请问学到哪里啦？ 推理出来的视频回贴至原视频有没有好的方法呀？

[thorory]

[abinggo]

能够复现和训练，但是还是画质还是有点糊，贴回去的方法我在b站上有看到一个方法 https://www.bilibili.com/video/BV1qm411R7Lb/?spm_id_from=333.1007.top_right_bar_window_history.content.click&vd_source=6f33a1d32167510196f37e3f7168c96e   亭曈 ***@***.***  

…

------------------ 原始邮件 ------------------ 发件人: ***@***.***>; 发送时间: 2024年4月25日(星期四) 下午2:39 收件人: ***@***.***>; 抄送: ***@***.***>; 主题: [abinggo/genefacepp] 请问学到哪里啦？ 推理出来的视频回贴至原视频有没有好的方法呀？ (Issue #1) — Reply to this email directly, view it on GitHub, or unsubscribe. You are receiving this because you are subscribed to this thread.Message ID: ***@***.***>

[thorory]

谢谢，请问你训练的时候有碰到过训练一下，就自动退出的情况吗？ 并且没有输出报错。
val: Smooth head trajectory (rotation and translation) with a window size of 7 | Copied codes to checkpoints/motion2video_nerf/may_head/codes/20240426161252. | cond_prenet Trainable Parameters: 0.050M | blink_embedding Trainable Parameters: 0.000M | blink_encoder Trainable Parameters: 0.001M | cond_att_net Trainable Parameters: 0.004M | position_embedder Trainable Parameters: 1.807M | ambient_net Trainable Parameters: 0.029M | ambient_embedder Trainable Parameters: 1.807M | sigma_net Trainable Parameters: 0.041M | direction_embedder Trainable Parameters: 0.000M | color_net Trainable Parameters: 0.019M | dropout Trainable Parameters: 0.000M | sr_net Trainable Parameters: 0.271M Sanity Val: 0%| | 0/2 [00:00<?, ?step/s]/mnt/data/llm/avatar/GeneFacePlusPlus/tasks/radnerfs/dataset_utils.py:427: UserWarning: To copy construct from a tensor, it is recommended to use sourceTensor.clone().detach() or sourceTensor.clone().detach().requires_grad_(True), rather than torch.tensor(sourceTensor). sample['lm68'] = torch.tensor(self.lm68s[idx].reshape([68*2])) Setting up PyTorch plugin "bias_act_plugin"... Done. Setting up PyTorch plugin "upfirdn2d_plugin"... Done. /mnt/data/llm/avatar/GeneFacePlusPlus/tasks/radnerfs/dataset_utils.py:427: UserWarning: To copy construct from a tensor, it is recommended to use sourceTensor.clone().detach() or sourceTensor.clone().detach().requires_grad_(True), rather than torch.tensor(sourceTensor). sample['lm68'] = torch.tensor(self.lm68s[idx].reshape([68*2])) Sanity Val: 100%|████████████████████████████████████████████████████████████████████████████████████████████████████| 2/2 [00:02<00:00, 1.15s/step] | Validation results@250000: {'total_loss': 19.5990600586, 'mse_loss': 0.0250960551, 'sr_mse_loss': 0.0267604357, 'lpips_loss': 3.4612528086, 'sr_lpips_loss': 4.3024814129, 'sr_lip_lpips_loss': 2.9367291927, 'lambda_ambient': 8.8467416763} 250000step [00:00, ?step/s, ambient_loss=3.4e-5, head_psnr=16.2, lambda_ambient=8.88, lpips_loss=3.44, lr_optimizer0_params_group0=5e-5, lr_optimizer0_params_group1=0.0005, lr_optimizer0_params_group2=0.00025, mse_loss=0.024, num_non_facemask=57910, sr_lip_lpips_loss=2.85, sr_lpips_loss=inf, sr_mse_loss=0.0257, weights_entropy_loss=0.0472]| Training end.. Epoch 0 ended. Steps: 250001. {'total_loss': inf, 'mse_loss': 0.023985084146261215, 'weights_entropy_loss': 0.047238558530807495, 'num_non_facemask': 57910.0, 'ambient_loss': 3.40268925356213e-05, 'sr_mse_loss': 0.025656795129179955, 'lpips_loss': 3.4387094974517822, 'sr_lpips_loss': inf, 'sr_lip_lpips_loss': 2.845965623855591, 'lambda_ambient': 8.882060050964355, 'head_psnr': 16.200586318969727} 250000step [00:00, ?step/s, ambient_loss=3.4e-5, head_psnr=16.2, lambda_ambient=8.88, lpips_loss=3.44, lr_optimizer0_params_group0=5e-5, lr_optimizer0_params_group1=0.0005, lr_optimizer0_params_group2=0.00025, mse_loss=0.024, num_non_facemask=57910, sr_lip_lpips_loss=2.85, sr_lpips_loss=inf, sr_mse_loss=0.0257, weights_entropy_loss=0.0472]