Ecosyste.ms: Issues
An open API service for providing issue and pull request metadata for open source projects.
GitHub / modelscope/ms-swift issues and pull requests
#1627 - torch._C._LinAlgError: linalg.cholesky 即使讲quant_n_samples提到了2048依然报错
Issue -
State: open - Opened by ff1Zzd about 1 month ago
- 1 comment
#1626 - deepseek v2推理报错
Issue -
State: closed - Opened by zhangfan-algo about 1 month ago
- 1 comment
#1625 - refactor internvl2
Pull Request -
State: closed - Opened by Jintao-Huang about 1 month ago
#1617 - SWIFT 2.4 TO DO LIST
Issue -
State: open - Opened by tastelikefeet about 1 month ago
- 10 comments
#1616 - 加载本地下好的sd模型进行微调时,仍然需要去网页端下载模型
Issue -
State: open - Opened by jamesbondzhou about 1 month ago
- 1 comment
#1613 - Best Practices for Inference and Fine-Tuning with MiniCPM-V 2.6
Issue -
State: open - Opened by Jintao-Huang about 1 month ago
- 75 comments
Labels: good first issue
#1611 - GLM4V运行时报错
Issue -
State: open - Opened by det-tu about 1 month ago
- 5 comments
#1605 - 为什么增大显卡数量无法加快训练速度?
Issue -
State: closed - Opened by jhrsya about 1 month ago
- 3 comments
#1602 - 目前支持minicpm_v_v2_5_chat的多卡推理吗?
Issue -
State: closed - Opened by dengruoqing about 1 month ago
- 3 comments
#1594 - New multimodal framework
Pull Request -
State: closed - Opened by Jintao-Huang about 2 months ago
#1589 - 魔搭NPU训练部署交流群
Issue -
State: open - Opened by Jintao-Huang about 2 months ago
- 1 comment
Labels: good first issue, npu
#1577 - 全量微调full断点续训(resume_from_checkpoint)问题
Issue -
State: closed - Opened by jhrsya about 2 months ago
- 6 comments
#1574 - DPO使用deepspeed出现报错
Issue -
State: open - Opened by badmic about 2 months ago
- 1 comment
#1570 - 微调, deepspeed出现报错
Issue -
State: open - Opened by badmic about 2 months ago
- 8 comments
#1549 - DDP in DPO tuning on MLLM
Issue -
State: open - Opened by yepzhang about 2 months ago
- 2 comments
#1547 - GLM 4V 9B RuntimeError: Expected all tensors to be on the same device, but found at least two devices
Issue -
State: closed - Opened by nico2rdj about 2 months ago
- 2 comments
#1532 - InternVL2的grounding任务自定义数据集如何一个prompt里融合多个<box>或<ref-object>
Issue -
State: open - Opened by Harley-Sun about 2 months ago
- 3 comments
#1528 - 关于推理时获得logits结果
Issue -
State: closed - Opened by Starfulllll about 2 months ago
- 3 comments
Labels: enhancement
#1528 - 关于推理时获得logits结果
Issue -
State: closed - Opened by Starfulllll about 2 months ago
- 3 comments
Labels: enhancement
#1520 - 微调llama3_1系列模型出现报错
Issue -
State: closed - Opened by badmic about 2 months ago
- 5 comments
#1507 - failed (exitcode: -11) local_rank: 5 (pid: 11514) of binary: /home/jovyan/data-ws-enr/zconda/envs/swift_ft/bin/python
Issue -
State: open - Opened by shyzzz521 about 2 months ago
- 8 comments
Labels: multi-node
#1504 - swift 量化多模态大模型internvl2-26B,报错
Issue -
State: open - Opened by wangdong1992 about 2 months ago
- 8 comments
Labels: enhancement
#1502 - 用改框架微调GLM-4V-9B模型,存在重复输出和任务无法拟合的问题。
Issue -
State: closed - Opened by FanWan about 2 months ago
- 3 comments
#1486 - 评测Internvl76B时,完成了部署但是没有推理
Issue -
State: closed - Opened by ZhiyuYUE about 2 months ago
- 1 comment
#1484 - The deployment of llama3.1-405b-instruct
Issue -
State: closed - Opened by Jintao-Huang about 2 months ago
- 4 comments
Labels: good first issue
#1460 - 微调InternLM-Xcomposer2d5时报错
Issue -
State: closed - Opened by bonre about 2 months ago
- 11 comments
#1428 - 现在支持vllm的多LoRA部署吗?
