JasonChi

Wondering how to get such 3D skeleton pose. I searched for it in AMASS but could not find it

我看到蒸馏用的loss是： def direct_distill_loss(self, **inputs): self.scheduler.set_timesteps(inputs["num_inference_steps"]) models = {name: getattr(self, name) for name in self.in_iteration_models} for progress_id, timestep in enumerate(self.scheduler.timesteps): timestep = timestep.unsqueeze(0).to(dtype=self.torch_dtype, device=self.device) noise_pred = self.model_fn(**models, **inputs, timestep=timestep, progress_id=progress_id) inputs["latents"]...

I also encounter this problem. My set is python 3.10.18 cuda 12.8 torch 2.10.0.dev20250924 flash-attn 2.8.3