Issues filed for karpathy/neuraltalk2

View Full Project
Do you use neuraltalk2? Leave a review!

Rate of open issues in the last 60 days

neuraltalk2 open issues (View Closed Issues)
  • about 2 years how to get bleu/meteor/cider in other datasets like flickr30k
  • about 2 years /torch/install/share/lua/5.1/hdf5/ffi.lua:56: ')' expected near '_close' at line 1436
  • about 2 years If I use cuda8.0, is there a compatibility issues?
  • about 2 years When I run eval.lua, I failed
  • about 2 years Http Server
  • about 2 years Could neuraltalk2 caption images by using other language such as Chinese?
  • about 2 years ready-trained network does not recognize the image correctly
  • over 2 years Why the eval CIDEr: 0.673 is less than 0.9
  • over 2 years Has anyone trained on any other datasets on top of MS COCO?
  • over 2 years Who has trained a model as good as the pretrained checkpoint model ?
  • over 2 years THCTensorMathPointwise.cu line=40 error=8 : invalid device function
  • over 2 years unknown Torch class <torch.CudaTensor>
  • over 2 years can't run prepro.py
  • over 2 years Why make clones of LSTM and lookup_table in LanguageModel.lua
  • over 2 years Problem about preprocessing
  • over 2 years Error: unable to locate HDF5 header file at hdf5.h
  • over 2 years Error in CuDNN: CUDNN_STATUS_NOT_INITIALIZED
  • over 2 years Extreme bias towards cell phones, dogs, bananas, doughnuts and signs.
  • over 2 years Insane memory usage - is there a tradeoff?
  • over 2 years Can the program output the location of identified objects in the image?
  • over 2 years gradCheckLM and gradCheck can't pass gradcheck test
  • over 2 years When to update the feature embedding layer ?
  • over 2 years when i execute eval.lua ,It always appears as the following error.
  • over 2 years Possible error while saving language model to checkpoint.
  • over 2 years Error while running "th eval.lua ..."
  • over 2 years File.lua failed to load function from bytecode binary string: not a precompiled chunk
  • over 2 years Is that possible to generate a confidence, like 81.23% for a image caption?
  • over 2 years Load CPU Model Check Point Error
  • over 2 years evalution with error
  • over 2 years COCO_train2014_000000221400.jpg
  • over 2 years Can't evaluate BLEU/METEOR/CIDEr scores while running eval.lua
  • over 2 years th eval.lua questions
  • over 2 years How can I use my features of size 4096
  • over 2 years Torch opencv install error /Users/yuanhuaan/desktop/src/torch-opencv/src/Common.cpp:177:24: error: implicit instantiation of undefined template 'std::__1::array<int, 3>'
  • over 2 years <image.load> missing file name
  • over 2 years BRNN vs RNN
  • over 2 years Curious Code Segment in eval.lua
  • over 2 years Evaluation for Flickr dataset?
  • over 2 years Labeling each frame of a video data
  • over 2 years "True" batch training
  • over 2 years training error
  • over 2 years beam search ranked by ppl instead of probability
  • over 2 years Error when training: "DataLoader.lua:84: split train not found" ?
  • over 2 years More Specific Captions - Train New Dataset?
  • almost 3 years Net doesn't seem to train on my own data
  • almost 3 years Test atan2 is already defined
  • almost 3 years Installing on OSX
  • almost 3 years Did anyone successfully run this on Windows?
  • almost 3 years Missing libjpeg library
  • almost 3 years $th eval.lua -model model -image_folder image -num_images 1 ????
  • almost 3 years Can Neural talk 2 be modified for reading text in an Image i.e. Photo OCR ?
  • almost 3 years No model .t7 file being created while training on own data
  • almost 3 years Live webcam stream choppy/delay
  • almost 3 years Wrong JPEG library
  • almost 3 years Am I the only one getting low CIDEr score with default setting?? I barely got 0.3~0.4 with 277500 iterations and what is CNN finetuning ?
  • almost 3 years Issue when running on Jetson TX1
  • almost 3 years Thank you NeuralTalk2
  • almost 3 years "Camera Dropped Frame"
  • almost 3 years Getting error CUDNN_STATUS_NOT_INITIALIZED
  • almost 3 years train.lua error , cuda runtime error (2) : out of memory
  • almost 3 years val_images_use option error
  • almost 3 years Readme claims the ipython notebook downloads the ms coco dataset, but it does not.
  • almost 3 years Instructions have problems
  • almost 3 years prepro.py crashes on Unicode captions
  • almost 3 years libcudnn.dylib not found
  • about 3 years Error in CuDNN: CUDNN_STATUS_NOT_SUPPORTED
  • about 3 years Distributing training on several hosts
  • about 3 years size mismatch, m1: [1 x 512], m2: [25088 x 4096]
  • about 3 years Training freezes and high use of virtual memory (ubuntu)
  • about 3 years The same as "attempt to concatenate local 'ext' (a nil value)" #11
  • about 3 years Issues with CPU based training
  • about 3 years Dockerfile & image available for those interested
  • about 3 years Car model? Motorcycle or truck? Alpr ?
  • about 3 years Bug on ARM hardware
  • about 3 years Image Sizes: 256 to 224 with augmentation?
  • about 3 years Can it work in another direction (from text to image)?
  • about 3 years Using OpenCL instead of CUDA
  • about 3 years CNN Feature Extraction Automatically Runs on Multiple GPUS?
  • about 3 years Inaccurate captioning of images
  • about 3 years System took control over my computer, and wants world domination
  • about 3 years Checkpoint crash
  • about 3 years Error when running eval.lua ("attempt to index field 'tensorOutput' (a nil value)")
  • about 3 years COCO images and Torch's image reader
  • about 3 years another "out of memory issue" when reading the pretained model
  • almost 2 years strange issue while finetuning
  • almost 2 years Segmentation fault on Jetson TX1 during training
  • almost 2 years GPU checkpoint: cannot load incompatible bytecode
  • almost 2 years when I use the code to evaluate BLEU/METEOR/CIDEr scores,I failed!
  • about 2 years How to make eval.lua read image name instead on invoking its own ID?
  • about 2 years JPEG images
  • about 2 years problem in prepro.py
  • about 2 years cant generate "coco/coco_raw.json" in neuraltalk2
  • about 2 years r
  • about 2 years Error when training started from another model
  • about 2 years : Error: unable to locate HDF5 header file at hdf5.h,when I change my conf.lua