Issue -
State: closed - Opened by jidechao 2 months ago
- 2 comments
#1389 - QwenVL Expected all tensors to be on the same device
Issue -
State: closed - Opened by wellhowtosay 2 months ago
- 8 comments
#1387 - AttributeError: module 'inspect' has no attribute 'getargspec'. Did you mean: 'getargs'?
Issue -
State: closed - Opened by wellhowtosay 2 months ago
- 6 comments
#1358 - 脚本使用问题
Issue -
State: closed - Opened by taiji01 2 months ago
- 5 comments
#1286 - minicpm-v-v2_5-chat 微调vpm显存溢出
Issue -
State: open - Opened by lyc728 3 months ago
- 7 comments
Labels: bug
#1281 - Internvl 微调后的模型,python代码在哪里加载自己模型的路径
Issue -
State: open - Opened by LiJunY 3 months ago
- 4 comments
#1266 - Not working with python virtualenv
Issue -
State: open - Opened by heinrichI 3 months ago
- 2 comments
#1243 - RuntimeError: Sync:torch_npu/csrc/framework/OpCommand.cpp:190 NPU error, error code is 507015 昇腾显卡910B qwen1.5,2.0微调报错
Issue -
State: closed - Opened by aoyinke 3 months ago
- 7 comments
Labels: npu
#1222 - 多机多卡训练出现问题
Issue -
State: open - Opened by Jamly7 3 months ago
- 12 comments
Labels: multi-node
#1219 - 更新完modeling_chatglm.py后出现bug
Issue -
State: closed - Opened by chensongcan 3 months ago
- 13 comments
#1193 - Fix dataset concatenation
Pull Request -
State: closed - Opened by tastelikefeet 3 months ago
- 1 comment
#1185 - glm4v微调到一半会显示网络问题,请问有其他加载本地输入的方法吗
Issue -
State: closed - Opened by Nanuion 3 months ago
- 3 comments
#1179 - 训练及导出显存消耗不正常
Issue -
State: closed - Opened by tastelikefeet 3 months ago
- 1 comment
Labels: bug
#1176 - 关于"自我认知微调最佳实践"和"自定义模型和数据集"的说明优化
Issue -
State: closed - Opened by ivandgetic 3 months ago
- 3 comments
#1166 - 预训练模板错误
Issue -
State: closed - Opened by tiandiweizun 3 months ago
- 3 comments
#1163 - GLM4v使用Lora模型微调后,merge模型后运行报错,
Issue -
State: closed - Opened by hyyuan123 3 months ago
- 6 comments
#1155 - Representing results of Agent best practice with Qwen2-7b-instruct outputs unexpected <|endoftext|> and <|im_start|>
Issue -
State: open - Opened by edgeinfinity-wzt 3 months ago
Labels: question
#1151 - 量化训练qwen2-72b-instruct 的权重突入中断
Issue -
State: closed - Opened by lxb0425 3 months ago
- 1 comment
#1147 - 用户如何自己实现新的参数高效微调算法?
Issue -
State: closed - Opened by 2catycm 3 months ago
#1140 - minicpm-v 2.5 sft训练时跑验证集会额外占用10GB显存是正常现象吗?
