Mynet实现

运行命令：--g_ckpt=/home/joselyn/workspace/0419-course/stylenerf_pkl/ffhq_256.pkl --outdir=/home/joselyn/workspace/0419-course/logs/StyleNeRF-main/train_mynet/debug 　--data=/home/joselyn/workspace/0419-course/logs/StyleNeRF-main/dataset_create/Mydataset

记录一些中间变量

一个batch对应的相机参数如下：

经过rec_ws_1,rec_cm_1 = E(img_1) 之后：

因为无法从camera_matrices反向计算得到mode，
所以，保存数据集的时候，直接把mode存下来。

运行测试

报错

File "/home/joselyn/workspace/0419-course/StyleNeRF-main1/dnnlib/camera.py", line 174, in look_at
    x_axis = normalize(torch.cross(up[None,:].expand_as(z_axis), z_axis, dim=-1))
RuntimeError: expected scalar type Float but found Double

和train_encoder对应的部分进行对比。
train_encoder 在调用camera_matrices = G.synthesis.get_camera(batch, device, mode=c_samples) 之前：

发现后者的device在 cpu上面。
报错：

File "/home/joselyn/workspace/0419-course/StyleNeRF-main1/dnnlib/camera.py", line 174, in look_at
    x_axis = normalize(torch.cross(up[None,:].expand_as(z_axis), z_axis, dim=-1))
RuntimeError: Expected object of scalar type Float but got scalar type Double for argument #2 'other' in call to _th_cross_kernel_out

修改为float类型。

运行成功

尝试使用预训练过的encoder
命令：python -W ignore train_mynet.py --g_ckpt=/home/joselyn/workspace/0419-course/stylenerf_pkl/ffhq_256.pkl --e_ckpt=/home/joselyn/workspace/0419-course/logs/StyleNeRF-main/train_encoder_1GPU/checkpoints/network-snapshot-000900.pkl --outdir=/home/joselyn/workspace/0419-course/logs/StyleNeRF-main/train_mynet/train_test --data=/home/joselyn/workspace/0419-course/logs/StyleNeRF-main/dataset_create/Mydataset --max_steps=100

效果不错。