Ecosyste.ms: Issues
An open API service for providing issue and pull request metadata for open source projects.
GitHub / OFA-Sys/Chinese-CLIP issues and pull requests
#376 - 关于对已初始化好的模型进行并发抽取图像向量问题
Issue -
State: open - Opened by TheRainWalkStreet 11 days ago
#375 - 训练自己的数据集ACC一直很低
Issue -
State: open - Opened by cuppersd about 2 months ago
#374 - 文本行为特征定位坐标
Issue -
State: open - Opened by shengruicai 2 months ago
- 1 comment
#373 - Add tqdm to finetune process
Issue -
State: open - Opened by rujiangyouyou 3 months ago
#372 - 怎样批量encode图片?
Issue -
State: open - Opened by Royhuiy 3 months ago
#372 - 怎样批量encode图片?
Issue -
State: open - Opened by Royhuiy 3 months ago
#371 - 关于调用cn_clip进行特征提取报错格式错误的问题
Issue -
State: open - Opened by Seing-yu 3 months ago
#370 - 报错:ModuleNotFoundError: No module named 'flash_attn.flash_attention'
Issue -
State: open - Opened by Charimanhua 3 months ago
- 3 comments
#369 - Insecure called of torch.load in the function of load_from_name
Issue -
State: open - Opened by alpha731 3 months ago
#368 - 关于clip score的计算方式
Issue -
State: open - Opened by thbupt 4 months ago
#367 - 单机四卡训练报错
Issue -
State: closed - Opened by ctgushiwei 4 months ago
#366 - 构建数据的流程能不能更具体点,或者提供原始转tsv和jsonl的脚本
Issue -
State: closed - Opened by ctgushiwei 4 months ago
#365 - No model called 'RoBERTa-wwm-Base'
Issue -
State: open - Opened by ranck626 4 months ago
#364 - ncnn
Issue -
State: open - Opened by Heiheicxx 4 months ago
#363 - Fluctuating embedding value
Issue -
State: open - Opened by blizzardwj 4 months ago
#362 - 模型微调异常
Issue -
State: open - Opened by darkprices 4 months ago
- 2 comments
#361 - The protocol error while running the build the build_lmdb_dataset.py
Issue -
State: open - Opened by darkprices 5 months ago
#360 - 微调好的finetune模型,做图片分类,用api方法不行
Issue -
State: open - Opened by xiaoAugenstern 5 months ago
#359 - welcome to clip.cpp project, support openai clip and chinese clip model, quantization type(q4,q8_0,f16,f32), metal gpu and cuda speed
Issue -
State: open - Opened by yysu-888 5 months ago
#358 - 保存微调后模型,效果没有微调过程中好
Issue -
State: open - Opened by wangly1998 5 months ago
#357 - MUGE微调报错:ERROR:torch.distributed.elastic.multiprocessing.api:failed (exitcode: -7)
Issue -
State: open - Opened by Guan-Y 5 months ago
#356 - 不兼容gguf格式?
Issue -
State: open - Opened by aiportor 5 months ago
- 1 comment
#355 - cuDNN mismatch tensorRT(onnx,tensorRT)
Issue -
State: open - Opened by ranck626 5 months ago
#354 - 有简单的办法生成热力图吗,很需要
Issue -
State: open - Opened by saymyname77 6 months ago
#353 - 支持英文吗?
Issue -
State: open - Opened by Phoenix724 6 months ago
#352 - 多卡微调多线程报错:Exception in thread
Issue -
State: open - Opened by fanshuaiyao 6 months ago
- 6 comments
#351 - get_image_features 段错误(吐核)
Issue -
State: open - Opened by yanqiangmiffy 6 months ago
#350 - onnx部署没问题,tensorrt部署有问题
Issue -
State: open - Opened by paulpaul91 6 months ago
- 3 comments
#349 - train script using torchrun for torch version >= 2.0.0
Pull Request -
State: open - Opened by SuperAZHE 6 months ago
#348 - 可以在fiftyone中实现文搜图跨模态检索吗
Issue -
State: open - Opened by seasnakes 6 months ago
#347 - AttributeError: 'ChineseCLIPConfig' object has no attribute 'hidden_size'
Issue -
State: open - Opened by Manjiz 6 months ago
#346 - 大佬可以训练一个像jina-clip-v1的中文版这样的话,以后做搜索或是 构建多模态 RAG 应用 就牛了~~
Issue -
State: open - Opened by gg22mm 6 months ago
- 1 comment
#345 - 使用下列代码,如何加载自己训练的 'ViT-L-14'模型指定路径的.pt文件。(比如加载Epoch为2的./data/model/epoch2.pt模型)
Issue -
State: open - Opened by laitianan 6 months ago
- 8 comments
#344 - 翻译有点糟糕
Issue -
State: open - Opened by ccl-private 6 months ago
#343 - The context length to use; all baseline models use 52 as the context length
Issue -
State: open - Opened by qiuchen001 7 months ago
#342 - 通过pip安装的cn_clip库方法load_from_name不能加载本地模型
Issue -
State: open - Opened by qiuchen001 7 months ago
- 1 comment
#341 - 请教下,提取图文特征json数据后,应该怎么配合clip-retrieval使用搭建自己的检索demo?
