-
Notifications
You must be signed in to change notification settings - Fork 4.2k
Issues: hiyouga/LLaMA-Factory
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
Author
Label
Projects
Milestones
Assignee
Sort
Issues list
连续做增量预训练,每次的loss都是从6.2左右开始降低到3.9
pending
This problem is yet to be addressed
#6035
opened Nov 14, 2024 by
tszslovewanpu
1 task done
请问llama3.2 vision版本的sft模块近期会发布吗?
pending
This problem is yet to be addressed
#6033
opened Nov 14, 2024 by
bupianlizhugui
1 task done
部署 最新的0.9.0 版本,看不到qwen2以上的系列,如何处理
pending
This problem is yet to be addressed
#6031
opened Nov 14, 2024 by
wangyingle666
1 task done
训练时使用的QLoRA 4rank,进行cuda模型合并导出时出现,KeyError: 'base_model.model.model.model.layers.14.mlp.down_proj'
pending
This problem is yet to be addressed
#6030
opened Nov 14, 2024 by
xiaoheiyue
1 task done
LLaMA-Factory在昇腾Ascend910_c单机八卡运行Qwen2.5-3B模型的全参训练出现npu oom存储超载的问题
npu
This problem is related to NPU devices
pending
This problem is yet to be addressed
#6028
opened Nov 14, 2024 by
beinyann
1 task done
请问给其他模型如何配置正确的template模板
pending
This problem is yet to be addressed
#6026
opened Nov 14, 2024 by
CLL112
1 task done
经过sft或者dpo后,合并lora后导出的gguf,在ollama上回答效果不佳
pending
This problem is yet to be addressed
#6020
opened Nov 13, 2024 by
NeilL0412
1 task done
Customized Reward Model
pending
This problem is yet to be addressed
#6017
opened Nov 13, 2024 by
yuepengs
1 task done
使用原生DDP训练后,15min后显示sockettimeout。
pending
This problem is yet to be addressed
#6015
opened Nov 13, 2024 by
liuhash
1 task done
多卡4090 DeepSpeed ZeRO-3 LoRA 微调 Qwen2.5-14B-Instruct 显存占用超出预期
pending
This problem is yet to be addressed
#6011
opened Nov 13, 2024 by
Lanture1064
1 task done
Serious misalignment in LLaVA implementation
pending
This problem is yet to be addressed
#6008
opened Nov 12, 2024 by
oncleJules
1 task done
Qwen2.5-coder-7B微调运行一段时间后崩溃
pending
This problem is yet to be addressed
#6007
opened Nov 12, 2024 by
xierbut
1 task done
Unsloth FileExistsError: [WinError 183]
pending
This problem is yet to be addressed
#6005
opened Nov 12, 2024 by
rogersohandsome
1 task done
创建的public url一直在加载
pending
This problem is yet to be addressed
#6004
opened Nov 12, 2024 by
Wang-YLi
1 task done
indices should be either on cpu or on the same device as the indexed tensor (cpu)
pending
This problem is yet to be addressed
#6003
opened Nov 12, 2024 by
sxm7078
qwen2-vl-2b全参微调后,推理出现重复输出
pending
This problem is yet to be addressed
#6001
opened Nov 12, 2024 by
hunter2009pf
1 task done
单机多卡运行报错,8卡h20
pending
This problem is yet to be addressed
#6000
opened Nov 12, 2024 by
yewenpeng
1 task done
在微调tools之后,通过api启动,返回了非json格式的arguments,但是在webui上测试是正常的
pending
This problem is yet to be addressed
#5998
opened Nov 12, 2024 by
tydtydtyd
1 task done
Qwen2-vl lora train?
solved
This problem has been already solved
#5996
opened Nov 12, 2024 by
kike-0304
1 task done
ChatGLM3-6b在阿里云上使用lora微调模型后,训练完成后,点击chat加载模型出现以西报错
pending
This problem is yet to be addressed
#5995
opened Nov 12, 2024 by
wsjqq
1 task done
After using lora to fine-tune the GLM-4 model, the chat template format is wrong
pending
This problem is yet to be addressed
#5986
opened Nov 11, 2024 by
sixsixcoder
1 task done
quantization_bit: 4出现raise NotImplementedError( [rank4]: NotImplementedError: Cannot copy out of meta tensor; no data! Please use torch.nn.Module.to_empty() instead of torch.nn.Module.to() when moving module from meta to a different device.
pending
This problem is yet to be addressed
#5985
opened Nov 10, 2024 by
Minchic
1 task done
qwen2-vl全量微调时解冻Vit和merger/connect
pending
This problem is yet to be addressed
#5981
opened Nov 10, 2024 by
Michael4933
1 task done
input getting stripped while finetuning qwen2-vl
pending
This problem is yet to be addressed
#5980
opened Nov 9, 2024 by
grvsh02
1 task done
Previous Next
ProTip!
Exclude everything labeled
bug
with -label:bug.