update torch2trt.py to make it work with windows and add logs to have…

[MaraScott]

… a sense of progression

[pfn]

I tried fetching this and running it, still encountering the same error:

personalive] $ python torch2trt_win.py
C:\Users\pfnguyen\Programs\miniconda3\envs\personalive\lib\site-packages\transformers\utils\generic.py:441: FutureWarning: `torch.utils._pytree._register_pytree_node` is deprecated. Please use `torch.utils._pytree.register_pytree_node` instead.
  _torch_pytree._register_pytree_node(
C:\Users\pfnguyen\Programs\miniconda3\envs\personalive\lib\site-packages\transformers\utils\generic.py:309: FutureWarning: `torch.utils._pytree._register_pytree_node` is deprecated. Please use `torch.utils._pytree.register_pytree_node` instead.
  _torch_pytree._register_pytree_node(
[DBG] Loading config: ./configs/prompts/personalive_online.yaml
[DBG] Loading PoseGuider...
[DBG] Loading MotionEncoder...
[DBG] Loading UNet3DConditionModel...
[DBG] Loading VAE...
[DBG] Creating scheduler...
[DBG] Building wrapped model...
[DBG] Exporting ONNX to: ./pretrained_weights/onnx/unet/unet.onnx
['pred_video', 'latents', 'pose_cond_fea_out', 'motion_hidden_states_out', 'motion_out', 'latent_first']
开始导出 ONNX 模型到: ./pretrained_weights/onnx/unet/unet.onnx ...
C:\Users\pfnguyen\Programs\miniconda3\envs\personalive\lib\site-packages\diffusers\models\embeddings.py:175: TracerWarning: Converting a tensor to a Python boolean might cause the trace to be incorrect. We can't record the data flow of Python values, so this value will be treated as a constant in the future. This means that the trace might not generalize to other inputs!
  if self.height != height or self.width != width:
... snip ...
RuntimeError: Expected tensor for argument #1 'input' to have the same type as tensor for argument #2 'weight'; but type torch.cuda.FloatTensor does not equal torch.cuda.HalfTensor (while checking arguments for cudnn_batch_norm)
personalive] $

this is in msys on windows, could be that the versions of packages are too new, e.g. pytorch 2.9.1+cu130 (to support blackwell)

yep, downgrading to pytorch 2.1.2 makes the error go away (however, there's no blackwell optimization, so it sucks)