Ecosyste.ms: Issues
An open API service for providing issue and pull request metadata for open source projects.
GitHub / DLLXW/baby-llama2-chinese issues and pull requests
#84 - 关于分词器处理后的预训练语料是通过哪个代码生成的
Issue -
State: open - Opened by livevivaer 2 months ago
#84 - 关于分词器处理后的预训练语料是通过哪个代码生成的
Issue -
State: open - Opened by livevivaer 2 months ago
#83 - 想问一下 为什么做数据清洗时保存数据为Parquet格式,后面做分词时候还是用的json
Issue -
State: open - Opened by yangwenche 3 months ago
#83 - 想问一下 为什么做数据清洗时保存数据为Parquet格式,后面做分词时候还是用的json
Issue -
State: open - Opened by yangwenche 3 months ago
#82 - ChatGLMTokenizer类
Issue -
State: open - Opened by licx102359 3 months ago
- 2 comments
#82 - ChatGLMTokenizer类
Issue -
State: open - Opened by licx102359 3 months ago
- 2 comments
#81 - 预训练输入最后的切片不会导致模型的输入少一个长度吗?
Issue -
State: open - Opened by AI-Study-Han 4 months ago
#81 - 预训练输入最后的切片不会导致模型的输入少一个长度吗?
Issue -
State: open - Opened by AI-Study-Han 4 months ago
#80 - 模型的回答较长,输出结果不完整要怎么解决
Issue -
State: open - Opened by MSJeinlong 4 months ago
#80 - 模型的回答较长,输出结果不完整要怎么解决
Issue -
State: open - Opened by MSJeinlong 4 months ago
#79 - smallvocab tokenizer
Issue -
State: open - Opened by iangellove 5 months ago
#79 - smallvocab tokenizer
Issue -
State: open - Opened by iangellove 5 months ago
#78 - 请问语言模型的强化学习有可以参考的开源项目吗?
Issue -
State: open - Opened by AI-Study-Han 5 months ago
- 1 comment
#78 - 请问语言模型的强化学习有可以参考的开源项目吗?
Issue -
State: open - Opened by AI-Study-Han 5 months ago
- 1 comment
#77 - 请问大数据量怎么加载呢?
Issue -
State: open - Opened by CaesarGo 6 months ago
#77 - 请问大数据量怎么加载呢?
Issue -
State: open - Opened by CaesarGo 6 months ago
#76 - 请问哪步加的 Positional embeddings
Issue -
State: closed - Opened by buhe 6 months ago
- 1 comment
#76 - 请问哪步加的 Positional embeddings
Issue -
State: closed - Opened by buhe 6 months ago
- 1 comment
#75 - chatglm_tokenizer 模块是在哪个软件包中?
Issue -
State: open - Opened by PANASV 6 months ago
- 2 comments
#75 - chatglm_tokenizer 模块是在哪个软件包中?
Issue -
State: open - Opened by PANASV 6 months ago
- 2 comments
#74 - 预训练阶段,每条训练样本混杂着不同的句子(不同句子用<eos>隔开)
Issue -
State: open - Opened by Itochiee 6 months ago
#74 - 预训练阶段,每条训练样本混杂着不同的句子(不同句子用<eos>隔开)
Issue -
State: open - Opened by Itochiee 6 months ago
#73 - 请问在处理微调数据集时为何要限制文本长度?
Issue -
State: open - Opened by jzzzf 6 months ago
- 1 comment
#73 - 请问在处理微调数据集时为何要限制文本长度?
Issue -
State: open - Opened by jzzzf 6 months ago
- 1 comment
#72 - 作者,这个项目支持断点续训嘛
Issue -
State: open - Opened by 1737686924 7 months ago
- 2 comments
#72 - 作者,这个项目支持断点续训嘛
Issue -
State: open - Opened by 1737686924 7 months ago
- 2 comments
#71 - 请问支持tensorrt llm部署吗
Issue -
State: open - Opened by Ss-shuang123 7 months ago
#71 - 请问支持tensorrt llm部署吗
Issue -
State: open - Opened by Ss-shuang123 7 months ago
#70 - 交个作业吧
Issue -
State: closed - Opened by yasohasakii 7 months ago
#70 - 交个作业吧
Issue -
State: closed - Opened by yasohasakii 7 months ago
#69 - proces single file in foreach,avoid oom
Pull Request -
State: open - Opened by maoxiangyi 7 months ago
#69 - proces single file in foreach,avoid oom
Pull Request -
State: open - Opened by maoxiangyi 7 months ago
#68 - 预训练模型参数和eval参数维度不匹配的问题
Issue -
State: open - Opened by 1019245175 7 months ago
#68 - 预训练模型参数和eval参数维度不匹配的问题
Issue -
State: open - Opened by 1019245175 7 months ago
#67 - c4-zh数据有问题
Issue -
State: closed - Opened by yasohasakii 7 months ago
- 3 comments
#67 - c4-zh数据有问题
Issue -
State: closed - Opened by yasohasakii 7 months ago
- 3 comments
#66 - 关于运行一段时间,机器断电,如何继续训练
Issue -
State: open - Opened by GromZhang 8 months ago
- 2 comments
#66 - 关于运行一段时间,机器断电,如何继续训练
Issue -
State: open - Opened by GromZhang 8 months ago
- 2 comments
#65 - fix: Fix attribute error and reduce memory usage during data processing
Pull Request -
State: open - Opened by noahc1510 8 months ago
#65 - fix: Fix attribute error and reduce memory usage during data processing
Pull Request -
State: open - Opened by noahc1510 8 months ago
#64 - 请问单卡16G显存的4060Ti能训练吗?
