issues
search
yanqiangmiffy
/
InstructGLM
ChatGLM-6B 指令学习|指令数据|Instruct
MIT License
654
stars
51
forks
source link
issues
Newest
Newest
Most commented
Recently updated
Oldest
Least commented
Least recently updated
Problems in train_deepspeed.py with ZeRO stage 1|2|3
#34
zjhJOJO
opened
1 year ago
0
用tokenizer_dataset_rows.py转换自己的数据报错datasets.builder.datasetgeneraationerror
#33
cat1222
opened
1 year ago
0
使用lora训练,使用web_demo加载lora权重后,结果跟原生chatglm结果一样,lora权重没生效,这个是什么原因呢
#32
AnddyWang
opened
1 year ago
1
用BelleGroup/train_1M_CN训练后,为什么用数据集里的问题测,回答不一样
#31
czhcc
opened
1 year ago
0
调教后的逻辑能力如何?
#30
daiaji
opened
1 year ago
1
train_lora最低需要多大显存GPU可以训练?除了batch size 还有别的参数可以降低显存使用吗?
#29
twosnowman
opened
1 year ago
2
用peft加载lora后,generate时报错ValueError: 130000 is not in list,加载lora之前推理是正常的
#28
BIGPPWONG
opened
1 year ago
0
请问下train_deepspeed.py 怎么引入lora.pt
#27
AlexXx-Wu
opened
1 year ago
1
torch.distributed.elastic.multiprocessing.errors.ChildFailedError
#26
MonkeyTB
closed
1 year ago
1
ValueError: ChatGLMForConditionalGeneration does not support gradient checkpointing.
#25
deepeye
opened
1 year ago
7
预测时,torch.set_default_tensor_type(torch.cuda.HalfTensor)的问题
#24
reborm
closed
1 year ago
0
datasets.builder.InvalidConfigName: Bad characters from black list '<>:/\|?*' found in 'data/belle_data.json'. They could create issues when creating a directory for this config on Windows filesystem.
#23
deepeye
opened
1 year ago
1
RuntimeError: torch.cat(): expected a non-empty list of Tensors
#22
hrdxwandg
closed
1 year ago
7
4张32G的可以吗,作者可以用你写的其他开源数据集finetune看看效果吗,再放出转换和训练代码
#21
hangzeli08
opened
1 year ago
2
4张 12G的 3060能训练吗
#20
zlszhonglongshen
opened
1 year ago
2
运行web_demo_alpaca_lora.py报错,是单纯的显存不够嘛
#19
tianmala
opened
1 year ago
1
训练python train_lora.py的时候显示 ModuleNotFoundError: No module named 'configuration_chatglm'
#18
dragononly
opened
1 year ago
3
测试数据打不开https://huggingface.co/datasets/BelleGroup/generated_train_0.5M_CN
#17
dragononly
opened
1 year ago
2
微调2:BELLE中文指令数据的问题
#16
czhcc
opened
1 year ago
1
Lora+DeepSpeed多机多卡的问题
#15
zyds
closed
1 year ago
0
怎么能把lora参数merge回原始模型呢?
#14
AItechnology
closed
1 year ago
1
运行 finetune.py 遇到问题:OSError: /data/pretrained-chatglm-6b/ does not appear to have a file named config.json
#13
xubuvd
opened
1 year ago
1
最新update的代码中,web_demo推理时报错
#12
feyxong
opened
1 year ago
2
关于训练完成后,生成的答案总是带一些莫名奇妙的Q,A数据,真的不造是哪里出了问题,还望大佬赐教!谢谢!
#11
UMU689
opened
1 year ago
1
修改README.md中的错别字
#10
SunYanCN
closed
1 year ago
0
关于多轮对话的疑问
#9
ZeyuTeng96
opened
1 year ago
0
web_demo_belle生成结果时有大段重复的问题
#8
JiayiFu
opened
1 year ago
10
请问支持多卡吗,怎么改造?
#7
hjyMM2018
closed
1 year ago
2
RuntimeError: expected scalar type Half but found Float
#6
fulQuan
closed
1 year ago
1
ValueError: 150000 is not in list
#5
superhg
opened
1 year ago
5
ValueError: Please specify `target_modules` in `peft_config`
#4
MrInouye
closed
1 year ago
3
24G显存的3090可以训练吗?
#3
franklyd
closed
1 year ago
1
请问有训练好的权重可以下载吗?
#2
EagleChen
closed
1 year ago
1
基于原始chatglm-6b训练效果好还是基于alpaca的lora继续微调效果好呢?
#1
suc16
closed
1 year ago
4