jaclyn

Hi, thank you for sharing. I noticed on the homepage that they mentioned the calculation method for EF score. It states that the volume is estimated by integrating the left...

Hi, thanks for sharing! I am trying to train Painter using toy_dataset. I found in /Painter/data/pairdataset.py that the data processing section is designed with a weight parameter called **type_weight_list.** In...

### Prerequisite - [X] I have searched [Issues](https://github.com/open-mmlab/mmocr/issues) and [Discussions](https://github.com/open-mmlab/mmocr/discussions) but cannot get the expected help. - [X] The bug has not been fixed in the [latest version (0.x)](https://github.com/open-mmlab/mmocr) or...

Following your previous suggestions, I set up distributed computing during training and executed the command as shown in the figure below. The corresponding settings in the config.yaml remained unchanged. However,...

Where is mcloader defined?

2


关于 Qwen2_VL-2B 微调时显存不足的问题

8

您好， 我在使用自己的图文数据集尝试微调 **Qwen2_VL-2B**，训练时设置了使用 default-zero2 ，出现显存占用不均，单张显卡显存不足报error的情况。下面我运行的完整sh: ``` SIZE_FACTOR=8 MAX_PIXELS=602112 CUDA_VISIBLE_DEVICES=0,1,2,3,4,5,6,7 NPROC_PER_NODE=8 swift sft \ --model_type qwen2-vl-2b-instruct \ --model_id_or_path /projects/Qwen2_VL/Qwen2-VL-2B-Instruct \ --sft_type lora \ --lora_target_modules ALL \ --tuner_backend peft \ --deepspeed default-zero2...