Jiarui Fang（方佳瑞）
Jiarui Fang（方佳瑞）
> hehe 友商别黑，现在已经很多做Decoder-only LLM的inference了，其实挑战和encoder有比较大的差别了。这issue post的两年前的大模型还不是指decoder-only的。
Currently, a config file is not necessary. See this as the latest CAI GPT example https://github.com/hpcaitech/ColossalAI/blob/main/examples/language/gpt/README.md
> Please run pre-commit on your changed files. I have already rerun it.
you have to map your local directory to docker with `-v` to do column mapping. For docker volumn mapping usage please refer https://www.geeksforgeeks.org/mounting-a-volume-inside-docker-container/
Hi, can you tell me your method to **use a GPT model I trained using ColossalAI with huggingface/transformers**? Pointing out which example is your implementation reference would be helpful.
Hello, how long did you train the model. Epoches? Iterations? Considering that train a diffusion model from scratch is very slow to converge.
Actually, I am not an expert on SP and PP. I can help you to contact the author of the SP paper. @FrankLeeeee can you help with this project?
You can fix the bug with the correct PL version. I closed the issue.
Thanks for using our new feature, this is a known bug and we are working on fixing it. I will reply you in this issue if it is fixed.
That's right, policy cpu is more stable than auto. I suggest you use CPU for large model. We will check the problem in auto implementation. you can refer the benchmark...