training rework, best weights so far*

After quite a bit of experimentation it would appear that a far better way to train on self-play data in this setup is to randomise not the moves but the opening. The new selfplay and match binaries have lost the epsilon parameter and now play a random opening of some length before switching to pure, deterministic engine inference. This also resulted in a cleanup of nn_train.py.

As an added bonus, in this configuration it makes sense to harvest training data from both sides in self-play and consequently the two sides are given distinct game ids (this was almost surely an oversight in the original version of TD(lambda) that i had). I also noticed that rand32 was not using the top 32bits of xorshitf*.

*Under default settings nn_train.py produces a 5x5 weight set that doesn't lose a single game to any other cttei that i cared to look at, when measured using `racetrack` (which has a diverse balanced opening book).

1 files changed, 1 insertions, 1 deletions

diff --git a/include/xorshift64.h b/include/xorshift64.h
index 2b13f60..120e337 100644
--- a/include/xorshift64.h
+++ b/include/xorshift64.h
@@ -30,7 +30,7 @@ extern uint64_t xors;
   }
 
 #define RANDOM64 (xors * 0x2545F4914F6CDD1D)
-#define RANDOM32 ((uint32_t)RANDOM64)
+#define RANDOM32 ((uint32_t)(RANDOM64>>32))
 
 static inline uint64_t rand64(void) {
   XORSHIFT64;