Jack Chen

最好能分以下类，搞个一二三级标题，目前基本是平铺出来的。 比如 pruning，knowledge distillation等技术，可以放到一起

feature request: support pytorch torchdynamo

1

Torch-TensorRT acts as an extension to TorchScript. But there are models that are not supported by pytorch jit tracing, see [warning in jit tracing doc](https://pytorch.org/docs/stable/generated/torch.jit.trace.html). Would you guys have plan...

Implement BatchNorm in triton

2

batchnorm(bn) is very popular in CV, almost every conv op will be followed by bn. I see [layernorm](https://triton-lang.org/master/getting-started/tutorials/05-layer-norm.html#) in triton achieved best HBM bandwidth. So I'm curious about implement [batchnorm](https://github.com/NVIDIA/apex/blob/master/apex/parallel/sync_batchnorm.py)...

### Is there an existing issue for this? - [X] I have searched the existing issues and checked the recent builds/commits ### What would your feature do ? hi guys...

use torchDynamo instead of tracing to get graph?

3

hi guys, this project is awesome comparing with torch pipe, I have some models that are [not supported by tracing](https://pytorch.org/docs/stable/fx.html#limitations-of-symbolic-tracing). Would you guys have plan to support [torchDynamo](https://github.com/pytorch/torchdynamo) to get...

TE 1.4 fused attn caused a NaN when backward

1

I met an nan using TE 1.4, after I use pytorch autograd.detect_anomaly, it captured the NaN traceback: ``` Error detected in FusedAttnFuncBackward. ... [rank5]: RuntimeError: Function 'FusedAttnFuncBackward' returned nan values...

will you guys have roadmap to support multimodal modes such as qwen2 vl, llava series?