neuraltalk2 closed issues

  • about 2 years how to generate multiple captions?
  • over 2 years Skimage module import error when language_eval is set to 1
  • over 2 years Necessary to clone LSTM for every time step?
  • over 2 years How long does CPU based training take?
  • over 2 years loadcaffe.lua:20: attempt to index a nil value (Train.lua)
  • over 2 years train.lua: misc.utils not found: No LuaRocks module found for misc.utils
  • over 2 years prepro.py: Image file not found
  • over 2 years Unspecified launch failure
  • over 2 years Is Neural Talk2 a potential candidate for Optical Formula Recognition?
  • over 2 years require libcudnn.so.5. while I have libcudnn.so.7
  • over 2 years Can't evaluate BLEU/METEOR/CIDEr scores
  • over 2 years where to get the COCO_train2014_000000152328.jpg
  • over 2 years attempt to index upvalue 'C' (a nil value)
  • over 2 years call_python_caption_eval.sh: Permission denied
  • over 2 years Using batches for Linear modules - mismatch error
  • over 2 years MAC OSX 10.11 - eva.lua does not work
  • over 2 years Problem: Same captions for every images
  • over 2 years [CaptureDelegate doFireTimer:]: unrecognized selector sent to instance 0x7fe0a2cb8820
  • almost 3 years Cannot evaluate on MSCOCO validation set
  • almost 3 years Error with h5py in preprocessing
  • almost 3 years Using another CNN feature extractor
  • almost 3 years How does neuraltalk2 extract regions fron image?
  • almost 3 years Why random initialization of image encoding still work.
  • almost 3 years No output .t7 model from running train.lua. Is this expected?
  • almost 3 years Issue while running eval.lua
  • almost 3 years cannot open eval.lua: No such file or directory
  • almost 3 years Can i run neuraltalk2 MacbookAir without CUDA?...
  • almost 3 years cuda runtime error (2) : out of memory at THCStorage.cu
  • almost 3 years option -start_from in train.lua : what's the expected behaviour
  • about 3 years Out of memory error when finetuning is enabled
  • about 3 years Training on SBU dataset
  • about 3 years Getting error on CPU only mode
  • about 3 years Live camera feed from lua-camera not working
  • about 3 years Error when processing training images - No such file or directory IOError
  • about 3 years RGB order?
  • about 3 years libcudnn.so* not found
  • about 3 years Error while running eval.lua
  • about 3 years File names / file paths to the .json file?
  • about 3 years read error: read 0 blocks instead of 1
  • about 3 years "attempt to concatenate local 'ext' (a nil value)"
  • about 3 years Dependency issues. loadcaffe, protobuf, gcc etc.
  • about 3 years Getting identical sequence all the time...
  • about 3 years Out of memory training with 6G memory GPU
  • about 3 years Error when processing training images
  • about 3 years Adding notes on memory requirements?
  • about 3 years link missing
  • almost 2 years why use randomly embedding of image feature can train the language model