chenhao
chenhao
### Electerm Version(Electerm版本) v1.21.93 ### Platform detail (平台详情) mac os ### What steps will reproduce the bug?(重新问题的详细步骤) 打开后,在首页输入命令执行,再退出Electerm,重新进入后,首页终端自动执行了上次输入的命令 ### What should have happened?(期望的结果) 重新打开后展示新终端 ### Would this happen in other...
### Is there an existing issue for this? - [X] I have searched the existing issues ### Current Behavior 使用model.chat   使用model.stream_chat   或者说界面上的三个参数,model.chat和model.stream_chat有什么区别? ### Expected Behavior _No...
**Describe the bug** A clear and concise description of what the bug is. **To Reproduce** Steps to reproduce the behavior: 1. Go to '...' 2. Click on '....' 3. Scroll...
### 🐛 Describe the bug ``` colossalai run --nproc_per_node=4 train_sft.py \ > --pretrain "/data/chenhao/train/ColossalAI/to/llama-7b-hf/" \ > --model 'llama' \ > --strategy colossalai_zero2 \ > --log_interval 10 \ > --save_path "/data/chenhao/train/ColossalAI/Coati-7B"...
When running the program on 8 cards across two machines, I encounter a ChildFailedError.  master节点: ``` torchrun --nproc_per_node=8 --nnodes=2 --node_rank=0 --master_addr=10.xxx.xx.xx --master_port=1234 train.py config/train_shakespeare_char.py ``` worker节点: ``` torchrun --nproc_per_node=8...
### What feature you'd like to see or proposal(期望什么新功能/特性或者建议) 现在只支持横分屏或者竖分屏,不能同时
https://huggingface.co/IDEA-CCNL/Randeng-Pegasus-523M-Summary-Chinese 推理代码有两个问题 # 问题一:AttributeError: 'PegasusTokenizer' object has no attribute 'vocab',需要把PegasusTokenizer中vocab的初始化放在super.__init__函数之前 ``` Traceback (most recent call last): File "/data/chenhao/codes/kmai-model/src/test.py", line 7, in tokenizer = PegasusTokenizer.from_pretrained("/data/models/kmai/IDEA-CCNL/Randeng-Pegasus-523M-Summary-Chinese") File "/data/chenhao/anaconda/envs/kmai/lib/python3.10/site-packages/transformers/tokenization_utils_base.py", line 2045, in from_pretrained...
When I remove a parameter from my filter, the return result is abnormal When my filter parameter is this ``` "filter": { "must": [ { "key": "category", "match": { "value":...
python3 index.py zh ../data/dureader.para test_index 出现这个错误 