Geary.Z

您好，我是汪老师的学生。RichpediaMEL里每条数据并未给出实体对应的图片名或者路径，从网盘下载图片也未能从图片名中看出端倪，请问实体与图片之间是怎么对应的？

How to add slug for terminal

1

The URL for terminal seems always randomly generated, even if I add slug in the configuration file like this: ```toml [account] api_key = "xxxx" [terminal.Terminal] slug = "a100server" name =...

In some multi-node environment like SLURM，there are some environment vars that contain special chars and can trigger errors when being exported. For example, there is a var `SLURM_JOB_CPUS_PER_NODE=64(x2)` when requesting...

Support GPU Training

22

This PR supports training Cambrian-8b on GPU with deepspeed zero2. Main modification includes 1. We remove all the `.float()` that satisfies TPU's precision and use bf16 unifiedly. 2. We fix...

本地推理效果和huggingface demo不一样

4

感谢大佬开源的模型！ 我在huggingface demo上试用了一下，效果很不错： 但是在本地用相同的配置跑同一张图片，检测效果却变差了： 命令：`p2t predict -i "/Users/geary.z/Desktop/截屏2024-09-18 18.10.00.png" --save-debug-res debug.jpg --file-type text_formula --layout-config "{\"scores_thresh\": 0.45}"` 输出： ``` For a single mutation in the learngenes, we consider its possibility independently...

Wrong logo of LMM-R1 in the survey

1

Thanks for your introduction about LMM-R1 in the survey! However, we find the logo of LMM-R1 in the survey is wrongly represented as microsoft. LMM-R1 is developed by Southeast University...

Merge lmm-r1 for Multimodal PPO

2

Breaking changes: - Support Qwen2.5-VL, Phi3.5/4 - Easily customize new model - Integrate dynamic sampling #968 - Integrate lora+vllm #869 - prompt with visual information is required to be a...