TurboTransformers v0.5.1

Albert Model uses the model-aware-allocator.

TurboTransformers v0.5.0

Add Model Aware Allocator for Bert Model.

TurboTransformers v0.4.2

Add Quantized Bert using onnxruntime.

TurboTransformers v0.4.1

Using onnxruntime-cpu as CPU backend, parallel to our own home-grown implementation.

TurboTransformer v0.3.0

Support Transformer decoder used in OpenNMT-py.
New GPU memory allocator.
Be Compatible with Pytorch v1.5.0.

TurboTransformer v0.2.1

Add blis to BLAS options.

TurboTransformer v0.0.1

Bert Acceleration on CPU and GPU.