CEGT - rating lists June 30th 2019

[Modern Times]

You may need to change this ?

float_type HALF

To FULL ?

[Modern Times]

I get this:

C:\Engines\Scorpio GPU>scorpio.bat
feature done=0
Number of cores 8 of 8
ht 16777216 X 16 = 256.0 MB
eht 524288 X 8 = 8.0 MB
pht 32768 X 24 = 0.8 MB
treeht 1342169600 X 10 = 12799.9 MB
processors [128]
EgbbProbe 4.3 by Daniel Shawul
egbb_cache 4084 X 8216 = 32.0 MB
0 egbbs loaded !
Loading neural network : ./net-maddex.uff
nn_cache 131072 X 1568 = 196.0 MB
Loading graph on /gpu:0
UFFParser: Parser error: Reshape: Reshape: Volume mismatch
Fail to parse network ./net-maddex.uff
C:\Engines\Scorpio GPU>

[Werner]

Here is the Output now with float type HALF:
feature done=0
Number of cores 4 of 4
ht 16777216 X 16 = 256.0 MB
eht 524288 X 8 = 8.0 MB
pht 32768 X 24 = 0.8 MB
treeht 1342169600 X 10 = 12799.9 MB
processors [128]
EgbbProbe 4.3 by Daniel Shawul
egbb_cache 4084 X 8216 = 32.0 MB
180 egbbs loaded !
Loading neural network : c:\Users\6600K\Arena\Engines\ScorpioNet\Scorpio-30-Jun-2019\net-maddex.uff
nn_cache 131072 X 1552 = 194.0 MB
Loading graph on /gpu:0
0. main_input 7168 = 112 8 8
1. value_head 1 = 1 1 1
2. policy_head 1858 = 1858 1 1
Neural network loaded !
loading_time = 2s
# rnbqkbnr/pppppppp/8/8/8/8/PPPPPPPP/RNBQKBNR w KQkq - 0 1
# [st = 11114ms, mt = 29250ms , hply = 0 , moves_left 10]
63 35 121 1839 e2-e4 c7-c5 Ng1-f3 d7-d6 d2-d4 c5xd4 Nf3xd4 Ng8-f6 Nb1-c3 a7-a6 Bc1-e3
64 28 239 4250 c2-c4 e7-e5 g2-g3 Ng8-f6 Bf1-g2 Bf8-c5 Nb1-c3 Ke8-g8 Ng1-f3
65 19 356 6676 d2-d4 d7-d5 c2-c4 e7-e6 Nb1-c3 Ng8-f6 Ng1-f3 Bf8-e7 g2-g3 Ke8-g8 Bf1-g2
66 27 481 9437 d2-d4 e7-e6 e2-e4 d7-d5 Nb1-c3 Ng8-f6 e4-e5 Nf6-d7 f2-f4 c7-c5 Ng1-f3 Nb8-c6 Bc1-e3 Bf8-e7
67 31 607 12171 d2-d4 c7-c6 e2-e4 d7-d5 Nb1-c3 d5xe4 Nc3xe4 Bc8-f5 Ne4-g3 Bf5-g6 Ng1-f3 Nb8-d7 h2-h4 h7-h6 Bf1-d3
68 32 724 14740 d2-d4 Ng8-f6 Ng1-f3 d7-d5 c2-c4 e7-e6 Nb1-c3 Bf8-e7 g2-g3 Ke8-g8 Bf1-g2
69 37 841 17203 e2-e4 d7-d6 d2-d4 g7-g6 Nb1-c3 Bf8-g7 Ng1-f3 Ng8-f6 Bc1-e3 Ke8-g8
70 29 959 19589 e2-e4 e7-e5 Nb1-c3 Nb8-c6 Ng1-f3 Ng8-f6 d2-d4 e5xd4 Nf3xd4 Bf8-b4 Nd4xc6 b7xc6 Bf1-d3 d7-d5 e4xd5
71 35 1032 21047 e2-e4 e7-e5 Ng1-f3 Ng8-f6 Nf3xe5 d7-d6 Ne5-f3 Nf6xe4 d2-d4 d6-d5 Bf1-d3 Bf8-b4 c2-c3 Bb4-d6 Ke1-g1 Ke8-g8

