Scorpio 2.8.7 MCTS+NN windows version

Discussion of anything and everything relating to chess playing software and machines.

Moderators: bob, hgm, Harvey Williamson

Forum rules
This textbox is used to restore diagrams posted with the [d] tag before the upgrade.
User avatar
AdminX
Posts: 5289
Joined: Mon Mar 13, 2006 1:34 pm
Location: Acworth, GA
Contact:

Re: Scorpio 2.8.7 MCTS+NN windows version

Post by AdminX » Sat Sep 08, 2018 6:52 pm

Daniel Shawul wrote:
Sat Sep 08, 2018 5:51 pm
Daniel Shawul wrote:
Sat Sep 08, 2018 5:15 pm
Nice!!

Lets try and get the gpu version working.

Daniel
I have now made it so that the necessary CUDA 9.2 dlls come pre-packaged as well.
If you have a GPU already, you should have the driver i.e. nvcuda.dll so no need to put that in.
The process for installing egbbdll for the gpu should be equally easy as the CPU counterpart now.

Download egbbdll for the gpu here https://github.com/dshawul/Scorpio/rele ... ws-gpu.zip

Extract it somewhere and set the "Path" environement variable. I should be able to write an install script for this soon.

Daniel
Done. I've used the previously used eenvironement variable path. However now engine is non responsive. IE: Not Moving.

Correction: Seems to be working
"Good decisions come from experience, and experience comes from bad decisions."
__________________________________________________________________
Ted Summers

Daniel Shawul
Posts: 3810
Joined: Tue Mar 14, 2006 10:34 am
Location: Ethiopia
Contact:

Re: Scorpio 2.8.7 MCTS+NN windows version

Post by Daniel Shawul » Sat Sep 08, 2018 7:12 pm

AdminX wrote:
Sat Sep 08, 2018 6:52 pm
Daniel Shawul wrote:
Sat Sep 08, 2018 5:51 pm
Daniel Shawul wrote:
Sat Sep 08, 2018 5:15 pm
Nice!!

Lets try and get the gpu version working.

Daniel
I have now made it so that the necessary CUDA 9.2 dlls come pre-packaged as well.
If you have a GPU already, you should have the driver i.e. nvcuda.dll so no need to put that in.
The process for installing egbbdll for the gpu should be equally easy as the CPU counterpart now.

Download egbbdll for the gpu here https://github.com/dshawul/Scorpio/rele ... ws-gpu.zip

Extract it somewhere and set the "Path" environement variable. I should be able to write an install script for this soon.

Daniel
Done. I've used the previously used eenvironement variable path. However now engine is non responsive. IE: Not Moving.

Correction: Seems to be working
Ok make sure you have all the dlls. If there is something missing it will be non-responsive.
Also if it works set the number of threads to very high number like 128 to see the benefit of batching.
With low values such as mt=2 you shouldnt see much benefit from using the GPU.
Could you post a picture for the GPU run like you did before ?

Code: Select all

c:\dev\egbbdll\gpu>dir
 Volume in drive C is OS
 Volume Serial Number is 78F4-686E

 Directory of c:\dev\egbbdll\gpu

09/08/2018  05:54 PM    <DIR>          .
09/08/2018  05:54 PM    <DIR>          ..
07/09/2018  12:29 AM        59,784,704 cublas64_92.dll
09/07/2018  09:43 PM            81,106 CUDA.txt
09/07/2018  09:43 PM            38,963 CUDNN.txt
09/07/2018  09:58 PM       336,443,392 cudnn64_7.dll
06/12/2018  11:53 PM        87,017,472 cufft64_92.dll
06/12/2018  11:53 PM        47,990,784 curand64_92.dll
06/12/2018  11:53 PM       114,022,400 cusolver64_92.dll
09/08/2018  05:34 PM       437,867,649 egbbdll64-nn-windows-gpu.zip
09/08/2018  05:18 AM            93,184 egbbdll64.dll
07/10/2018  09:11 PM       101,838,848 tensorflow.dll
              10 File(s)  1,185,178,502 bytes
               2 Dir(s)  151,305,957,376 bytes free

User avatar
AdminX
Posts: 5289
Joined: Mon Mar 13, 2006 1:34 pm
Location: Acworth, GA
Contact:

