Named Entity Recognition (LSTM + CRF) - Tensorflow.
Key parameters and specifications
Key parameters and specifications
Parameter
Value
Language
Python
Category
Model Development
Subcategory
Architecture
Stars
1,956
Forks
433
Contributors
8
Created date
2016-07-09
Updated date
2026-10-11
Stars in 1 day
0
Stars in 7 days
-1
Rank
9,194
Previous rank
9,231
Rank change
37
7-day stars rank
16,988
Previous 7-day stars rank
16,983
7-day stars rank change
-5
Total stars rank
6,467
Previous total stars rank
6,461
Total stars rank change
-6
Country
USA
README summary
**Status:** Archive (code is provided as-is, no updates expected) pixel-cnn++ This is a Python3 / Tensorflow implementation of PixelCNN++, as described in the following paper: **PixelCNN++: A PixelCNN Implementation with Discretized Logistic Mixture Likelihood and Other Modifications**, by Tim Salimans, Andrej Karpathy, Xi Chen, Diederik P. Our work builds on PixelCNNs that were originally proposed in van der Oord et al. PixelCNNs are a class of powerful generative models with tractable likelihood that are also easy to sample from
. The core convolutional neural network computes a probability distribution over a value of one pixel conditioned on the values of pixels to the left and above it. Below are example samples from a model trained on CIFAR-10 that achieves **2. 03 of the PixelCNN in van der Oord et al. ): Samples from the model (**left**) and samples from a model that is conditioned on the CIFAR-10 class labels (**right**): This code supports multi-GPU training of our improved
PixelCNN on CIFAR-10 and Small ImageNet, but is easy to adapt for additional datasets. Training on a machine with 8 Maxwell TITAN X GPUs achieves 3. 0 bits per dimension in about 10 hours and it takes approximately 5 days to converge to 2. Setup To run this code you need the following: a machine with multiple GPUs Python3 Numpy, TensorFlow and imageio packages: Training the model Use the script to train the model. To train the default model
on CIFAR-10 simply use: You might want to at least change the and which point to paths on your system to download the data to (if not available), and where to save the checkpoints. To train on fewer GPUs we recommend using to narrow the visibility of GPUs to only a few and then run the script. Don't forget to modulate the flag accordingly. Have a look at the classes in the folder. You have to write an analogous
data iterator object for your own dataset and the code should work well from there. Pretrained model checkpoint You can download our pretrained (TensorFlow) model that achieves 2. Citation If you find this code useful please cite us in your work: pixel-cnn-rotations
Frequently asked questions
How many stars and forks are recorded?
openai/pixel-cnn has 1,956 stars and 433 forks.
What is the repository rank?
openai/pixel-cnn is ranked 9,194.
What language and category are recorded?
openai/pixel-cnn has 1,956 stars, uses Python and belongs to Model Development.
Which repository is closest in rank?
openai/pixel-cnn is compared with philipperemy/tensorflow-1.4-billion-password-analysis, ranked 9,196 with 1,956 stars.
README
Status: Archive (code is provided as-is, no updates expected)
pixel-cnn++
This is a Python3 / Tensorflow implementation
of PixelCNN++, as described in the following
paper:
PixelCNN++: A PixelCNN Implementation with Discretized Logistic Mixture Likelihood and Other Modifications, by
Tim Salimans, Andrej Karpathy, Xi Chen, Diederik P. Kingma, and Yaroslav Bulatov.
Our work builds on PixelCNNs that were originally proposed in van der Oord et al.
in June 2016. PixelCNNs are a class of powerful generative models with tractable
likelihood that are also easy to sample from. The core convolutional neural network
computes a probability distribution over a value of one pixel conditioned on the values
of pixels to the left and above it. Below are example samples from a model
trained on CIFAR-10 that achieves 2.92 bits per dimension (compared to 3.03 of
the PixelCNN in van der Oord et al.):
Samples from the model (left) and samples from a model that is conditioned
on the CIFAR-10 class labels (right):
This code supports multi-GPU training of our improved PixelCNN on CIFAR-10
and Small ImageNet, but is easy to adapt
for additional datasets. Training on a machine with 8 Maxwell TITAN X GPUs achieves
3.0 bits per dimension in about 10 hours and it takes approximately 5 days to converge to 2.92.
Setup
To run this code you need the following:
a machine with multiple GPUs
Python3
Numpy, TensorFlow and imageio packages:
pip install numpy tensorflow-gpu imageio
Training the model
Use the train.py script to train the model. To train the default model on
CIFAR-10 simply use:
python3 train.py
You might want to at least change the --data_dir and --save_dir which
point to paths on your system to download the data to (if not available), and
where to save the checkpoints.
I want to train on fewer GPUs. To train on fewer GPUs we recommend using CUDA_VISIBLE_DEVICES
to narrow the visibility of GPUs to only a few and then run the script. Don't forget to modulate
the flag --nr_gpu accordingly.
I want to train on my own dataset. Have a look at the DataLoader classes
in the data/ folder. You have to write an analogous data iterator object for
your own dataset and the code should work well from there.
Pretrained model checkpoint
You can download our pretrained (TensorFlow) model that achieves 2.92 bpd on CIFAR-10 here (656MB).
Citation
If you find this code useful please cite us in your work:
@inproceedings{Salimans2017PixeCNN,
title={PixelCNN++: A PixelCNN Implementation with Discretized Logistic Mixture Likelihood and Other Modifications},
author={Tim Salimans and Andrej Karpathy and Xi Chen and Diederik P. Kingma},
booktitle={ICLR},
year={2017}
}
