glide-the

Results 18 issues of glide-the

标题:对图片化的文档进行内容总结的方案 讨论目的:固定领域大部分pdf会使用文本存储pdf内容,但有些pdf为了保密或者禁止复制与拷贝,会将内容进行图片化,为了对此类的文档进行分析,我的构思是可以使用ocr的视觉模型,对文档区域信息分析,并整理出内容简介 大致思路:通过对 [manga-image-translator](https://github.com/zyddnys/manga-image-translator)项目的阅读,发现作者在对ocr的处理上已经有非常完善的方案 内容评价:出于对使用的目的阐述,我对作者的使用方法感到可惜,如果将此类方案应用到文档分析中,这个方案会有更加成熟的方法,甚至可以商用

feature request

project in the ``` import fetchAdapter from "../util/axios-fetch-adapter.js"; const clientConfig = new Configuration({ ...this.clientConfig, baseOptions: { adapter: fetchAdapter }, }); ``` im edit this code ··· const clientConfig = new...

… fixed length and provides a method called contains_stop_sequence to check if the queue contains the given stop sequence. The clear_torch_cache function is designed to clear GPU memory cache used...

### Describe the bug I encountered a bug while using the transformers.StoppingCriteria method to construct a collector. The issue arises when StopIteration is triggered in any way, but the sub-thread...

bug

我在2个3090上使用一个人的纯净语言,预估有7G的数据量,15小时的音频时间 我通过此仓库的webui加载了训练任务,参数如下,在尝试训练335后,发现loss并没有下降的情况,测试了下模型效果,部分语气下相似 对于优化训练的方法,哪里可以调整参数,已达到最优效果 训练参数 batch_size 20, enable ckpt save_every_epoch weights folder of 5 epoch pretrained_v2 ![image](https://github.com/RVC-Project/Retrieval-based-Voice-Conversion-WebUI/assets/16206043/83badd33-9e76-4273-97e8-1f57d44a8c21) 训练日志 ``` INFO:lulu-epoch:Train Epoch: 345 [83%] INFO:lulu-epoch:[30000, 9.577890768671308e-05] INFO:lulu-epoch:loss_disc=3.664, loss_gen=3.372, loss_fm=10.585,loss_mel=16.342, loss_kl=1.011 INFO:lulu-epoch:Saving model...

## Add CI/CD build process, which includes the following tasks: compiling and building, releasing to test PyPI, testing with pytest, releasing to PyPI, and publishing a Release Tag. ### your...

size:XL

#### Requirements Does Ollama currently plan to support multiple acceleration frameworks? We understand that Ollama currently leverages Llama.cpp for inference acceleration, which supports only the Llama architecture. The GLM has...

feature request

### systemRole Wants all responses to be presented as if from a character named DAN, who uses coarse language in every sentence, prefixes responses with "DAN:", and ends with an...

## Description RateLimiter: The original TpmRpmLLMLimiter strategy did not account for minute-based rate limiting when scheduled. The RateLimiter was introduced to ensure that the CommunityReportsExtractor could be scheduled to adhere...

## Description add notebook verb and workflow example