Issue -
State: closed - Opened by XiaoluJiayou 8 months ago
- 1 comment
#64 - 请问单卡16G显存的4060Ti能训练吗?
Issue -
State: closed - Opened by XiaoluJiayou 8 months ago
- 1 comment
#63 - Problem with tokenizer?
Issue -
State: open - Opened by shokhjakhonone 8 months ago
- 3 comments
#62 - 请问下这个报错是哪里配置的不对吗?
Issue -
State: open - Opened by beginner-wj 8 months ago
#61 - 请问下这个报错是什么信息?
Issue -
State: closed - Opened by beginner-wj 8 months ago
#60 - 为了丰富和扩充本项目,这里开源了使用deepspeed进行训练的代码和权重(1.75B)
Issue -
State: closed - Opened by AI-Study-Han 9 months ago
#59 - Ignore the `freqs_cis` buffer so that DDP does not broadcast it at construction time
Issue -
State: open - Opened by xiaoguzai 9 months ago
#58 - 跑训练报错
Issue -
State: closed - Opened by singeleaf 10 months ago
- 3 comments
#57 - 自己用
Pull Request -
State: closed - Opened by life-peace 10 months ago
#56 - fix: multi gpu ddp save error
Pull Request -
State: closed - Opened by billvsme 10 months ago
#55 - 配置优化器的部分为什么,大于或等于2D的参数会被衰减,小于2D不会衰减?
Issue -
State: closed - Opened by zerozhoujie 10 months ago
- 2 comments
#54 - /track1/train_valid.json
Issue -
State: closed - Opened by cj401 10 months ago
- 1 comment
#53 - 如何修改,支持4k上下文,以及16k上下文呢?
Issue -
State: closed - Opened by 937739823 10 months ago
- 1 comment
#52 - 交个作业
Issue -
State: closed - Opened by ljg-lixufeng 10 months ago
#51 - 提示:在训练中加入complie = True后再sft中也需要同步,不然会造成模型载入错误
Issue -
State: closed - Opened by Hong-Shuo 11 months ago
#50 - Attention!! 推理代码里面的致命笔误是导致大家看到效果不好的原因。望周知!
Issue -
State: closed - Opened by DLLXW 11 months ago
#49 - 多个节点多卡的pretrain
Issue -
State: closed - Opened by lixin716 11 months ago
- 2 comments
#48 - 修改了词表大小后与预训练模型的维度不匹配,大家怎么处理的呀
Issue -
State: open - Opened by ghost 11 months ago
#47 - 模型效果
Issue -
State: closed - Opened by AI-Study-Han 11 months ago
- 3 comments
#46 - 没有找到此文件
Issue -
State: closed - Opened by servlet1111 12 months ago
#45 - transformers最新版本会报错
Issue -
State: open - Opened by somewordstoolate 12 months ago
- 2 comments
#44 - 模型参数量计算
Issue -
State: open - Opened by zxx20231119 12 months ago
- 2 comments
#43 - 前期数据处理差异
Issue -
State: closed - Opened by wujianqiangwjq about 1 year ago
- 1 comment
#42 - 一个很诡异的错误 IndexError: index 35930 is out of bounds for axis 1 with size 2048
Issue -
State: closed - Opened by zhaodice about 1 year ago
- 1 comment
#41 - 第一轮练完了正在跑第二轮。不能能加个脚本转格式能让obabooga使用?
