bhsueh_NV comments

Results 639 comments of


                                            bhsueh_NV

mt5 support？

Please refer https://github.com/NVIDIA/FasterTransformer/issues/225.

mt5 support？

mt5 is supported in latest release. You can refer https://github.com/NVIDIA/FasterTransformer/blob/main/docs/t5_guide.md#running-mt5.

mt5 support？

Close this bug because it is inactivated. Feel free to re-open this issue if you still have any problem.

The fp16 inference of pytorch swintransformer op got `nan` output.

```bash CUDA Error: (null) /workdir/xxx/packages/v5.0_tag/FasterTransformer-release-v5.0_tag/3rdparty/trt_fused_multihead_attention/fused_multihead_attention_v2.h 682 ``` This error means that you don't call fused mha successfully. Can you provide the docker image you use and the building steps you...

The fp16 inference of pytorch swintransformer op got `nan` output.

I believe CUDA 11.0 is runnable. I try to build the cpp example by `nvcr.io/nvidia/pytorch:20.07-py3`, which contains CUDA 11.0. I can run the cpp example successfully by following scripts: ```bash...

In fp16 mode, slice the result of vit, but the result is nan

What's your n ?

In fp16 mode, slice the result of vit, but the result is nan

Try to run without half2.

In fp16 mode, slice the result of vit, but the result is nan

For half, you need to set data_type_factor to be 1.

In fp16 mode, slice the result of vit, but the result is nan

You can print the values of inputs and outputs of your slice kernel under FP32 and FP16.

In fp16 mode, slice the result of vit, but the result is nan

Close this bug because it is inactivated. Feel free to re-open this issue if you still have any problem.