# Move Value=(V,P,V+P) Policy Visits PV
#----------------------------------------------------------------------------------
# 1 (0.533,0.527,0.530) 23.01 7382 d2-d4 Ng8-f6 e2-e3 d7-d5 c2-c4 e7-e6 Ng1-f3 Bf8-e7 Bf1-e2 Ke8-g8 Ke1-g1
# 2 (0.551,0.527,0.539) 22.39 9130 e2-e4 e7-e5 Ng1-f3 Ng8-f6 Nf3xe5 d7-d6 Ne5-f3 Nf6xe4 d2-d4 d6-d5 Bf1-d3 Bf8-b4 c2-c3 Bb4-d6 Ke1-g1 Ke8-g8
# 3 (0.519,0.527,0.519) 9.44 1008 c2-c4 e7-e5 g2-g3 Ng8-f6 Bf1-g2 Bf8-c5 Nb1-c3 Ke8-g8 Ng1-f3 Nb8-c6
# 4 (0.519,0.527,0.519) 9.30 1007 Ng1-f3 d7-d5 d2-d4 Ng8-f6 c2-c4 e7-e6 Nb1-c3 Bf8-e7 g2-g3 Ke8-g8 Bf1-g2
# 5 (0.522,0.527,0.522) 4.41 531 g2-g3 d7-d5 Ng1-f3 c7-c5 Bf1-g2 Ng8-f6 Ke1-g1 Nb8-c6 d2-d4 c5xd4
# 6 (0.528,0.527,0.528) 4.08 616 e2-e3 Ng8-f6 d2-d4 d7-d5 c2-c4 e7-e6 Ng1-f3 Bf8-e7 Bf1-e2
# 7 (0.500,0.527,0.500) 3.09 200 b2-b3 e7-e5 Bc1-b2 Nb8-c6 e2-e3 Ng8-f6 Ng1-f3 e5-e4 Nf3-d4 Bf8-c5 Nd4xc6
# 8 (0.513,0.527,0.513) 2.85 243 c2-c3 d7-d5 d2-d4 Ng8-f6 Ng1-f3 e7-e6 Bc1-f4 Bf8-d6 e2-e3
# 9 (0.500,0.527,0.500) 2.65 170 d2-d3 d7-d5 Ng1-f3 Ng8-f6 g2-g3 c7-c5 Bf1-g2 Nb8-c6
# 10 (0.515,0.527,0.515) 2.64 238 Nb1-c3 d7-d5 d2-d4 Ng8-f6 Bc1-f4 a7-a6 e2-e3 e7-e6 Ng1-f3 c7-c5 Bf1-e2
# 11 (0.464,0.527,0.464) 2.49 91 f2-f4 d7-d5 Ng1-f3 Ng8-f6 e2-e3 c7-c5 b2-b3 g7-g6
# 12 (0.476,0.527,0.476) 2.11 92 b2-b4 e7-e5 Bc1-b2 Bf8xb4 Bb2xe5 Ng8-f6 c2-c3 Bb4-e7 e2-e3 Ke8-g8
# 13 (0.497,0.527,0.497) 1.94 118 h2-h3 d7-d5 d2-d4 c7-c5 e2-e3 Ng8-f6 Ng1-f3 Nb8-c6
# 14 (0.502,0.527,0.502) 1.87 123 a2-a3 d7-d5 d2-d4 Ng8-f6 Ng1-f3 g7-g6 c2-c4 Bf8-g7
# 15 (0.431,0.527,0.431) 1.62 43 Ng1-h3 d7-d5 d2-d4 c7-c5 d4xc5 Ng8-f6 g2-g3 e7-e5
# 16 (0.474,0.527,0.474) 1.40 61 a2-a4 e7-e5 e2-e4 Ng8-f6 Nb1-c3 Bf8-b4 Ng1-f3 Ke8-g8
# 17 (0.449,0.527,0.449) 1.28 41 h2-h4 d7-d5 d2-d4 c7-c5 d4xc5 Ng8-f6 Ng1-f3 e7-e6
# 18 (0.375,0.527,0.375) 1.24 22 g2-g4 d7-d5 h2-h3 e7-e5 Bf1-g2 Nb8-c6 Nb1-c3
# 19 (0.413,0.527,0.413) 1.21 28 f2-f3 e7-e5 Nb1-c3 Ng8-f6 d2-d4 e5xd4
# 20 (0.439,0.527,0.439) 0.99 28 Nb1-a3 e7-e5 e2-e3 d7-d5 d2-d4 e5-e4 c2-c4

