polestar comments

Results 3 comments of


                                            polestar

renderer.render()not enough values to unpack (expected 4, got 2)

> (gaussian_splatting) xxx@amax:/data/xxx/dreamgaussian-main$ python main.py --config configs/image.yaml input=data/zelda_rgba.png save_path=zelda_rgba [INFO] load image from data/zelda_rgba.png... Number of points at initialisation : 5000 [INFO] loading zero123... Loading pipeline components...: 100%|████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████| 6/6 [00:00

When using gradient accumulation, does the order of optimizer.zero_grad() affect training?

> [@polestarss](https://github.com/polestarss) Our gradient accumulation is implemented using `accelerate`, so there's no need to worry—it works as expected. When we call `accelerator.backward(loss)`, the gradients are temporarily stored in GPU memory...

使用deepspeed zero_stage 3训练，遇到张量shape不一致的问题该怎么解决？

I suffered the same problem