Awesome Open Source
Awesome Open Source

CI PyPI version Documentation Gitter Forum


CTranslate2 is a C++ and Python library for efficient inference with Transformer models.

The project implements a custom runtime that applies many performance optimization techniques such as weights quantization, layers fusion, batch reordering, etc., to accelerate and reduce the memory usage of Transformer models on CPU and GPU. The following model types are currently supported:

  • Encoder-decoder models: Transformer base/big, M2M-100, NLLB, BART, mBART, Pegasus, T5, Whisper
  • Decoder-only models: GPT-2, OPT

Compatible models should be first converted into an optimized model format. The library includes converters for multiple frameworks:

The project is production-oriented and comes with backward compatibility guarantees, but it also includes experimental features related to model compression and inference acceleration.

Key features

  • Fast and efficient execution on CPU and GPU
    The execution is significantly faster and requires less resources than general-purpose deep learning frameworks on supported models and tasks thanks to many advanced optimizations: layer fusion, padding removal, batch reordering, in-place operations, caching mechanism, etc.
  • Quantization and reduced precision
    The model serialization and computation support weights with reduced precision: 16-bit floating points (FP16), 16-bit integers (INT16), and 8-bit integers (INT8).
  • Multiple CPU architectures support
    The project supports x86-64 and AArch64/ARM64 processors and integrates multiple backends that are optimized for these platforms: Intel MKL, oneDNN, OpenBLAS, Ruy, and Apple Accelerate.
  • Automatic CPU detection and code dispatch
    One binary can include multiple backends (e.g. Intel MKL and oneDNN) and instruction set architectures (e.g. AVX, AVX2) that are automatically selected at runtime based on the CPU information.
  • Parallel and asynchronous execution
    Multiple batches can be processed in parallel and asynchronously using multiple GPUs or CPU cores.
  • Dynamic memory usage
    The memory usage changes dynamically depending on the request size while still meeting performance requirements thanks to caching allocators on both CPU and GPU.
  • Lightweight on disk
    Quantization can make the models 4 times smaller on disk with minimal accuracy loss.
  • Simple integration
    The project has few dependencies and exposes simple APIs in Python and C++ to cover most integration needs.
  • Configurable and interactive decoding
    Advanced decoding features allow autocompleting a partial sequence and returning alternatives at a specific location in the sequence.

Some of these features are difficult to achieve with standard deep learning frameworks and are the motivation for this project.

Installation and usage

CTranslate2 can be installed with pip:

pip install ctranslate2

The Python module is used to convert models and can translate or generate text with few lines of code:

translator = ctranslate2.Translator(translation_model_path)

generator = ctranslate2.Generator(generation_model_path)

See the documentation for more information and examples.


We translate the En->De test set newstest2014 with multiple models:

  • OpenNMT-tf WMT14: a base Transformer trained with OpenNMT-tf on the WMT14 dataset (4.5M lines)
  • OpenNMT-py WMT14: a base Transformer trained with OpenNMT-py on the WMT14 dataset (4.5M lines)
  • OPUS-MT: a base Transformer trained with Marian on all OPUS data available on 2020-02-26 (81.9M lines)

The benchmark reports the number of target tokens generated per second (higher is better). The results are aggregated over multiple runs. See the benchmark scripts for more details and reproduce these numbers.

Please note that the results presented below are only valid for the configuration used during this benchmark: absolute and relative performance may change with different settings.


Tokens per second Max. memory BLEU
OpenNMT-tf WMT14 model
OpenNMT-tf 2.26.1 (with TensorFlow 2.9.0) 283.0 3475MB 26.93
OpenNMT-py WMT14 model
OpenNMT-py 2.2.0 (with PyTorch 1.11.0) 474.2 1543MB 26.77
- int8 510.6 1455MB 26.72
CTranslate2 2.17.0 1220.2 1072MB 26.77
- int16 1534.8 920MB 26.82
- int8 1737.5 771MB 26.89
- int8 + vmap 2122.4 666MB 26.62
OPUS-MT model
Transformers 4.19.2 230.1 2840MB 27.92
Marian 1.11.0 756.6 13819MB 27.93
- int16 718.4 10395MB 27.65
- int8 853.3 8166MB 27.27
CTranslate2 2.17.0 988.0 995MB 27.92
- int16 1285.7 847MB 27.51
- int8 1469.1 847MB 27.71

Executed with 8 threads on a c5.metal Amazon EC2 instance equipped with an Intel(R) Xeon(R) Platinum 8275CL CPU.


Tokens per second Max. GPU memory Max. CPU memory BLEU
OpenNMT-tf WMT14 model
OpenNMT-tf 2.26.1 (with TensorFlow 2.9.0) 1289.3 2667MB 2407MB 26.93
OpenNMT-py WMT14 model
OpenNMT-py 2.2.0 (with PyTorch 1.11.0) 1271.4 2993MB 3553MB 26.77
FasterTransformer 4.0 2941.3 5869MB 2327MB 26.77
- float16 6497.4 3917MB 2325MB 26.83
CTranslate2 2.17.0 3644.1 1231MB 646MB 26.77
- int8 5393.6 975MB 522MB 26.83
- float16 5454.7 815MB 550MB 26.78
- int8 + float16 6158.6 687MB 523MB 26.80
OPUS-MT model
Transformers 4.19.2 811.1 4013MB 3044MB 27.92
Marian 1.11.0 2172.9 3127MB 1869MB 27.92
- float16 2722.0 2985MB 1715MB 27.93
CTranslate2 2.17.0 3042.5 1167MB 486MB 27.92
- int8 4573.1 1007MB 511MB 27.89
- float16 4718.4 783MB 552MB 27.85
- int8 + float16 5300.5 687MB 508MB 27.81

Executed with CUDA 11 on a g4dn.xlarge Amazon EC2 instance equipped with a NVIDIA T4 GPU (driver version: 510.47.03).

Additional resources

Related Awesome Lists
Top Programming Languages

Get A Weekly Email With Trending Projects For These Topics
No Spam. Unsubscribe easily at any time.
C Plus Plus (257,795
Deep Learning (39,275
Translation (13,664
Gpu (9,488
Cpu (8,589
Deep Neural Networks (4,053
Decoding (3,350
Openmp (1,600
Machine Translation (957
Neon (933
Avx (873
Quantization (827
Parallel Computing (772
Neural Machine Translation (224
Inference Engine (195
Gemm (148
Transformer Models (55
Intrinsics (50
Mkl (45
Thrust (19
Opennmt (18
Onednn (3