Kunhao ZHENG

had the same issue here, i tried adding `.ref()` to the `v` there and the error message becomes ``` 2022-06-17 09:24:55.671316: W tensorflow/core/framework/op_kernel.cc:1733] INTERNAL: 'cuLaunchKernel(function, gridX, gridY, gridZ, blockX, blockY,...

Could you provide more details about your hardware information. - How many GPUs you are using to load the model? Out of the box, the 13b model cannot be loaded...

You may wanna take a look at https://github.com/facebookresearch/codellama/issues/82 for quantization if the usecase is inference only. If you could run batch_size = 1, then as discussed above https://github.com/facebookresearch/codellama/issues/77#issuecomment-1703664262, the gradient...

Hi Brando! By exact accuracy you means the accuracy breaking down to each statement, or the sota accuracy like the paper you mentioned here?

I can provide several paper that I know reporting accuracies on miniF2F, more or less in chronological order. The list may be incomplete and any fix welcome! - minif2f: https://arxiv.org/abs/2109.00110....

as both stan and jesse left openai, do we have anyone who can merge this PR? @spolu, probably Daniel?

> 👀 Is there any updates...? Hi! Thank you for your interest! The fixes are presented in FAIR's fork. https://github.com/facebookresearch/miniF2F

Hi Brando! I have no idea about classical SMT/ATP solvers, would you name some of them?

as far as i know there is no undergoing efforts on porting minif2f to these solvers. would like to know if this is in principle feasible?

Hi @yutakang ! It is correct. The previous maintainers don't have the admin right to merge PR in this repo hosting miniF2F. Currently, several new fixes are into FAIR's repo....