zhuzhu

 All my arguments are as follows: { "batch_size": 8, "board_interval": 50, "checkpoint_path": null, "d_reg_every": 16, "dataset_type": "ffhq_encode", "delta_norm": 2, "delta_norm_lambda": 0.0002, "encoder_type": "Encoder4Editing", "exp_dir": "./experiment", "id_lambda": 0.5, "image_interval": 100,...

int8 quantified model

1

Hi! @htwang14 Thanks for your nice work! I have seen the Conv2dQuant and ConvTrans2dQuant operators in the in models/models.py file, but the feature map before the quantized convolution and weights...

**Describe the feature** **Motivation** 是否可以借助contrastive self-distillation来压缩超分模型，同时尽可能不降低图像质量 **Related resources** [CSD loss](https://arxiv.org/abs/2105.11683)

测试支持的batch_size[Feature]

2

### What's the feature? 测试时能不能支持samples_per_gpu > 1的情况，只要设置的dataset大小不能保证被num_gpu和samples_per_gpu均匀分配就会报错 ### Any other context? _No response_

https://github.com/tensorboy/pytorch_Realtime_Multi-Person_Pose_Estimation/blob/5ce5a4fbeb3cd6ffcd7aec18d1b5779d518a8736/evaluate/coco_eval.py#L153 https://github.com/tensorboy/pytorch_Realtime_Multi-Person_Pose_Estimation/blob/5ce5a4fbeb3cd6ffcd7aec18d1b5779d518a8736/evaluate/coco_eval.py#L159

您好，感谢您的工作，这个标注工具只能标注头顶和脖子吗

code seems incompleted!

5

In your code just use the 2D joints to train model, and there are no test data, can you provide full code?

unsupported cog version

1

Hello, your demo on replica can not run no longer. It prompts you have used the unsupported cog version. Can you update it?

dynamic dimension shape input

1

Does QNN support input of dynamic dimensions, such as the width and height of input pictures are usually not fixed in super-resolution models. Any plans to implement this feature? ONNX...