[Feature] will modelscope support prompt-based tuning for large language models? about modelscope HOT 3 CLOSED

Thanks for your attention to the modelscope large language models. We have now provided the prompt-based tuning for the Chinese GPT-3 base/large/1.3B/2.7B models. You just need to handle the data and refer to the code of model card to start training. Later, we will consider supporting training larger models, including GPT-3 175B.

Chinese GPT-3：https://modelscope.cn/models/damo/nlp_gpt3_text-generation_13B/summary

from modelscope.

Thanks for your quick reply and looking forward to GPT-3 175B. I found that the fine-tuning examples exist in GPT-3 large repo and GPT-3 poetry-generation, but not in GPT-3 13B, and I will follow the tutorials. For GPT-3 13B and GPT-3 30B, is there a plan to release these checkpoints for the community?

from modelscope.

We haven't had a plan to release the 13B and 30B for downloading, yet the finetune procedures are the same with the smaller one. So I will suggest you to run and test with smaller size firstly.

from modelscope.