liufeiran

teracli would core dump when exit

8

``` #0 0x00007f2a47a45cc9 in __GI_raise (sig=sig@entry=6) at ../nptl/sysdeps/unix/sysv/linux/raise.c:56 #1 0x00007f2a47a490d8 in __GI_abort () at abort.c:89 #2 0x00007f2a48350535 in __gnu_cxx::__verbose_terminate_handler() () from /usr/lib/x86_64-linux-gnu/libstdc++.so.6 #3 0x00007f2a4834e6d6 in ?? () from /usr/lib/x86_64-linux-gnu/libstdc++.so.6 #4...

can't work with IPv6

3

提示 connection refused

simcse_unsup_loss 只有一个loss函数，有监督跟无监督用的loss函数是一样的吗。？

pretrain的数据为啥用的是sft的数据。

1

这两个的数据格式是一样的吗

bad result on test image

6

on a test image .  then ,using pretrained model and run command as below. **python inference.py --mode recognize --structure_config_path structure_config.json --structure_model_path /home/koo/桌面/po/TATR-v1.1-All-msft.pth --structure_device cpu --image_dir /home/koo/桌面/pic/ --out_dir ./output -o...

APP的targetSdkversion由23改成26后，调用分享，APP闪退。

ResponseBatch 返回结果不正确

5

同一batch中的prompt，有的prompt返回正确，有的prompt返回不正确，检查后发现是padding 的问题，代码中并没有padding mask的相关实现啊。

Pangu 2.6b 启动失败。

3

Traceback (most recent call last): File "/mnt/d/ai/RLHF/test.py", line 3, in tokenizer = AutoTokenizer.from_pretrained("/mnt/d/ai/pretrain_models/pangu", trust_remote_code=True) File "/home/liu/anaconda3/envs/rlhf/lib/python3.10/site-packages/transformers/models/auto/tokenization_auto.py", line 774, in from_pretrained return tokenizer_class.from_pretrained(pretrained_model_name_or_path, *inputs, **kwargs) File "/home/liu/anaconda3/envs/rlhf/lib/python3.10/site-packages/transformers/tokenization_utils_base.py", line 2028, in from_pretrained...

使用默认参数训练，模型效果数据能提供下吗