HUAFOR issues

Results 9 issues of


                                            HUAFOR

请问训练模型所需要的最少GPU是多少呢？

学校小作坊没有那么多卡，看论文提到是用8张A100训练瑟瑟发抖，不知道能不能作为一个baseline改进。。。

大哥，代码好像一直跑不通啊，一直报类似的错The size of tensor a (64) must match the size of tensor b (128) at non-singleton dimension 1

When the code will be released?

When will the code be released, please?

Amazing work! When will the code be released? Thank you.

Awesome work！I would like to know how to reproduce the performance values in the experimental table 1 of your paper?

非常感谢作者您为社区带来的开源贡献！我想知道如何通过您提供的checkpoint来复现您论文中表格1里面的精度数据(以lrs2为例) inference_single.py只提供了一段测试输出的视频，没有提供相应的精度数据~ ![image](https://github.com/Weizhi-Zhong/IP_LAP/assets/58834906/4df5f77d-4c2b-40ee-98e0-87f6ddad8d54)

Results of ABC->D weights published on huggingface are higher than those reported in the paper

Is this normal? https://hf-mirror.com/roboflamingo/RoboFlamingo/blob/main/ABC_D/checkpoint_gripper_post_hist_1_aug_10_4_traj_cons_ws_12_mpt_dolly_3b_4.pth ![image](https://github.com/RoboFlamingo/RoboFlamingo/assets/58834906/58c58272-463c-4341-82e2-fa890b727d37) ![image](https://github.com/RoboFlamingo/RoboFlamingo/assets/58834906/61c1f3ce-09e4-468e-b927-30cf6c1c193a)

Other ways of DirectInversion

想请教下作者，直接存储prompt_src在inversion过程中每一步的latent而不是loss，然后直接用于sampling过程的非编辑区域，相当于非编辑区域用prompt_src部分的latent，编辑区域用prompt_tgt的latent，写法是不是和原来是等价的？