# nodes = 170133 <0% qnodes> time = 10810ms nps = 15738 eps = 0 nneps = 2000
# Tree: nodes = 29173 depth = 19 pps = 1978 visits = 21173
# qsearch_calls = 0 search_calls = 0
move e2e4

@Daniel: why do you use 4CPU + GPU ? Is 1 CPU not enough ?

[Modern Times]

HALF is not compatible with my GPU. I have to use FLOAT.

But I still get the errors

UFFParser: Parser error: Reshape: Reshape: Volume mismatch
Fail to parse network ./net-maddex.uff

[Daniel Shawul]

Change nn_type to 1 like Werner did. Note that he initially got the same error message as you got.

I forgot to mention that in the steps. Had the install script worked, it would have done all that.

For those who want to try INT8, you have to generate calibration data first by running

Code: Select all

./scorpio.bat runinpnn calibrate.epd calibrate.dat

Then you change float_type to INT8 in scorpio.ini.

I think older GPUs like 11xx series have better INT8 support than HALF precision and you might get more than 2knps with that.
On a GTX 20xx series half precision gives me 10x more node than that i.e. 20knps

@Werner The cpu is not really used when you are running on GPU (no alpha-beta search or anything like that even though the old version
used to do that). It will use 128 threads even when you are running on 1 core. I could try and find a programatic way of restricting
all threads to run on specified number of cores though ..

Daniel

[Modern Times]

Change nn_type to 1 like Werner did. Note that he initially got the same error message as you got.

Still the same error. I'll try again tomorrow.

[Werner]

I have started a match between Scorpio and StoofVlees
40 moves in 4 min ponder off
both engines use 2CPU and the GTX 1650

after 10 games:
1: StoofVleesII a10 8,5/10 ·········· 1111=1=11= 12,75
2: Scorpio 2.9.0 MCTS+NN Maddex 1,5/10 0000=0=00= ·········· 12,75

6 StoofVlees II a9 3323
Scorpio around 3023 so near by Arasan 21.3 x64 1CPU
(bad start - but after the 1st win 3100 Points are possible)
next result after 20 games.

[Guenther]

I have started a match between Scorpio and StoofVlees
40 moves in 4 min ponder off
both engines use 2CPU and the GTX 1650

after 10 games:
1: StoofVleesII a10 8,5/10 ·········· 1111=1=11= 12,75
2: Scorpio 2.9.0 MCTS+NN Maddex 1,5/10 0000=0=00= ·········· 12,75

6 StoofVlees II a9 3323
Scorpio around 3023 so near by Arasan 21.3 x64 1CPU

Werner, do you know, if Gian-Carlo will keep Stoofvlees private?
Or is there a slight chance for a release? :)

[Werner]

Hi Guenther,
I think there is a slight chance for the release when the engines is ready: as Deep Sjeng 4.0

[Werner]

I have started a match between Scorpio and StoofVlees
40 moves in 4 min ponder off
both engines use 2CPU and the GTX 1650

after 10 games:
1: StoofVleesII a10 8,5/10 ·········· 1111=1=11= 12,75
2: Scorpio 2.9.0 MCTS+NN Maddex 1,5/10 0000=0=00= ·········· 12,75

6 StoofVlees II a9 3323
Scorpio around 3023 so near by Arasan 21.3 x64 1CPU
(bad start - but after the 1st win 3100 Points are possible)
next result after 20 games.

after 20 games
1 StoofVleesII a10 2200 +301 1111½1½11½11½½1111½1 17.0/20
2 Scorpio 2.9.0 MCTS+NN Maddex 2200 -301 0000½0½00½00½½0000½0 3.0/20