https://github.com/seannaren/warp-ctc

Pytorch Bindings for warp-ctc
https://github.com/seannaren/warp-ctc

Last synced: about 10 hours ago
JSON representation

Pytorch Bindings for warp-ctc

• Host: GitHub
• URL: https://github.com/seannaren/warp-ctc
• Owner: SeanNaren
• License: apache-2.0
• Created: 2017-01-24T20:35:30.000Z (almost 8 years ago)
• Default Branch: pytorch_bindings
• Last Pushed: 2023-07-02T07:31:58.000Z (over 1 year ago)
• Last Synced: 2024-12-18T06:04:22.510Z (7 days ago)
• Language: Cuda
• Homepage:
• Size: 379 KB
• Stars: 757
• Watchers: 17
• Forks: 271
• Open Issues: 98
• Metadata Files:
  • Readme: README.md
  • License: LICENSE

Awesome Lists containing this project

README

        
# PyTorch bindings for Warp-ctc

[![Build Status](https://travis-ci.org/SeanNaren/warp-ctc.svg?branch=pytorch_bindings)](https://travis-ci.org/SeanNaren/warp-ctc)

This is an extension onto the original repo found [here](https://github.com/baidu-research/warp-ctc).

## Installation

Install [PyTorch](https://github.com/pytorch/pytorch#installation) v0.4.

`WARP_CTC_PATH` should be set to the location of a built WarpCTC

(i.e. `libwarpctc.so`).  This defaults to `../build`, so from within a

new warp-ctc clone you could build WarpCTC like this:

```bash

git clone https://github.com/SeanNaren/warp-ctc.git

cd warp-ctc

mkdir build; cd build

cmake ..

make

```

Now install the bindings:

```bash

cd pytorch_binding

python setup.py install

```

If you try the above and get a dlopen error on OSX with anaconda3 (as recommended by pytorch):

```bash

cd ../pytorch_binding

python setup.py install

cd ../build

cp libwarpctc.dylib /Users/$WHOAMI/anaconda3/lib

```

This will resolve the library not loaded error. This can be easily modified to work with other python installs if needed.

Example to use the bindings below.

```python

import torch

from warpctc_pytorch import CTCLoss

ctc_loss = CTCLoss()

# expected shape of seqLength x batchSize x alphabet_size

probs = torch.FloatTensor([[[0.1, 0.6, 0.1, 0.1, 0.1], [0.1, 0.1, 0.6, 0.1, 0.1]]]).transpose(0, 1).contiguous()

labels = torch.IntTensor([1, 2])

label_sizes = torch.IntTensor([2])

probs_sizes = torch.IntTensor([2])

probs.requires_grad_(True)  # tells autograd to compute gradients for probs

cost = ctc_loss(probs, labels, probs_sizes, label_sizes)

cost.backward()

```

## Documentation

```

CTCLoss(size_average=False, length_average=False)

    # size_average (bool): normalize the loss by the batch size (default: False)

    # length_average (bool): normalize the loss by the total number of frames in the batch. If True, supersedes size_average (default: False)

forward(acts, labels, act_lens, label_lens)

    # acts: Tensor of (seqLength x batch x outputDim) containing output activations from network (before softmax)

    # labels: 1 dimensional Tensor containing all the targets of the batch in one large sequence

    # act_lens: Tensor of size (batch) containing size of each output sequence from the network

    # label_lens: Tensor of (batch) containing label length of each example

```