silence-moon

train on vimeo90K

1

Hi, it is very pretty for your work. when I try to train your model on vimeo90K, I go t he follow result: I want to ask is it normal?

使用data中的数据进行PT微调，但是发现微调过程中损失一直为nan，同时训练到一半的时候也会出现Current loss scale already at minimum - cannot decrease scale anymore这个错误。最主要还是损失为nan，请问大佬，这是什么情况？（注：当前使用的模型是当前可以从官网下载的最新的chatglm3-6b的模型） 运行命令： CUDA_VISIBLE_DEVICES=0 deepspeed --master_port 520 train.py \ --train_path data/spo_0.json \ --model_name_or_path /mnt/workspace/chatglm3-6b/ \ --per_device_train_batch_size 4 \ --max_len 1560 \...

Distance issue displayed during orbslam3 positioning

1

This is a very excellent project. When I was using my own camera for testing, I found that on the same road section, when using ORBSLAM3 to locate the composition,...

启动之后没有任何反应

21

系统启动起来之后，无图像也没有建图的输出。若是改变模型路径，则会提示找不到模型，说明应该是在运行，但是具体卡在哪里了，不太清楚 

problem of isaac_ros_visual_slam_dual_realsense_fusion.launch.py

2

Thanks for your great work. hardware: Jetson-AGX Xavier platform: docker isaac_ros_visual_slam version: dp3.1 description: I can launch single realsense camera on slam system. However, when I try to launch the...

cannot localize in map

1

I want to use the latest isaac-ros-visual-slam version of isaac 3.2 to achieve localize in an exist map. I send localization service by visual_slam/localize_in_map to localize current stereo image and...