issues
search
hiyouga
/
LLaMA-Factory
Unify Efficient Fine-Tuning of 100+ LLMs
Apache License 2.0
25.26k
stars
3.13k
forks
source link
issues
Newest
Newest
Most commented
Recently updated
Oldest
Least commented
Least recently updated
CUDA out of memory | QLORA | Llama 3 70B | 4 * NVIDIA A10G 24 Gb
#4559
russellorv
closed
2 days ago
5
llama board测评模型效果差
#4558
xiao-liya
closed
2 days ago
0
Get "RuntimeError: 'weight' must be 2-D" Error when finetuning llama3-8b using ZeRO3 and customised dataset
#4557
NeWive
closed
2 days ago
2
[bug] pissa init脚本报错
#4556
letterk
closed
2 days ago
1
API_PORT=8000 llamafactory-cli api examples/inference/qwen2_vllm.yaml报错
#4555
frostjsy
closed
2 days ago
1
微调llama3-8b的时候,eval_loss不断上升,考虑到了使用多个数据集混合,但还是没有效果,应该怎么解决?
#4554
MemoryOldTime
closed
2 days ago
0
llamafactory-cli支持输出向量吗?
#4553
frostjsy
closed
3 days ago
1
Help! 训练结束加载大模型,使用训练数据集中的问题向大模型提问,答案不是训练数据集中的答案。训练耗时不足2分钟,怀疑训练失败,请大佬们帮忙诊断一下!!
#4552
muliu
closed
3 days ago
4
llamafactory-cli api examples/inference/qwen2_vllm.yaml报错
#4551
frostjsy
closed
3 days ago
0
Llama3 微调文本分类数据集准备的格式,以及微调后模型回答输出非标签内容
#4550
git3210
closed
3 days ago
1
【问题】为什么要把可训练参数精度强行转换为全精度?
#4549
LaniakeaS
closed
2 days ago
7
请教多轮工具调用的数据格式
#4548
lwj2001
closed
2 days ago
0
lora微调Qwen-14b-chat后导入模型报错
#4547
ntz2000
closed
3 days ago
1
出一个中文版的操作手册
#4546
countree
closed
3 days ago
1
vllm部署freeze微调后的qwen2-57b-instruct报错
#4545
hexixiang
closed
3 days ago
1
fix docker-compose path
#4544
MengqingCao
closed
3 days ago
1
how to authenticate to access huggingface
#4543
ares0027
closed
3 days ago
1
什么时候可以支持glm4v等多模态模型微调呀
#4542
yang182
closed
3 days ago
1
scripts/pissa_init.py to initialize PiSSA for a quantized model.
#4463
Katehuuh
closed
3 days ago
1
对sft阶段的数据进行packing之后,同一条训练数据内的指令之间是否会相互影响?
#4462
Zheng-Jay
closed
3 days ago
0
support flash-attn in Dockerfile
#4461
hzhaoy
closed
1 day ago
1
## feature request ## 支持 ZeRO3 infinity
#4460
Xingxiangrui
closed
4 days ago
1
deepspeed zero3 出现 training_eval_loss 图为空白
#4459
ycjcl868
closed
4 days ago
1
PPO 跑example例子报错:value should be one of int, float, str, bool, or torch.Tensor
#4458
xudong2019
closed
2 days ago
1
关于基座模型和对话模型的疑问
#4457
VirgilG72
closed
2 days ago
0
docker容器内没有example和data文件
#4456
yaya159456
closed
4 days ago
0
最新代码中没有llamafactory-cli ,怎么合并权重
#4455
Liufeiran123
closed
4 days ago
2
lora微调后的glm4模型不生成回答
#4454
RyanCcc114
closed
2 days ago
0
8*A800 80G lora训练qwen2-72B模型 内存占用异常
#4453
999wwx
closed
2 days ago
2
LoRA微调和全参微调的时候总是会出现过拟合,在无法提高数据集大小的情况下,应该如何解决这个问题呢
#4452
MemoryOldTime
closed
4 days ago
2
如何指定已划分好的训练集和验证集?
#4451
Owen1u
closed
4 days ago
1
Memory Error during tokenization while fine tuning LLava1.5-7B-Chat more than 8000 images
#4450
Hassaan68
closed
4 days ago
2
qlora微调Qwen2-57B。使用单卡A6000显存占用40G,使用双卡A6000则是两张卡各占40G显存,请问是什么原因?
#4447
PhysicianHOYA
closed
4 days ago
0
Bug Fix: `off` is parsed as `False` in yaml file
#4446
stceum
closed
4 days ago
0
auto-label npu issue
#4445
MengqingCao
closed
4 days ago
0
stop word of template of qwen
#4444
HackGiter
closed
4 days ago
1
请问支持 early stopping 吗?
#4443
laozhai507
closed
4 days ago
0
请问改工程可以用来glm4的增量预训练吗
#4442
gyh123wqe
closed
4 days ago
1
训练glm4报错:RuntimeError when using flash attention with 8-bit quantization,同样的参数训llama3则没问题
#4441
fst813
closed
4 days ago
1
How to pre-train Llava1.5 from vicuna1.5?
#4440
yuzhms
closed
4 days ago
1
预训练方式lora微调Qwen2 base模型,是否需要添加template
#4439
Shame-fight
closed
4 days ago
1
单机多卡微调glm4-9B设置max_grad_norm=1,但是仍然出现了梯度爆炸的问题
#4438
KATHERINE0511
closed
4 days ago
1
fsdp+fp16 全参数微调是否支持呢
#4437
zhangfan-algo
closed
5 days ago
1
Low MMLU of llama2
#4436
Zkli-hub
closed
4 days ago
1
1*8 H20微调qwen2-72b-instruct,保存模型失败
#4435
liulfy
closed
4 days ago
2
沒有過擬和的狀況,但是loss到一個點後就會難以下降,並且推理對話會有重覆內容
#4434
ahsbdcpu
closed
4 days ago
1
为什么训练不支持predict_with_generate,有什么原因吗? 需要计算metrics
#4433
jinwik
closed
5 days ago
1
Error during inference LLaMA2 + LoRA: RuntimeError: expected mat1 and mat2 to have the same dtype, but got: c10::Half != float
#4432
WJMacro
closed
4 days ago
1
glm-4-9b-chat在使用web_demo进行推理时报错
#4429
lanhaibin
closed
4 days ago
0
RuntimeError: disagreement between rank0 and rank6 while fully supervised finetuning deepseek v2 lite during the evaluation phase.
#4428
falcon-xu
closed
4 days ago
3
Previous
Next