Re: Scorpio 2.8.7 MCTS+NN windows version

Post by AdminX » Sat Sep 08, 2018 7:24 pm

Code: Select all

D:\Chess Engines\ScorpioNN>scorpio use_nn 1 mt 128 montecarlo 1 frac_alphabeta 0 backup_type 6 go quit
feature done=0
ht 4194304 X 16 = 64.0 MB
eht 524288 X 8 = 8.0 MB
pht 32768 X 24 = 0.8 MB
treeht 335539200 X 40 = 12799.8 MB
processors [128]
processors [128]
EgbbProbe 4.1 by Daniel Shawul
509 egbbs loaded !
Loading neural network ...
2018-09-08 15:21:14.220729: I c:\users\user\source\repos\tensorflow\tensorflow\core\platform\cpu_feature_guard.cc:141] Your CPU supports instructions that this TensorFlow binary was not compiled to use: AVX AVX2
2018-09-08 15:21:14.535000: I c:\users\user\source\repos\tensorflow\tensorflow\core\common_runtime\gpu\gpu_device.cc:1392] Found device 0 with properties:
name: GeForce GTX 1060 6GB major: 6 minor: 1 memoryClockRate(GHz): 1.835
pciBusID: 0000:01:00.0
totalMemory: 6.00GiB freeMemory: 4.97GiB
2018-09-08 15:21:14.677391: I c:\users\user\source\repos\tensorflow\tensorflow\core\common_runtime\gpu\gpu_device.cc:1392] Found device 1 with properties:
name: GeForce GTX 1060 6GB major: 6 minor: 1 memoryClockRate(GHz): 1.835
pciBusID: 0000:05:00.0
totalMemory: 6.00GiB freeMemory: 4.97GiB
2018-09-08 15:21:14.687719: I c:\users\user\source\repos\tensorflow\tensorflow\core\common_runtime\gpu\gpu_device.cc:1471] Adding visible gpu devices: 0, 1
2018-09-08 15:21:16.312770: I c:\users\user\source\repos\tensorflow\tensorflow\core\common_runtime\gpu\gpu_device.cc:952] Device interconnect StreamExecutor with strength 1 edge matrix:
2018-09-08 15:21:16.318185: I c:\users\user\source\repos\tensorflow\tensorflow\core\common_runtime\gpu\gpu_device.cc:958]      0 1
2018-09-08 15:21:16.321937: I c:\users\user\source\repos\tensorflow\tensorflow\core\common_runtime\gpu\gpu_device.cc:971] 0:   N N
2018-09-08 15:21:16.327157: I c:\users\user\source\repos\tensorflow\tensorflow\core\common_runtime\gpu\gpu_device.cc:971] 1:   N N
2018-09-08 15:21:16.331470: I c:\users\user\source\repos\tensorflow\tensorflow\core\common_runtime\gpu\gpu_device.cc:1084] Created TensorFlow device (/job:localhost/replica:0/task:0/device:GPU:0 with 4734 MB memory) -> physical GPU (device: 0, name: GeForce GTX 1060 6GB, pci bus id: 0000:01:00.0, compute capability: 6.1)
2018-09-08 15:21:16.575703: I c:\users\user\source\repos\tensorflow\tensorflow\core\common_runtime\gpu\gpu_device.cc:1084] Created TensorFlow device (/job:localhost/replica:0/task:0/device:GPU:1 with 4734 MB memory) -> physical GPU (device: 1, name: GeForce GTX 1060 6GB, pci bus id: 0000:05:00.0, compute capability: 6.1)
2018-09-08 15:21:16.885257: I c:\users\user\source\repos\tensorflow\tensorflow\core\common_runtime\gpu\gpu_device.cc:1471] Adding visible gpu devices: 0, 1
2018-09-08 15:21:16.890123: I c:\users\user\source\repos\tensorflow\tensorflow\core\common_runtime\gpu\gpu_device.cc:952] Device interconnect StreamExecutor with strength 1 edge matrix:
2018-09-08 15:21:16.896153: I c:\users\user\source\repos\tensorflow\tensorflow\core\common_runtime\gpu\gpu_device.cc:958]      0 1
2018-09-08 15:21:16.900452: I c:\users\user\source\repos\tensorflow\tensorflow\core\common_runtime\gpu\gpu_device.cc:971] 0:   N N
2018-09-08 15:21:16.905858: I c:\users\user\source\repos\tensorflow\tensorflow\core\common_runtime\gpu\gpu_device.cc:971] 1:   N N
2018-09-08 15:21:16.909802: I c:\users\user\source\repos\tensorflow\tensorflow\core\common_runtime\gpu\gpu_device.cc:1084] Created TensorFlow device (/job:localhost/replica:0/task:0/device:GPU:0 with 4734 MB memory) -> physical GPU (device: 0, name: GeForce GTX 1060 6GB, pci bus id: 0000:01:00.0, compute capability: 6.1)
2018-09-08 15:21:16.920119: I c:\users\user\source\repos\tensorflow\tensorflow\core\common_runtime\gpu\gpu_device.cc:1084] Created TensorFlow device (/job:localhost/replica:0/task:0/device:GPU:1 with 4734 MB memory) -> physical GPU (device: 1, name: GeForce GTX 1060 6GB, pci bus id: 0000:05:00.0, compute capability: 6.1)
Neural network loaded !
loading_time = 5s
[st = 11114ms, mt = 29250ms , hply = 0 , moves_left 10]
63 16 112 3603  Nb1-c3 d7-d5 d2-d4 Ng8-f6 Ng1-f3
64 13 226 7133  d2-d4 d7-d5 e2-e3 Ng8-f6 Nb1-c3
65 12 341 9644  d2-d4 d7-d5 Ng1-f3 Ng8-f6 Nb1-c3
66 12 456 11525  d2-d4 d7-d5 Ng1-f3 Ng8-f6 Nb1-c3
67 12 573 13592  d2-d4 d7-d5 e2-e3 Ng8-f6 c2-c4 Bc8-g4 Ng1-f3
68 12 690 17588  d2-d4 d7-d5 Nb1-c3 Ng8-f6 Bc1-f4 h7-h6 e2-e3
69 13 805 20600  d2-d4 d7-d5 Nb1-c3 Ng8-f6 f2-f3 Nb8-c6 e2-e3 e7-e6 Ke1-f2 a7-a6
70 13 920 26230  d2-d4 d7-d5 Nb1-c3 Ng8-f6 Bc1-f4 Nf6-e4 e2-e3 Nb8-c6 Ng1-f3 e7-e6 h2-h3
71 13 1034 31579  d2-d4 d7-d5 Nb1-c3 g7-g6 Bc1-f4 Bf8-g7 a2-a3 Ng8-f6 f2-f3 e7-e6 e2-e4

