Issues
- 0
huggingface下载的数据有点大
#82 opened by qilong-zhang - 0
Skypile-150B好像有40%都是精确重复的
#81 opened by fyubang - 2
SkyPile-150B数据集的重复率有点小高
#72 opened by genggui001 - 0
为什么不直接用model返回的loss而要自己计算呢?
#79 opened by Chandler-Bing - 0
英文评测中的eval loss
#80 opened by XXares - 1
如何使用load_from_disk加载skypile-150B
#78 opened by 00ffcc - 3
eval_loss结果与原文结果不一致
#76 opened by APiaoG - 1
关于Chatglm的ppl
#75 opened by hxsz1997 - 2
chat模型快发布了吗
#58 opened by sea-boat - 3
ppl 的测试脚本只输出了 loss,不输出 ppl?
#74 opened by SefaZeng - 1
SkyPile-150B数据集的数据类型
#73 opened by HaimianYu - 1
- 0
关于MOCK_GSM8K_TEST的问题部分
#71 opened by cobraheleah - 1
报错Some modules are dispatched on the CPU or the disk. Make sure you have enough GPU RAM to fit
#67 opened by gabrielpondc - 6
官方提供 磁力链接 数据源吗
#66 opened by hggq - 3
中文领域数据perplexity评估脚本问题
#68 opened by zhliuworks - 1
- 3
- 3
BUG:ceval, cmmlu和mmlu中选项ABCD的概率计算错误
#57 opened by naturesphere - 1
请问数据清洗,预处理代码,有计划开源吗?
#62 opened by zdaiot - 1
关于该模型的部署,官方有推荐的框架吗?
#59 opened by liwenju0 - 1
当前eval_loss脚本不支持chatglm系列模型
#64 opened by wbq9224 - 1
BUG:评测loss计算中attention_mask有误
#55 opened by zhangbin1997 - 4
请问什么时候会再度开放开源数据集?
#48 opened by Johnson-Ding - 1
关于评测集的选择和使用
#51 opened by zhangbin1997 - 1
泄露检测的ref集合问题
#61 opened by dongZheX - 17
请问您开源的150B数据集huggingface上怎么不能下载了?
#30 opened by sunzhuojun - 1
评测数据集MOCK_GSM8K_TEST使用方式
#54 opened by cafeii - 6
请问测试数据会公开么?
#46 opened by 4IK1d - 3
Questions about eval_loss.py
#50 opened by chengeharrison - 1
Skywork 团队有兴趣推出一个 7B 的蒸馏版本以支持推测采样和低资源设备推理吗?
#49 opened by tq-xyy - 5
PPL领域数据计算Average结果按照每个领域结果平均对不上
#45 opened by autumnCanTell - 0
SkyPile-150B not found
#44 opened by siruizhang30 - 8
eval loss标准化是否可以理解为平均每个token的loss
#43 opened by Davidgzx - 6
验证集评测结果复现与报告表不符
#32 opened by fengzi258 - 1
The eval_loss_tp.py file is missing.
#41 opened by zinccat - 0
词表中没有\n字符,应该采用什么字符来表示换行
#40 opened by ChenYang24 - 0
save的checkpoint里面没有bin文件
#39 opened by QuanhuiGuan - 3
关于预训练数据拼接
#35 opened by young-chao - 2
RuntimeError: CUDA error: CUBLAS_STATUS_NOT_INITIALIZED when calling `cublasCreate(handle)`
#38 opened by QuanhuiGuan - 3
咨询一下预训练阶段第一次预训练和第二次预训练的数据使用问题
#29 opened by zgctmac - 0
- 2
the dataset (Skypile-150B) can not be download
#36 opened by nicosouth - 1
目前的开源模型版本是否支持工具调用?
#33 opened by danjier7 - 1
Skypile-150B数据里是否包含Skypile-STEM数据?
#34 opened by zgctmac - 2
请问能不能测试下GPT-4的 L_{test} - L_{train}
#31 opened by wxj630 - 1
- 1
does it support FlashAttention-2?
#23 opened by ericxsun - 0
Error: bash_scripts/skywork_eval_loss.sh
#28 opened by fengzi258 - 4
Skywork-13B-Chat
#22 opened by Tonsjkjkas