David J Wu comments

Results 304 comments of


                                            David J Wu

A question in the "match" module

What are you setting nnMaxBatchSize to?

A question in the "match" module

That's interesting! * What happens if you use only 1x4090 instead of 3x4090? * What happens if you run a different number of games in parallel than 288? * Do...

match: Wrong useNHWC/FP16 settings are given to bot if there are bot dedups

Thanks for reporting, yeah this is an ugly conflict between the neural net deduplication code and the parameter indexing code where neither part was written with attention to the other....

Can't compile for OpenCL 1.2

Do you know where the lib and include dirs are for your OpenCL installation? If so, it may work to specify `-DOpenCL_LIBRARY=... -DOpenCL_INCLUDE_DIR=...` but replacing the ... with the appropriate...

FATAL ERROR: Failed test assert: buf.result->policyProbs[buf.result->getPos(Location::ofString("E16",board),board)] >= 0.95

Thanks for the report and thanks for the contributions! Unfortunately, it sounds like your GPU may have malfunctioned. The path mentioned in an error is a path to the source...

There is a flying dagger game on the Wildfox Go gaming platform, flying dagger is to play 2 moves at a time。can find the combination with the highest win rate from the candidate moves by exhaustive enumeration, or modify the code to realize it.

In principle, I think you could use the "avoidMoves" parameter documented at https://github.com/lightvector/KataGo/blob/master/docs/Analysis_Engine.md to prohibit every possible legal move for the opponent except for passing on the second ply, as...

There is a flying dagger game on the Wildfox Go gaming platform, flying dagger is to play 2 moves at a time。can find the combination with the highest win rate from the candidate moves by exhaustive enumeration, or modify the code to realize it.

Cool. You've probably already seen the documentation at https://github.com/lightvector/KataGo/blob/master/docs/Analysis_Engine.md but please read it more carefully, for example you tried to provide an "sgf" field but the documentation does not ever...

v1.13.1 TensorRT katago failed to use or generate trtcache when genconfig

There's a chance that your TensorRT and CUDA versions are too recent. I suspect the NVIDIA libraries are not always backwards compatible, and the precompiled exes are using older versions...

The introduction of Fixed Variance Initialization and One Batch Norm is different from code.

Variance is proportional to the index of blocks + 1, because every time you have a (+), you are adding two things that we idealize as uncorrelated random vectors. When...

The introduction of Fixed Variance Initialization and One Batch Norm is different from code.

I updated https://github.com/lightvector/KataGo/blob/master/docs/KataGoMethods.md#fixed-variance-initialization-and-one-batch-norm with an additional diagram to make this more clear. The point of this is to choose an initialization scale so that the entire operation the net is...