Where communities thrive


  • Join over 1.5M+ people
  • Join over 100K+ communities
  • Free without limits
  • Create your own community
People
Activity
  • Feb 15 09:42
    hariharansrc closed #1045
  • Feb 15 09:41
    hariharansrc opened #1045
  • Feb 09 17:37
    1412540568 opened #1044
  • Feb 07 16:07
    progr7 opened #1043
  • Dec 31 2020 00:10
    tak1000 opened #1042
  • Nov 17 2020 17:35
    ZW007 commented #1035
  • Nov 15 2020 14:28
    GiovanniCmpaner commented #1041
  • Nov 15 2020 01:38
    GiovanniCmpaner commented #1041
  • Nov 15 2020 01:37
    GiovanniCmpaner commented #1041
  • Nov 09 2020 08:56
    dmilos commented #1038
  • Oct 26 2020 21:18
    tak1000 closed #1041
  • Oct 26 2020 21:18
    tak1000 commented #1041
  • Oct 25 2020 20:07
    tak1000 opened #1041
  • Oct 01 2020 19:25
    teriterance opened #1040
  • Aug 10 2020 08:02
    Sammed98 opened #1039
  • Aug 04 2020 15:43
    Sammed98 commented #892
  • Aug 04 2020 14:54
    Sammed98 commented #1038
  • Aug 04 2020 13:56
    Joe-Mikey commented #1033
  • Aug 04 2020 13:55
    Joe-Mikey commented #1033
  • Aug 04 2020 13:50
    Sammed98 opened #1038
bhack
@bhack
Let's see if @karandesai-96 and @Randl have room for updates of theirs PR
Evgeniy Zheltonozhskiy
@Randl
I'm kinda stuck with the last PR, and I've really underestimated the amount of studies I'm having this semester
However, the PR is basically the last one, there is minor stuff here and there, but it can be used as is
Karan Desai
@kdexd
@bhack I am in for it, let's continue as per our roadmap. @Randl's network tensor integration PR should go first
I had a fresh install of ubuntu recently so will set up tiny-dnn again
does it even require a setup xD
I am almost done with my MS applications, might take max upto 27-28 December, but I can start meanwhile.
@edgarriba just to recall, we agreed on a tiniest possible API of providing support to read hdf into a Tensor object
I am travelling for about 18 hours though.
Edgar Riba
@edgarriba
@Randl I'll take a look at your PR
Evgeniy Zheltonozhskiy
@Randl
@edgarriba Thanks, if you have something tell me, I'll try to find some time this week to work on it too
Dean
@deanhu2
will there ever be support added for gpu through tiny-dnn?
bhack
@bhack
@deanhu2 tiny-dnn/tiny-dnn#866
Dean
@deanhu2
ah thanks for the link, hopefully it happens in future then
Dean
@deanhu2
would it take a particularly long time to convert certain (cnn mostly) layers of tiny_dnn to perform on the gpu? or is the system tied together in another way where a lot of refactoring further down would have to take place? I want to improve the performance and i like tiny dnn, it fits well with continuous integration setups
Manvendra Singh
@manu-chroma
Hi, I'm looking forward to contribute to tiny-dnn. I'm not very experienced with coding in C++ apart from my coursework, though I have experience in contributing to OS projects and experience in python. Is there a beginners issue I might take a shot at? Thanks.
bhack
@bhack
tiny-dnn/tiny-dnn#921
xiaoerlageid
@xiaoerlaigeid
Hi all!I want to participate gsoc on this project. where should i start ?
Ranjodh Singh
@singhranjodh
Hi! I want to contribute to tiny-dnn. Is there any checklist of the milestones available?
appanacca
@appanacca
how about using SYCL as beckebnd for gpu support?
Also I am interested in contribute to the project, there is anyone that can indicate priorities of the project and/or introduce things to work on ?
Alex Giokas
@alexge233
Hi, is tiny-dnn still maintained? The examples you have do not correspond to release 1.0.0.a3, and the master branch doesn't compile.
han-so1omon
@han-so1omon
is there a room open for reinforcement-learning related development?
han-so1omon
@han-so1omon
or has anyone integrated automatic differentiation?
Dendi Suhubdy
@dendisuhubdy
No, not yet.
