LuGY comments

Results 28 comments of


                                            LuGY

triton softmax support multi-batch

I found it would fail when the batch dimension comes to 2.

triton softmax support multi-batch

Update: It works incorrectly before this commit. For the way getting `bias_ptr` not supports multi-batch. And I have fixed the triton version. CUDA version may support multi-batch one day :(

Multimer model scores

Hi, it seems not so reasonable. Could you please share your input to help us reproduce it?

Multimer model scores

Hi, I have reproduced this problem. And it would be fixed soon.

Multimer model scores

Hi, I have made some experiments. And I found it is the preprocessing led to the difference, seems there is some randomness in preprocessing. Can you rerun alphafold with precomputed...

About the recycle dimension

Hi, this part follows the AlphaFold2 paper. See 1.10 of Supplementary information of AF2.

Did fastfold support bfloat16 inference?

Thanks for your question, this feature is still under development.

Dimension mismatch for einsum() in ops.py

Hi, it seems that these features were not produced by FastFold preprocess. Did you load a pickle file or other things as features?