How to convert functional.Tensor to numpy or a list of Tensor ?

[lzcchl]

Due to upgrading the version of trtllm from 0.8.0 to 0.9.0, I originally obtained the "host_max_attention_window_sizes" as a List in models/generation_mixin.py, but now it is a Tensor, so in order to meet my original interface, I must convert the Tensor to a List or numpy, I can find the function about that, is there any function I can use ?

[byshiue]

In 0.8.0, although host_max_attention_window_sizes is a list, but the list still store the TensorRT-LLM tensors instead of numpy or torch tensor. How do you use these tensors at the time?

Also, as a input tensor in TensorRT-LLM, this means that we will pass inputs in generation.py by torch tensor, so you should be able find them in the generation.py (in python runtime).