pppppM
pppppM
Would you like to reopen this pr and complete this feature together?
Sorry for taking so long to get back to you. Next, I will focus on MMRazor, and solve the related issues you raised as soon as possible.
Thanks for your issue. The documentation and examples in this part are not complete enough, and we will improve them as soon as possible.
Currently， pruner only supports CNN-based models. We have not verified the performance of AutoSlim on segmentation, and we will release related settings after verification.
Due to a lack of manpower, the progress of transferring AutoSlim to other tasks is not very satisfactory. We are reproducing BigNAS, if it goes well, we will release the...
Because there are several losses in distiller, they need to be distinguished according to `name`. The `name` will be poped before `build_loss`. https://github.com/open-mmlab/mmrazor/blob/4611be25e07b2e4d9d26d615215f06dc86a010ef/mmrazor/models/distillers/single_teacher.py#L64-L66
Implemented in dev-1.x https://github.com/open-mmlab/mmrazor/tree/dev-1.x
@xiaohangguo 收到！我们正在设计 RLHF 部分的实现方案，等有初版设计方案后，可以在讨论区详细讨论
我们目前正在开发基于 internlm2 的自定义工具调用微调的功能，敬请期待
Please complete the git commit id so that we can reproduce it.