Ravshan90
@Ravshan90
hello! does anybody use tinydnn with libdnn?
beru
@beru
libdnn looks really good but it's not header only and uses ViennaCL for OpenCL.
Ravshan90
@Ravshan90
yes, i installed ViennaCL, CUDA, but i have some errors when i build, can you help me?
Ravshan90
@Ravshan90
error LNK2019: unresolved external symbol "public: thiscall greentea::device::device(int,int,enum greentea::Backend)" (??0device@greentea@@QAE@HHW4Backend@1@@Z) referenced in function "public: thiscall std::_Ref_count_obj<class greentea::device>::_Ref_count_obj<class greentea::device><int,int,enum greentea::Backend>(int &&,int &&,enum greentea::Backend &&)" (??$?0HHW4Backend@greentea@@@?$_Ref_count_obj@Vdevice@greentea@@@std@@QAE@QAH0QAH0QAW4Backend@greentea@@@Z)
error LNK2019: unresolved external symbol _cuGetErrorString@8 referenced in function "private: class std::basic_string<char,struct std::char_traits<char>,class std::allocator<char> > __thiscall CLCudaAPI::CLCudaAPIError::GetErrorString(enum cudaError_enum)const " (?GetErrorString@CLCudaAPIError@CLCudaAPI@@ABE?AV?$basic_string@DU?$char_traits@D@std@@V?$allocator@D@2@@std@@W4cudaError_enum@@@Z)
beru
@beru
@Ravshan90 I can't help you now.
jianhui
@fantimond
Is this chat room still active?
Sylvain Corlay
@SylvainCorlay
now it is :)
Stanislav Arnaudov
@palikar
Hello to everyone! Is the library still being actively developed\used?
Adi Shavit
@adishavit
I just came across the note re: possible abandonment on the readme.
Are there any recommended alternatives?
Caleb
@caleb221
hi! does anyone know about support for an ESP-32?
taoliqiang1
@taoliqiang1
tiny_char_rnn bot: say something and I'll try to answer :D
Usage: @tiny_char_rnn query
taoliqiang1
@taoliqiang1
tiny_char_rnn bot: say something and I'll try to answer :D
Usage: @tiny_char_rnn query
Febin P Sunny
@codegit001
Hello, is AVX absolutely necessary for tinyDNN to run ?
Is it okay if I turn off using -mno-avx2 or editing the config file ?
Febin P Sunny
@codegit001
I am trying to get some traces for a research paper, using Gem5, for X86
beru
@beru
It's OK to turn it off. You may check CMakeLists.txt how C preprocessor symbols are defined.
taoliqiang1
@taoliqiang1
hello, I am @codegit001 In CMakeLists.txt, this line: option(USE_AVX "Build tiny-dnn with AVX library support" OFF), set USE_AVX --->OFF. AVX improves program effiency, it can be included,but not absolutely necessary.
Febin P Sunny
@codegit001
Thanks
taoliqiang1
@taoliqiang1
Hello, Beru. I debug char_rnn with Eclipse. I find it has some unusual output. In recurrent layer.h: line 29 , “bool reset_state = true”, that is to say the default value for variable “reset_state” is true, and it doesn’t update in any other place in the program. However, this will result in the following two lines codes not being executed ( in recurrent layer.h:line 189, 190) : "for (size_t b = 0; b < batch_size; b++) data[b] = buffer[b]; // copy state" . So, in forward computation, the computed state “h_next” and “c_next” will not be copied to lstm edge (in_data[1]:h_prev, in_data[2]:c_prev) and it is only be stored in the temporary variable “input_buffer”. When in backward propagation, the state "h_prev" and "c_prev" are always zero (their default value when initialized in layer.h, set_sample_count function ), this leads to the variables “dWh2c”, “dWh2f”, “dWh2i”, “dWh2o”, “dWx2f” , "db2f" (in lstm_cell_op_internal.h) will always be zero, the h matrix will not be updated and it is a constant value. I am doubt if this is a bug or What's wrong with my understanding?
@beru
Febin P Sunny
@codegit001
Hello
I would like to cite your framework in my research
How do I do that ? As you guys have not provided a format for this ! Do you guys have a paper in which this was published so that I can cite that ?