Issue -
State: closed - Opened by KasLoot 3 months ago
- 2 comments
#1137 - 请问swift训练MiniCPM-v2.0时应该如何制作带坐标的数据集
Issue -
State: closed - Opened by daihuidai 3 months ago
- 1 comment
#1135 - Error during Finetune MiniCPM 微调MiniCPM时出错 & zero3
Issue -
State: closed - Opened by youjiaSHTU 3 months ago
- 10 comments
Labels: bug
#1133 - GLM4v LORA微调后,断点训练失败
Issue -
State: closed - Opened by lyc728 3 months ago
- 8 comments
#1128 - Process hang with futex(0x7f403c0199d0, FUTEX_WAIT, 14826, NULL
Issue -
State: open - Opened by QiMingChina 3 months ago
- 1 comment
Labels: question
#1092 - 双卡80GiB A100对Qwen2-72B-Instruct进行自我认知微调的最佳实践
Issue -
State: closed - Opened by Jintao-Huang 4 months ago
- 7 comments
Labels: good first issue
#1077 - swift训练minicpm-v如何设置grounding格式
Issue -
State: closed - Opened by daihuidai 4 months ago
- 2 comments
#1075 - 是否支持自定义lr_scheduler
Issue -
State: open - Opened by zodiacg 4 months ago
Labels: enhancement
#1074 - 关于多图微调和推理问题
Issue -
State: closed - Opened by 1028686314 4 months ago
- 3 comments
#1062 - Loss and acc drop to 0 after several steps
Issue -
State: open - Opened by MindLostGuy 4 months ago
- 1 comment
#1061 - 又测了下,您说的最新的代码是指加了if batch[0].get('pixel_values') is not None:这个判断语句吗,在batch>1的情况下,还是会出现keyerror的
Issue -
State: closed - Opened by qzDiao 4 months ago
- 1 comment
#1060 - [Help]请问MiniCPM2.5支持Web端交互推理吗?
Issue -
State: closed - Opened by Egber1t 4 months ago
- 2 comments
#1056 - 不太明白Lisa为何能够节省显存,希望能够帮忙解答
Issue -
State: closed - Opened by ymsdu2004 4 months ago
- 1 comment
#1045 - 随机初始化预训练
Issue -
State: closed - Opened by JiwenJ 4 months ago
- 1 comment
#1041 - 如何批量导出MS数据集为swift能加载的格式
Issue -
State: closed - Opened by WSC741606 4 months ago
- 1 comment
#1040 - 是否支持idm-vton模型的微调呢?
Issue -
State: closed - Opened by awzhgw 4 months ago
Labels: enhancement
#1033 - swift微调多模态大模型后,比如Intern VL 1.5,可以使用lmdeploy部署吗?
Issue -
State: closed - Opened by wangdong1992 4 months ago
- 3 comments
Labels: enhancement
#1031 - bitsandbytes was compiled without GPU support
Issue -
State: closed - Opened by stellarxxu 4 months ago
- 1 comment
#1025 - react模板放在system和user的区别?
Issue -
State: closed - Opened by nauyiahc 4 months ago
- 2 comments
#1019 - 可以支持一下InternLM2-Math-Plus-Mixtral8x22B的微调吗
Issue -
State: open - Opened by zhangfan-algo 4 months ago
- 1 comment
Labels: enhancement, pending
#1008 - 微调minicpmv2时cpu占用率超高
Issue -
State: closed - Opened by strawhatboy 4 months ago
- 1 comment
#994 - 微调量化后qwen1half-14b-chat-gptq-int8推理时向量报错RuntimeError: probability tensor contains either `inf`, `nan` or element < 0
Issue -
State: closed - Opened by AnsongLi 4 months ago
- 1 comment
#972 - 使用zero3进行多机多卡全量微调,保存的模型权重不完整
Issue -
State: closed - Opened by ultrazhl98 4 months ago
- 3 comments
#959 - ValueError: model_type: 'yi-1_5-6b' is not registered
Issue -
State: closed - Opened by jacnmm4 4 months ago
- 5 comments
#950 - 我想将模型保存到本地,怎么才能保存。同时怎么调用本地的模型?