Issue -
State: open - Opened by limao999666 about 1 year ago
#40 - 为什么预训练时,做attention的时候不需要mask
Issue -
State: closed - Opened by LLH1818 about 1 year ago
#39 - 想问下训练的数据和epoch数
Issue -
State: open - Opened by YuzhouPeng about 1 year ago
- 4 comments
#38 - eos token是空字符串
Issue -
State: closed - Opened by Destiny-Lu about 1 year ago
- 4 comments
#37 - 进行多卡pretrain的时候,出现了如下异常
Issue -
State: open - Opened by GromZhang about 1 year ago
- 7 comments
#36 - 请问是从头开始预训练的,为什么在项目中体现到了llama2,初学者不太理解
Issue -
State: closed - Opened by GromZhang about 1 year ago
- 1 comment
#35 - 总结下几个问题
Issue -
State: open - Opened by Vincent-ZHQ about 1 year ago
- 3 comments
#34 - 为什么在pretrain 309行model.complie要加prefix '_orig_mod'?
Issue -
State: closed - Opened by ToxicNeil about 1 year ago
- 1 comment
#33 - RuntimeError: probability tensor contains either `inf`, `nan` or element < 0
Issue -
State: open - Opened by sunhao about 1 year ago
- 1 comment
#32 - Windows support modifications in pretrain script
Pull Request -
State: closed - Opened by jh01231230 about 1 year ago
#31 - Windows support modifications in pretrain script
Pull Request -
State: closed - Opened by jh01231230 about 1 year ago
- 1 comment
#30 - 没有SFT的话 推理会抱错,麻烦看看
Issue -
State: open - Opened by hopeforus about 1 year ago
- 2 comments
#29 - '../track1/train_valid.json。这个文件在哪里下载?
Issue -
State: open - Opened by hopeforus about 1 year ago
- 2 comments
#28 - sft dataset
Issue -
State: open - Opened by paopao0226 about 1 year ago
- 2 comments
#27 - Data process modifications
Pull Request -
State: closed - Opened by jh01231230 about 1 year ago
#26 - 要训练几个epoch,会有比较好的效果?
Issue -
State: closed - Opened by binwang672012 about 1 year ago
- 4 comments
#25 - 处理百度数据集的时间报错
Issue -
State: open - Opened by hopeforus about 1 year ago
- 6 comments
#24 - 交个作业
Issue -
State: open - Opened by AClolinta about 1 year ago
- 13 comments
#23 - 数据集问题
Issue -
State: open - Opened by zhihui-shao about 1 year ago
- 3 comments
#22 - sft.py运行报错 CUDA out of memory,请问咋解决?
Issue -
State: closed - Opened by qxj about 1 year ago
- 6 comments
#21 - 您好,请问显存为24G 3090预训练这个参数量大小的模型大概需要多久呀?
Issue -
State: open - Opened by LePanda026 about 1 year ago
- 3 comments
#20 - 可以提供一个训练好的模型吗?
Issue -
State: open - Opened by PeterouZh about 1 year ago
- 5 comments
#19 - fix: remove redundant pkg
Pull Request -
State: closed - Opened by jianhu-chen about 1 year ago
#18 - 单卡下运行pretrain.py 报错 Default process group has not been initialized, please make sure to call init_process_group.
Issue -
State: open - Opened by TristanShao about 1 year ago
- 10 comments
#17 - 大家在预训练的时候有遇到过loss为nan吗
Issue -
State: open - Opened by ZK-Zhou about 1 year ago
- 15 comments
#16 - Where to fetch medical_qa_144w.csv?
Issue -
State: closed - Opened by qxj about 1 year ago
- 1 comment
#15 - 在处理百度563baike时Memory error
Issue -
State: open - Opened by ZK-Zhou about 1 year ago
- 5 comments
#14 - 百度云垃圾
Issue -
State: closed - Opened by KKIverson about 1 year ago
- 2 comments
#13 - dataset_sft.py中loss_mask的切片为什么和X一致?
Issue -
State: open - Opened by BigaGrayWolf about 1 year ago
- 2 comments
#12 - Question about tokenizer
Issue -
State: closed - Opened by IshootLaser about 1 year ago
- 1 comment
#11 - 预训练完后执行python sft.py报错找不到文件
Issue -
State: closed - Opened by xamofb-xsk about 1 year ago
- 2 comments
#10 - sft使用的checkpoint问题
Issue -
State: closed - Opened by Deep1994 about 1 year ago
- 1 comment
#9 - medical_qa.bin 没有用上
Issue -
State: closed - Opened by Deep1994 about 1 year ago
- 3 comments
#8 - CUDA_VISIBLE_DEVICES=0,1 torchrun pretrain.py 只利用了一块GPU
Issue -
State: closed - Opened by BigaGrayWolf about 1 year ago
- 6 comments
#7 - 关于 GBK 编码的问题
Issue -
State: closed - Opened by lavinal712 about 1 year ago
- 4 comments
#6 - 上下文长度32K
Issue -
State: closed - Opened by CanvaChen about 1 year ago
- 1 comment