-
Notifications
You must be signed in to change notification settings - Fork 4.5k
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
Unsloth loss 似乎不太正常地低
pending
This problem is yet to be addressed
#6440
opened Dec 25, 2024 by
cyc00518
1 task done
QVQ-72B-Preview Support
pending
This problem is yet to be addressed
#6439
opened Dec 25, 2024 by
ZB052-A
加载完模型刚刚开始训练时,显示 torch.utils.checkpoint.CheckpointError: torch.utils.checkpoint: Recomputed values for the following tensors have different metadata than during the forward pass.
pending
This problem is yet to be addressed
#6438
opened Dec 25, 2024 by
yedingda
1 task done
求问:1.自己的模型是在llama3-8b的基础上训练的,想要继续训练自己的模型,提示模版应该选llama3还是default? 模型导出时应该选哪个提示模版?
pending
This problem is yet to be addressed
#6436
opened Dec 25, 2024 by
QinziLi-nanan
1 task done
the Difference of Memory Usage between Llama-factory and Transformer Trainer
pending
This problem is yet to be addressed
#6435
opened Dec 24, 2024 by
Znull-1220
1 task done
lora微调Mamba-Codestral-7B-v0.1出现了问题
pending
This problem is yet to be addressed
#6434
opened Dec 24, 2024 by
tongzeliang
1 task done
寒武纪:咱们是否能支持寒武纪?
pending
This problem is yet to be addressed
#6429
opened Dec 24, 2024 by
y149604146
1 task done
Ascend NPU 910B3采用deepspeed引擎训练,Q1:未调用NPU,Q2:NPU健康状态是否影响训练。
npu
This problem is related to NPU devices
pending
This problem is yet to be addressed
#6428
opened Dec 24, 2024 by
Lexlum
1 task done
奖励模型能否不是一个model,而是一个自己定义的函数
pending
This problem is yet to be addressed
#6423
opened Dec 23, 2024 by
cdhx
1 task done
Tokenizer does not derive the newer config
pending
This problem is yet to be addressed
#6415
opened Dec 21, 2024 by
xiaosu-zhu
1 task done
Questions about resuming training form ckpt
pending
This problem is yet to be addressed
#6414
opened Dec 21, 2024 by
Jiawei-Guo
1 task done
Why Speed per iteration slower when dataset is large
pending
This problem is yet to be addressed
#6410
opened Dec 20, 2024 by
coding2debug
1 task done
sft have bug while lora run successfully
pending
This problem is yet to be addressed
#6405
opened Dec 20, 2024 by
TimeFlysLeo
1 task done
How to reproduce the paper results?
pending
This problem is yet to be addressed
#6387
opened Dec 19, 2024 by
StiphyJay
1 task done
LLaMA-Factory对话预期之外存在问题
pending
This problem is yet to be addressed
#6386
opened Dec 19, 2024 by
3237522375
1 task done
如何把我训练的奖励模型放到ppo的工作管线里
pending
This problem is yet to be addressed
#6385
opened Dec 19, 2024 by
chcoo
1 task done
LLava Series (7B, 14B) freeze_vision_tower=false bug
pending
This problem is yet to be addressed
#6376
opened Dec 18, 2024 by
xirui-li
1 task done
训练PPO出现问题:ValueError: Target module ModuleDict( (default): Identity() (reward): Identity() ) is not supported. Currently, only the following modules are supported: This problem is yet to be addressed
torch.nn.Linear
, torch.nn.Embedding
, torch.nn.Conv2d
, transformers.pytorch_utils.Conv1D
.
pending
#6373
opened Dec 18, 2024 by
sunzjz
1 task done
多节点使用zero3速度很慢
pending
This problem is yet to be addressed
#6372
opened Dec 18, 2024 by
HelloWorld506
1 task done
webui加载qwen2-vl-7b进行chat报错
pending
This problem is yet to be addressed
#6371
opened Dec 18, 2024 by
laoqiongsuan
1 task done
Can you support fast resume with streaming option?
pending
This problem is yet to be addressed
#6352
opened Dec 16, 2024 by
JonghwanMun
1 task done
如何在 llama factory 中使用多个不同难度数据集实现课程式训练(Curriculum Learning)
pending
This problem is yet to be addressed
#6350
opened Dec 16, 2024 by
enerai
1 task done
Support phi-4 released by msft on 2024-12-16
pending
This problem is yet to be addressed
#6346
opened Dec 16, 2024 by
yx-lamini
1 task done
Previous Next
ProTip!
Find all open issues with in progress development work with linked:pr.