ImportError: libprotobuf.so.20: cannot open shared object file: No such file or directory
    can you help me?
    Prasanth Pulavarthi
    @prasanthpul
    @sveta-levitan the agenda is shown on the registration page.
    The registration page for 2019 workshop in Shanghai is here: https://www.microsoftevents.com/profile/form/index.cfm?PKformID=0x8257553abcd
    Wso
    @wsobra
    Dear all. I am working in a project where I had to create two new operators. The first is a categorical imputer and the second is an operator that checks if a categorical value is in a list of possible values, otherwise the operator substitutes for a label, such as "OTHERS". What is your opinion? Is there a better way of doing this? Does it make sense to open pull request adding these operators to the project?
    Prasanth Pulavarthi
    @prasanthpul
    @wsobra I suggest raising this in the onnx/operators gitter room or joining the next Operators SIG call. that'll let you discuss it with the folks focusing on operators.
    rr123rr
    @rr123rr
    Hello all. I was wondering if its possible to convert a R decision tree to a ONNX model? If yes, where can I find some documentation for this?
    Emad Barsoum
    @ebarsoum
    @wsobra if the new OP satisfy: https://github.com/onnx/onnx/blob/master/docs/AddNewOp.md, then please submit PR. As Prasanth said, use operator channel for all operator related issues or join our SIG meeting which we announce in the operator channel. @rr123rr in ONNX-ML we have tree and tree ensemble, we verified LightGBM and XGBoost conversion. So theoretically, R decision tree can be converted also. Look at https://github.com/onnx/onnx/blob/master/docs/Operators-ml.md
    Xiao
    @lingxiao1989
    Hi I have a question about model verification. Say I want to test resnet50 from Onnx model zoo with the same model defined in another framework, is there a way to compare their performance on the same dataset?
    amajid sinar
    @nakalbird_twitter
    hi guys
    so the thing is, i have a pytorch model that I converted into onnx model via tracing and scripting. when I try to run from netron, I was able to see the network without problem. However, when I tried to load using onnx.load(onnx_model), I got segmentation fault
    amajid sinar
    @nakalbird_twitter
    what confuses me is that the pytorch export using torch.onnx.export got no error whatsoever but upon reading using onnx.load(onnx_model) things starts to fall apart. thank you guys !!!
    Emad Barsoum
    @ebarsoum
    @nakalbird_twitter which version of PyTorch are you using? Can you open an issue in PyTorch and I will add the right people to it?
    rr123rr
    @rr123rr
    Hi. I have a question about the python onnxruntime library. I asked it on stackoverflow. Thanks for the help. https://stackoverflow.com/questions/59047288/error-making-prediction-with-python-onnxruntime
    Elijah Rippeth
    @erip
    I'm training a translation model w/ pytorch and I'm hoping to serve it using the JVM. Has anyone had success doing this? If so, is your code available for reference? :-)
    Prasanth Pulavarthi
    @prasanthpul
    Steering committee is putting together guidelines for new repos under ONNX. Please take a look at https://github.com/onnx/steering-committee/blob/master/repo-guidelines-draft.md and share your feedback
    Svetlana Levitan
    @sveta-levitan
    All, please help to review the PR for ONNX Training proposal, as it affects many parts of the spec. It is about to be merged to master soon, so please add your comments if you want to see any changes or if you support the proposal. It is here: onnx/onnx#2314