Issue -
State: open - Opened by JingMo 7 months ago
#340 - Update dataset_transform.py
Pull Request -
State: closed - Opened by youguohui 7 months ago
#339 - Add files via upload
Pull Request -
State: closed - Opened by youguohui 7 months ago
- 1 comment
#338 - 微调太复杂 简单的东西搞得这么复杂
Issue -
State: open - Opened by wangsheng1991 7 months ago
#337 - 进行 rn50 model-arch 转 onnx 的时候会报如下错误
Issue -
State: open - Opened by pisces365 7 months ago
- 1 comment
#336 - Is there a CLIP model that is smaller than ResNet50 and RBT3?
Issue -
State: open - Opened by guanfuchen 7 months ago
#335 - 运行,报错了,bash run_scripts/muge_finetune_vit-b-16_rbt-base.sh ${DATAPATH}
Issue -
State: open - Opened by Renjunhua30 7 months ago
#334 - 模型加载问题
Issue -
State: open - Opened by dayuanjie 7 months ago
#333 - 使用MUGE微调时,训练日志里面的Image2Text Acc与评测时的R@1的召回指标不一致?
Issue -
State: open - Opened by dulibubai 7 months ago
#332 - COCO-CN数据集申请
Issue -
State: open - Opened by zlw1113 7 months ago
#331 - 能否通过文本向量化中加入指令优化图片召回率
Issue -
State: open - Opened by Amphetaminewei 7 months ago
#330 - CN-CLIPViT-B/16模型在输入与向量化图片无关的输入时,如!@#¥,仍能有结果匹配
Issue -
State: open - Opened by wangyan828 7 months ago
- 1 comment
#329 - “参数命名不同”:Huggingface上的CLIP模型和GITHUB上的 CLIP模型 参数命名不同
Issue -
State: open - Opened by ranck626 7 months ago
#328 - onnx模型转换报错
Issue -
State: closed - Opened by guanxiaoguan 7 months ago
- 1 comment
#327 - 各位大佬,专家,请问该模型能否在Windows系统下进行运行呢?欢迎来交流
Issue -
State: open - Opened by LJK12581 8 months ago
#326 - Question: gradient accumulate
Issue -
State: open - Opened by zwhus 8 months ago
#325 - how to continue training with self-data? or Full Parameter Fine-Tuning
Issue -
State: open - Opened by StatYuDoc 8 months ago
- 1 comment
#324 - 手动修改ChineseCLIPVisionModel to ChineseCLIPVisionModelWithProjection 失败。
Issue -
State: open - Opened by ranck626 8 months ago
- 1 comment
#323 - ChineseCLIPVisionModel in huggingface can used?
Issue -
State: open - Opened by ranck626 8 months ago
- 3 comments
#322 - ViT-H/14训练脚本、部署能够提供吗?(使用ViT-B/16更改为ViT-H/14脚本,训练起来,模型各个方面参数都是下降的,但是部署成tensorRT,用图片预测,效果堪忧)
Issue -
State: open - Opened by laitianan 8 months ago
- 3 comments
#321 - 想要验证Flickr30K-CN Retrieval数据集应该使用那个脚本?
Issue -
State: open - Opened by Dagoli 8 months ago
#320 - 模型finetune这部操作应该使用什么软件?请各位大佬回复
Issue -
State: open - Opened by LJK12581 8 months ago
- 2 comments
#318 - fine tune报线程错误 求指导
Issue -
State: open - Opened by lll1024 9 months ago
- 3 comments
#317 - 请问在实现image caption上有什么建议吗
Issue -
State: open - Opened by luoxinggyyy 9 months ago
#316 - 请问有开源Chinese-Clip-Vit-G的计划么
Issue -
State: open - Opened by chengjianjie 9 months ago
#315 - 执行bash run_scripts/muge_finetune_vit-b-16_rbt-base.sh /home/liujiawei/datapath报错
Issue -
State: closed - Opened by lll1024 9 months ago
- 7 comments
#314 - 这个可以在华为的服务器部署吗?
Issue -
State: open - Opened by intjun 9 months ago
- 1 comment
#313 - 转换模型格式时的config.json文件是要自己提供吗
Issue -
State: open - Opened by Yang-bug-star 9 months ago
#312 - load_from_name 加入 flash-attn 支持
Issue -
State: open - Opened by ZechengLi19 10 months ago
- 4 comments
#311 - 请问训练脚本会在当验证集上性能下降时停止训练吗,还是说只是会汇报下验证集上的性能但训练继续
Issue -
State: open - Opened by Yang-bug-star 10 months ago
- 1 comment
#310 - 请问分词器有没有c++版本的实现?
