3 ms·Boosting Wan2.2 I2V Inference on 8xH100s, 56% Faster with Sequence Parallelism6 points by palakzat 1y agoblackboattech 1y agoHow does it scale for >8 H100s ?