光源殿下

I run the train with "python -m torch.distributed.launch --nproc_per_node=4" and using dataset=cityscape backbone=resnet50 batchsize=4 my GPU is Nvidia Titan X *4 12GB memory per card, and CUDA out of memory...

I've tried 2 ways to obtain a .pdiparams for C++ inferring: (我想要一个sam模型的静态模型,所以我先试了直接找paddle的export.py，但是没有这个文件，加代码我也没跑通， net = to_static(model, input_spec=[InputSpec（这里要填什么东西啊求助）, name='batched_input'), InputSpec(shape[None, 1], dtype=paddle.bool, name='multimaskk_output') ])这第一个batched_input的Inputspec该怎么填啊（我有点菜）） 1, Export from .pdparams.　 However, export.py is missing...

Paddle-Lite通过X2Paddle转化出的nb模型出错

4

我这边使用的yolov5的pth模型，转化成onnx之后，通过paddle-lite （2.13rc0）与 X2Paddle（1.4.1） 转化成了nb文件，试图部署在安卓环境下，报错  使用的命令是 x2paddle --framework=onnx --model=onnx_model.onnx --save_dir=pd_model --to_lite=True --lite_valid_places=arm --lite_model_type=naive_buffer 不知道是什么问题？以及问一句，这种转化方式的分辨率是和onnx一致么？精度是多少呢？可以自己定义精度么？

Error Detection with unexpected labels: [CLS]and[SEP]

2

I use offlinemode, and config_file = 'D:/deeplearning/Grounded-Segment-Anything-main/GroundingDINO/groundingdino/config/GroundingDINO_SwinT_OGC.py' # change the path of the model config file checkpoint_path = 'D:/deeplearning/Grounded-Segment-Anything-main/bert-base-uncased/groundingdino_swint_ogc.pth' # change the path of the model image_path = 'D:/deeplearning/Grounded-Segment-Anything-main/1.jpg' text_prompt...

fail on calibration

16

After carefully calibrate, I run the copy process, and met multiple failures. Now I print the angle, and just found that the main leader arm shows: >>> print(leader_pos) [ -89.82422...

UnboundLocalError: local variable 'safe_action' referenced before assignment

1

>>> robot.connect() Connecting main follower arm. Connecting main leader arm. Activating torque on main follower arm. >>> import tqdm >>> seconds = 30 >>> frequency = 200 >>> for _...