Pastel！
Pastel！
> 我是在window上运行报Fail to predict the shape on operator name: 'self.bn1.BatchNormalization', type: 'BatchNormalization'这个错误，在linux上运行正常。估计是安装包环境问题 我在linux上同样也遇到了这个问题...
Same question here.
> 1. Remove your your triton directory and clone it from github again. > 2. Setup `MAX_JOBS=`, where `` is a small number, depending on your processor. still struggling with...
> _No description provided._ best.pt 如果也是从tiny训出来的话 收敛会更快一些
> Can you share > > * the script to build engine > * the `config.pbtxt` of your backend settings I enabled the options with " --dtype float16 --remove_input_padding --use_gpt_attention_plugin...
> I don't see you setup the beam_width when you build the engine. Can you try adding `--max_beam_width 4`? I've added the option and the problem exists. [TensorRT-LLM][ERROR] Encountered error...
> Thanks. I find the limitation in batch manager that I missed. Could you modify this issue or open another issue to require this feature? OK
> The model you posted has been inspected, in which the size of buffers is only 36KB. So, the only way to lower the size of the model is to...
> So it will be optimized from 8 ops (10 tensors) to 5 ops (8 tensors) for each time step. Take a glance at the previous implementation of `AtenGRUOperator`, it...
> Well, it is doable and not hard at all if only GRU is involved, but for a better design, it takes some time. One week is enough? I'm available...