piekey2022
piekey2022
会报这个错误json.decoder.JSONDecodeError: Expecting ',' delimiter: line 112577 column 19 (char 2131822)
我有个文件，逐行进行分词，刚好第1595个句子分词后的list长度是48个词 如果我直接读第1595个句子进行分词，长度就是50个词 打印彼此的结果，会发现直接对这个句子进行分词的话，有个人名莱迪格会被分词成三个字。但逐行逐行的进行分词，到那一句的时候，莱迪格就不会被拆分成三个字，所以会少两个词。 为什么分词的结果会有出入呢
**Description** When I enabled max_queue_delay_microseconds to improve the response speed of the model, I found that there were occasional errors. I set max_queue_delay_microseconds to 70000. Then I sent three tensor...
这份代码的tokenizer和之前glm的tokenizer代码似乎不太一样，特别是没有build_inputs_for_generation这个函数，请问forward的时候input_ids和position_ids的构造方式和以前是一样的吗。以前似乎原文中必须包含mask，现在这个代码，我看generate函数的输入好像没有要求要有mask_token
### Is there an existing issue for this? - [X] I have searched the existing issues ### Current Behavior 在一张80g显存的卡上训练，之前训练10b的glm模型可以开到batchsize4，这个模型开到2就很容易爆显存。有用amp，不知道什么原因。 直接用huggingface的代码没办法做模型并行，不知道有什么好的办法 ### Expected Behavior _No response_ ### Steps To Reproduce...
A recent branch of peft is about to support multiple lora adapters. This implementation feels very suitable for the training in ppo stage. An sft model can be used as...
### System Info ```Shell - `Accelerate` version: 0.18.0 - Platform: Linux-5.4.0-128-generic-x86_64-with-glibc2.29 - Python version: 3.8.10 - Numpy version: 1.24.2 - PyTorch version (GPU?): 1.13.1+cu117 (True) - `Accelerate` default config: Not...
### Description 是否有用多轮的语料训练过，如果支持的话，请问prompt要如何构造呢？ ### Alternatives 谢谢
### System Info flagai 1.7.1 flash-attn 1.0.4 torch 1.13.1 python3.8 linux ### Information - [X] The official example scripts - [ ] My own modified scripts ### Tasks - [X]...
This api is very convenient to use, and I hope to see the new pokémon data join it as soon as possible.