WuChannn

Wrong with Table of Content

2

Hello, I get wrong position of table of content, and it covers the contents i want to show, how can i fix it ![Screen Shot 2019-03-27 at 4 51 00...

hello, how can I tune the position of github icon before twitter? I change the code in _config.yml, like ` github_username: yingjieYJH` `twitter_username: yingjieYJH ` but it doesn't work.

hello, I want to define the batch_size directly in input_data layer, and set batch_size in model.fit to None. However, I get an error " Incoming Tensor shape must be 4-D,...

what tag means

1

@Duankaiwen Hello, Kaiwen. I would like to know what tag means? I noticed there was ‘dists’ in kp_utils.py and it represented the distance between ’tl_tag’ and ‘br_tag’. Could you please...

RuntimeError: invalid argument 5: k not in range for dimension

11

@Duankaiwen hello, kaiwen when i test my test dataset using my own trained model, i met with this problem, and the following is my log: cfg_file: config/CenterNet-52.json loading all datasets......

pods_train: command not found

5

hi, I follow the installation instruction, and everything is ok until comes to "Train" part. After cd into playground/detection/coco/yolof/yolof.res50.C5.1x, I run the shell script `pods_train --num_gpus 1`, and get `pods_train:...

如标题描述

Can't pickle local object 'linear_scheduler.<locals>.lr_lambda'

2

when i use multiple gpus to train GeoLayoutLM, error as follows: Can't pickle local object 'linear_scheduler..lr_lambda' @wdp-007 @congyao @alibaba-oss Looking forward to your reply.

BGE-M3 unify微调时，forward函数中self.use_inbatch_neg的实际含义？

2

@hotchpotch @yzhliu @zh217 @neofung 首先，特别感谢您的优秀工作。在学习您的工作时遇到了理解上的问题，想请教您。如题所述，对于这个含义有点不清楚，请帮忙解答一下。我的理解如下，请帮忙看看是否正确：**use_inbatch_neg：同一个batch中query对应的neg数据是否参与loss计算。** 具体分析其中的代码（FlagEmbedding/baai_general_embedding/finetune/modeling.py forward函数），又有些不理解的地方： 1. if self.use_inbatch_neg: 为每个查询创建了一个查询索引，将查询索引乘以 group_size，确保了每个查询都指向其对应的文档组的第一个文档，这可以视为正样本。但是后续的loss计算为self.compute_loss(scores, target)，我的理解是只计算了正样本与query之间的loss，这里并没有体现出**use_inbatch_neg** 2. else:（也即 not self.use_inbatch_neg） 为每个查询创建了一个查询索引0, 表示每个查询只考虑第一个文档。第一个文档作为batch中的第一个文档，它只与第一个query对应，且为第一个query的正样本。将第一个query的正样本与所有的query求loss，这样是为了区分不同的query吗？那为什么不用第二个query的第一个文档与所有的query求loss呢？ 非常期待您的回复。

是否支持全参微调？

2

现在放出的微调接口是Lora微调的，是否可以在训练资源充足的情况下进行全参微调呢？是不是注释掉 `peft_lora.py`中的`model = get_peft_model(model, peft_config)`，就可以了？