I suggest scaling down the problem until some experience is gained. You don't need the full-fledge 40-block network to get started. You can probably get away with using only 20 blocks and maybe 128 features (from 256). That should save you about a factor of 8, plus you can use larger mini-batches.
You can also start with 9x9 go. That way games are shorter, and you probably don't need 1600 network evaluations per move to do well. Álvaro. On Fri, Oct 20, 2017 at 1:44 PM, Gian-Carlo Pascutto <g...@sjeng.org> wrote: > I reconstructed the full AlphaGo Zero network in Caffe: > https://sjeng.org/dl/zero.prototxt > > I did some performance measurements, with what should be > state-of-the-art on consumer hardware: > > GTX 1080 Ti > NVIDIA-Caffe + CUDA 9 + cuDNN 7 > batch size = 8 > > Memory use is about ~2G. (It's much more for learning, the original > minibatch size of 32 wouldn't fit on this card!) > > Running 2000 iterations takes 93 seconds. > > In the AlphaGo paper, they claim 0.4 seconds to do 1600 MCTS > simulations, and they expand 1 node per visit (if I got it right) so > that would be 1600 network evaluations as well, or 200 of my iterations. > > So it would take me ~9.3s to produce a self-play move, compared to 0.4s > for them. > > I would like to extrapolate how long it will take to reproduce the > research, but I think I'm missing how many GPUs are in each self-play > worker (4 TPU or 64 GPU or ?), or perhaps the average length of the games. > > Let's say the latter is around 200 moves. They generated 29 million > games for the final result, which means it's going to take me about 1700 > years to replicate this. I initially estimated 7 years based on the > reported 64 GPU vs 1 GPU, but this seems far worse. Did I miss anything > in the calculations above, or was it really a *pile* of those 64 GPU > machines? > > Because the performance on playing seems reasonable (you would be able > to actually run the MCTS on a consumer machine, and hence end up with a > strong program), I would be interested in setting up a distributed > effort for this. But realistically there will be maybe 10 people > joining, 80 if we're very lucky (looking at Stockfish numbers). That > means it'd still take 20 to 170 years. > > Someone please tell me I missed a factor of 100 or more somewhere. I'd > love to be wrong here. > > -- > GCP > _______________________________________________ > Computer-go mailing list > Computer-go@computer-go.org > http://computer-go.org/mailman/listinfo/computer-go
_______________________________________________ Computer-go mailing list Computer-go@computer-go.org http://computer-go.org/mailman/listinfo/computer-go
