💪 FP8 compatibility ! 🚀 Speed Up all Process 🚀 less VRAM consumption (Stay high, batch_size=1 for RTX4090 max, I'm trying to fix that) 🛠️ Better benchmark coming soon ...
In fact, openpi's official PyTorch implementation cannot align with the JAX version, and we have internally reproduced this result as well. Based on this issue, we re-implemented a new PyTorch version ...
Some results have been hidden because they may be inaccessible to you
Show inaccessible results