NNsauce

here is complete error: Epoch 0: : 0it [00:00, ?it/s]Update finite_difference_eps to 0.027204705103003882 Epoch 0: : 500it [00:26, 18.89it/s, loss=0.0754, train/inv_s=42.50, train/num_rays=512.0] Traceback (most recent call last): | 0/49 [00:00

I uh reduce chunk_size from 2048 to 1024, then it works. But why does it need so much more GPU memory when without mask?

3D imaging means 3D reconstruct? then why not NeRF?

@Cadene thanks so much man! On For quetion 2 , I am still not clear about the time alignment of video and state, are they recorded in a same thread...

> Well, regardless of the Gemma 3 version chosen (including non-QAT one), llmexport.py responds with `AttributeError: 'NoneType' object has no attribute 'weight'`. I got same error on Qwen2.5-VL-7B-Instruct, even under...

 it works after I change the value of 'lm_' of self.model_map['model']: '**model.lm_hea**d' -> '**lm_head**' (just like that value of Qwen2.5), then I pass the left converting and quanting. I...