diff options
| author | tslil <tslil@posteo.de> | 2026-09-11 12:56:12 +0100 |
|---|---|---|
| committer | tslil <tslil@posteo.de> | 2026-09-11 17:00:18 +0100 |
| commit | 6e90f1145b2c5c0f2dcc82d21ec45432274b8cad (patch) | |
| tree | 18db1c03ea7c3c4724028fc5154d4bca3d7200a5 /include/weights_6.c | |
| parent | 220a4b79e4b6d742a9ca7c278a1962b870500fbc (diff) | |
move from single candidate td-leaf with gating to tournament of a population
The goal in this change was to overcome the rock-paper-scissors phenomenon that was happening in capability space by having on a single candidate learn to beat the most recent version of itself. In some sense this lead to rotating around in capability space instead of chasing monotonic improvements.
I hope that flipping to a system in which many (N=5) candidates play each other and learn from their own set of diverse games in this pool, combined with --replay-iters, will lead to a given network having to have a much more comprehensive understanding of the space of tactics. I haven't yet had time to train this thorougly, will run those experiments on my desktop computer.
Diffstat (limited to 'include/weights_6.c')
| -rw-r--r-- | include/weights_6.c | 10 |
1 files changed, 5 insertions, 5 deletions
diff --git a/include/weights_6.c b/include/weights_6.c index 77ea409..915fc54 100644 --- a/include/weights_6.c +++ b/include/weights_6.c @@ -2,12 +2,12 @@ /* v000, random init */ #include "weights_6.h" -const float six_embedding[7] = +const float SIX_embedding[7] = { 0.724845707, -0.209938288, -0.347967505, 0.821479738, -1.34644675, 0.0270497333, 0.450257659 }; -const float six_dense_weights[SIX_DENSE_NUM][SIX_INP_NUM] NN_ALIGNED = +const float SIX_dense_weights[SIX_DENSE_NUM][SIX_INP_NUM] NN_ALIGNED = { {0.0375181399, 0.0131807914, 0.0733145475, -0.108837478, -0.155519173, -0.15236935, 0.161187366, -0.102666609, -0.105799064, 0.00281478628, 0.0976773277, -0.159017026, 0.0197313707, -0.0683622435, -0.145631433, -0.153941378, @@ -331,7 +331,7 @@ const float six_dense_weights[SIX_DENSE_NUM][SIX_INP_NUM] NN_ALIGNED = -0.133655116, 0.101930477, 0.119332045, -0.127442241, -0.129697219, -0.0865818784}, }; -const float six_dense_biases[SIX_DENSE_NUM] = +const float SIX_dense_biases[SIX_DENSE_NUM] = { -0.109864615, -0.00545889279, -0.0512907989, -0.0936995149, 0.145228133, 0.136148497, 0.082344763, 0.16194196, -0.0229825824, 0.116574578, -0.0256564897, -0.058733996, -0.14782922, -0.115881361, -0.139623746, 0.14362292, @@ -343,7 +343,7 @@ const float six_dense_biases[SIX_DENSE_NUM] = 0.161606282, -0.136956587, 0.0591680035, 0.117993973, -0.100358412, -0.0904804021, 0.161572158, 0.0794185847 }; -const float six_logit_weights[SIX_DENSE_NUM] NN_ALIGNED = +const float SIX_logit_weights[SIX_DENSE_NUM] NN_ALIGNED = { 0.0506259352, -0.0670735389, -0.0409830809, 0.0799648017, -0.0936499536, -0.122460753, 0.0720161647, -0.00714924932, 0.0374628901, -0.00931178033, -0.0230136961, 0.00667156279, 0.0823754966, -0.0555169284, -0.00904981792, 0.116692498, @@ -355,4 +355,4 @@ const float six_logit_weights[SIX_DENSE_NUM] NN_ALIGNED = -0.0314992517, 0.0338666737, 0.0178338587, -0.0309750885, -0.0261106789, -0.0142950565, 0.0768712908, -0.121748224 }; -const float six_logit_bias = -0.076127708; +const float SIX_logit_bias = -0.076127708; |