Issue -
State: closed - Opened by AlexJJJChen 4 months ago
- 1 comment
#945 - 多节点训练报错
Issue -
State: closed - Opened by zhangfan-algo 4 months ago
- 2 comments
#943 - 训练qwen14b,前面lr一直为0
Issue -
State: closed - Opened by jhjiang10 4 months ago
- 2 comments
#942 - 希望能应用TensorRT加速训练和推理
Issue -
State: open - Opened by WSC741606 4 months ago
Labels: enhancement
#940 - 量化后的模型推理报错怎么解决
Issue -
State: closed - Opened by greatheart1000 4 months ago
- 2 comments
#938 - deepseek-vl-7b模型使用deepspeed的ZeRo3报错
Issue -
State: closed - Opened by jiujiuma 4 months ago
- 5 comments
#935 - 用qwen-7b-int4和int8进行lora微调后,微调和推理没问题,但部署后,请求报错
Issue -
State: closed - Opened by nauyiahc 4 months ago
- 9 comments
#927 - 多机多卡推理
Issue -
State: closed - Opened by yu1nY 4 months ago
- 2 comments
#923 - DPO训练的时候grad_norm出现nan值
Issue -
State: open - Opened by rtz1998 4 months ago
- 5 comments
#922 - 2.0.4之后的版本的显存使用问题
Issue -
State: open - Opened by kratorado 4 months ago
- 6 comments
#905 - Eval 模块使用时 C-Eval 进行测试时,让人不确定日志输出
Issue -
State: closed - Opened by ShuMengZ 4 months ago
- 2 comments
Labels: bug
#903 - 请教一下 有没有办法扩展c4ai-command-r-plus的上下文长度呢
Issue -
State: closed - Opened by zhangfan-algo 4 months ago
Labels: enhancement
#898 - Lora认知微调,微调成功,直接进行推理也正确。但merge之后,再推理,生成乱码,并报错“TypeError: argument 'tokens': 'NoneType' object cannot be converted to 'PyString'”
Issue -
State: closed - Opened by dage0127 4 months ago
- 2 comments
#880 - 基于上次提的问题#691,后续改进后似乎依旧不能按微调的情况回复。
Issue -
State: closed - Opened by xiaolvtongxue-zt 4 months ago
- 4 comments
#862 - Infer internlm-xcomposer2 lead to `ValueError: Input length of input_ids is 0, but `max_length` is set to -1066.`
Issue -
State: closed - Opened by piqiuni 5 months ago
- 3 comments
#860 - VRAM requirement for full sft deepseek VL 7B
Issue -
State: closed - Opened by SinanAkkoyun 5 months ago
- 3 comments
#852 - llama3-8b-instruct awq量化oom
Issue -
State: closed - Opened by Edisonwei54 5 months ago
- 2 comments
Labels: question
#851 - 使用DDP运行时显存不够,但是使用Model Parallel时可以正常finetune,耗时很大
Issue -
State: closed - Opened by AlexJJJChen 5 months ago
- 7 comments
#836 - export problem: get_model_tokenizer_with_flash_attn() got multiple values for keyword argument 'automodel_class'
Issue -
State: open - Opened by AlexJJJChen 5 months ago
- 2 comments
Labels: bug
#828 - 请教一下,多节点多卡微调,用Slurm怎样运行?
Issue -
State: closed - Opened by rationalspark 5 months ago
#825 - 使用lora微调合并权重加载模型报错
Issue -
State: closed - Opened by qianliyx 5 months ago
- 1 comment
#811 - win10训练qwen1.5-moe-A2.7B-chat-gptq-int4速度缓慢
Issue -
State: closed - Opened by catundchat 5 months ago
- 2 comments
#807 - 能否支持/v1/embeddings的api调用
Issue -
State: open - Opened by chuanSir123 5 months ago
Labels: enhancement
#794 - Langchain-Chatchat部署训练后的模型后推理异常
Issue -
State: closed - Opened by WSC741606 5 months ago
- 9 comments
#786 - Support QLoRA with HQQ quantization
Issue -
State: closed - Opened by thincal 5 months ago
- 1 comment
Labels: enhancement
#774 - 多模态 minicpmv2的训练,使用的尺寸是多大,原始minicpm v-2的训练尺寸可以很大,查看源码只看到minicpmv2的输入尺寸是448,训练和推理时对于大图会在原图上进行处理吗,还是会统一直接resize缩放到448?
Issue -
State: closed - Opened by zhly0 5 months ago
- 2 comments
#773 - ddp_timeout能否作为sft命令行参数
Issue -
State: closed - Opened by WSC741606 5 months ago
- 2 comments
Labels: enhancement
#757 - 希望可以支持自我奖励的优化方法
Issue -
State: closed - Opened by WSC741606 5 months ago
#735 - fix zsh install
Pull Request -
State: closed - Opened by tastelikefeet 5 months ago
- 1 comment
#729 - swift安装后报sft: command not found
Issue -
State: closed - Opened by yangxin60-tal 5 months ago
- 2 comments
#713 - swift 1.8. -> 2.0.0 原lora脚本报错
Issue -
State: closed - Opened by hardlipay 5 months ago