hope366 / Lizzie-improvements

Lizzie - Leela Zero Interface
GNU General Public License v3.0
29 stars 9 forks source link

Mysterious Analysis Chart #19

Closed hebaeba closed 3 years ago

hebaeba commented 3 years ago

I set up a 40blocks and 60blocks katago net in Lizzie-improved-v2.6.2. The other settings are as default. Then I tried to analyze a newspaper Go problem with 40b, and the phenomenon shown in the attached image happened. Please see the expected procedure for the sub-go board. To be sure, I tried the same analysis with the latest package ( Lizzie-all-in-one-v2.7 ), and this problem seemes to happen when I switch engines after loading an SGF game. The network I used for my analysis is kata1-b40c256-s9738904320-d2372933741.bin.gz and kata1-b60c320-s4056272640-d2292804613.bin.gz.

The analyzed SGF game is as follows, which is "SGF with fixed arrival point". "(;GM[1]FF[4]SZ[19]KM[6.5]AB[qd][pd][od][oc][ne][md][mf][lc][pj][pn][qp][qo][rq][qr][sr][lo][lq][lr][fq][fo] AW[nd][nc][ob][mb][na][pc][qc][qb][pa][mm][nn][np][mq][pp][pq][qq][rr][dp][dr][cn][de] [dc](;B[mk])(;B[km])(;B[er]))"

LZ-katago40-27

hebaeba commented 3 years ago

The problem has been resolved in the new version. Thank you very much. Therefore, I am closing this ISSUE.