Jiarui Fang（方佳瑞）
Jiarui Fang（方佳瑞）
README写了，编译时候应该 cmake .. -DWITH_GPU=ON，或者你下载成了cpu-only的镜像了
I want to compare Turbo with ONNXRT on ALBERT. However, I found some ops in PyTorch does not support ONNX.
I have noticed this issue and left a comment on the PyTorch issue.
Accuracy is checked in unitests. Therefore benchmark is only for performance evaluation. Scripts in ./python/example compared the accuracy results
你在镜像里编译一下最新代码吧，你用的example代码是最新的，镜像里的代码不一定是。 你现在用turbo做backend，却用onnxrt的方式调用 https://github.com/Tencent/TurboTransformers/blob/f43f35b792/example/python/bert_example.py#L65
你要不直接用我提供的镜像试试？ FROM thufeifeibear/turbo_transformers_cpu:latest 在这个镜像里编译 你如果要自己从头开始编译，自己在container里看一下conda install哪个具体的包出问题，然后更改一下。
May be AutoModel is a new interface. Is `RobertaForSequenceClassification ` ok for you?
The Turbo is not ready for transformers 3.4.0. You can maintain a local turbo version for youself.
Hey, We noticed that the Transformers are becoming popular in CV. We currently have no plan to update new models. However, we believe it is easy to be adopted in...
Motivation: we use FBGEMM in order to have consistent accuracy as PyTorch dynamic quantization. As TurboTransformer's optimizations are focused on Non-GEMM operations, we can reuse PyTorch QLinear code as much...