Issue -
State: open - Opened by liumingzhu6060 10 months ago
- 3 comments
#309 - fb16速度比fb32要慢
Issue -
State: open - Opened by xxllp 10 months ago
#308 - 文搜图的原理
Issue -
State: open - Opened by sx-zyf-123 10 months ago
#307 - 请问怎么调用这个模型?怎么提供api服务?
Issue -
State: open - Opened by gubinjie 10 months ago
#306 - 请问demo用的base模型跟本项目给的CN-CLIP(ViT-B/16)是同一个模型吗?跟我本地的测试结果不太一样
Issue -
State: open - Opened by xiuxiuxius 10 months ago
#305 - 执行finetune流程时报错
Issue -
State: open - Opened by Cupies 10 months ago
- 2 comments
#304 - Delete
Issue -
State: closed - Opened by ChesonHuang 10 months ago
#303 - 请问在线demo中的文搜图或图搜图应用,对返回结果的逻辑处理是怎么样的呢?我设置了返回50张,但是有的文本只返回了32张。请问你们是对结果进行了阈值过滤吗,阈值设置的多少呢?用于检索的数据有哪些呢?
Issue -
State: open - Opened by xiuxiuxius 10 months ago
#302 - 微调clip_cn_vit-l-14-336 报错
Issue -
State: closed - Opened by xxllp 10 months ago
- 2 comments
#301 - finetune时报错,且Traceback疑似被截断,无法定位出错线程
Issue -
State: open - Opened by wrtppp 10 months ago
- 3 comments
#300 - 在finetune时报错KeyError: 'optimizer'
Issue -
State: open - Opened by wrtppp 10 months ago
- 2 comments
#299 - 无进行任何微调,直接用模型对总的数据集进行测试R@5值为50多,把总数据集切割为train、test和valid后直接测试R@5为10多
Issue -
State: open - Opened by EasonTuT 10 months ago
#298 - LOSS:nan 微调时LOSS异常
Issue -
State: open - Opened by wangbenchi 10 months ago
- 14 comments
#297 - 关于ACC和R@5的问题
Issue -
State: open - Opened by EasonTuT 10 months ago
- 14 comments
#296 - Downlodaded issue
Issue -
State: open - Opened by chenpinganshiiwo 10 months ago
- 1 comment
#295 - 为什么使用同样的词和图片得到的结果不一致?
Issue -
State: open - Opened by learning233 10 months ago
- 2 comments
#294 - text描述
Issue -
State: open - Opened by erlan-11 11 months ago
#293 - 使用RN50预训练模型和flick30k后得到的权重文件特别大
Issue -
State: open - Opened by TC120 11 months ago
- 2 comments
#292 - 图到文检索Recall计算出现The evaluation failed: image_ids
Issue -
State: closed - Opened by IF-chan 11 months ago
#291 - 这个问题太折磨了,找不到解决方法,有没有大神看一下
Issue -
State: open - Opened by iWangTing 11 months ago
- 14 comments
#290 - 运行Recall计算评测脚本时出现{text_id:: command not found
Issue -
State: open - Opened by IF-chan 11 months ago
#289 - 图文特征融合
Issue -
State: open - Opened by Amore-HDU 11 months ago
#288 - 好像包里少项了,按路径找过去确实没找到这东西
Issue -
State: open - Opened by 2398274582 11 months ago
- 1 comment
#287 - 关于对导入LMDB数据集在微调的时候出现并行的问题
Issue -
State: open - Opened by jakeallen123 11 months ago
- 1 comment
#286 - main.py: error: unrecognized arguments: --accum_freq=1
Issue -
State: open - Opened by iWangTing 11 months ago
- 2 comments
#285 - AttributeError: 'Namespace' object has no attribute 'use_flash_attention'
Issue -
State: open - Opened by iWangTing 11 months ago
- 1 comment
#284 - image_b64为空
Issue -
State: open - Opened by erlan-11 11 months ago
- 7 comments
#283 - 图文特征提取源码bug
Issue -
State: open - Opened by liudehua7539 11 months ago
- 1 comment
#282 - 在GPU 推理报错 Segmentation fault
Issue -
State: open - Opened by shiqwang 11 months ago
- 1 comment
#281 - 使用build_lmdb_dataset.py脚本转换lmdb文件的问题
Issue -
State: open - Opened by byraid218 11 months ago
- 13 comments
#280 - 请问在三个图文检索数据集上微调时,有使用两阶段微调的方式吗?
Issue -
State: open - Opened by xiuxiuxius 11 months ago
- 6 comments
#279 - windows单机单卡训练报错RuntimeError: Distributed package doesn't have NCCL built in
Issue -
State: open - Opened by bowenzc 11 months ago
- 2 comments
#278 - 关于执行sh脚本文件的报错 RuntimeError: NCCL error in: ../torch/lib/c10d/ProcessGroupNCCL.cpp:911, unhandled system error, NCCL version 2.7.8
Issue -
State: open - Opened by Fumon554 11 months ago