#  1      2    2051 e2-e4 e7-e5 d2-d4 d7-d5 Nb1-c3 e5xd4 Nc3xd5 Nb8-c6 Ng1-e2 Bc8-g4 f2-f3 Bg4-e6 Bc1-f4 Qd8-h4
#  2     19   27726 d2-d4 d7-d5 Nb1-c3 g7-g6 Bc1-f4 Bf8-g7 a2-a3 Ng8-f6 f2-f3 a7-a6 e2-e4 d5xe4 f3xe4 Ke8-g8 h2-h3 Nb8-c6 Ng1-f3 Bc8-e6 d4-d5 Nf6-h5 Bf4-e3 Nh5-g3 d5xc6 Ng3xh1 c6xb7 Ra8-b8
#  3      3    1767 Nb1-c3 d7-d5 d2-d4 Ng8-f6 Bc1-d2 Nb8-c6 e2-e3 e7-e5 d4xe5 Nc6xe5 Ng1-e2 Bc8-g4 f2-f3
#  4      4    2859 Ng1-f3 d7-d5 Nb1-c3 Ng8-f6 h2-h3 Nb8-c6 d2-d4 e7-e6 e2-e3 h7-h6 Bf1-d3 a7-a6 Ke1-g1 Bf8-d6 Rf1-e1 Ke8-g8 a2-a3 Bc8-d7
#  5     -5    1513 e2-e3 d7-d5 d2-d4 Ng8-f6 Ng1-e2 Nb8-c6 Nb1-c3 e7-e6 Ne2-f4 Bf8-d6 a2-a3 h7-h6 Bf1-e2 a7-a6 Ke1-g1 Ke8-g8 Be2-f3 Bc8-d7 Nf4-d3 e6-e5 d4xe5
#  6      0    1297 d2-d3 Nb8-c6 Bc1-d2 d7-d5 e2-e4 e7-e5 e4xd5 Qd8xd5 Nb1-c3 Qd5-d8 Ng1-e2
#  7     -6     794 g2-g3 e7-e5 e2-e4 d7-d5 e4xd5 Qd8xd5 f2-f3 Nb8-c6 Nb1-c3 Qd5-d4 d2-d3 f7-f6 Ng1-e2 Qd4-b6 Bf1-g2 Ke8-f7 Nc3-d5 Bf8-b4 Bc1-d2 Qb6-a5
#  8    -12     646 b2-b3 d7-d5 d2-d4 Nb8-c6 Nb1-c3 a7-a6 Ng1-f3 Ng8-f6 Bc1-b2 e7-e6 e2-e3 h7-h6 h2-h3
#  9    -17     446 f2-f3 e7-e5 d2-d3 d7-d5 e2-e4 Ng8-f6 Nb1-c3 Bf8-c5 Nc3xd5 Ke8-g8 Bc1-e3 Bc5-d4
# 10    -16     505 h2-h3 d7-d5 d2-d4 Nb8-c6 Nb1-c3 Ng8-f6 Ng1-f3 e7-e6 e2-e3 h7-h6 Bf1-d3 a7-a6 Ke1-g1 Bf8-d6 Rf1-e1
# 11    -14     634 c2-c3 d7-d5 d2-d4 Nb8-c6 Nb1-d2 e7-e6 e2-e4 Bf8-d6 e4-e5 Bd6-e7 Nd2-b3 Bc8-d7 Bf1-d3 a7-a6 Ng1-f3 h7-h6 h2-h3 Ra8-c8 Bc1-f4 f7-f5 a2-a3
# 12    -12     598 a2-a3 d7-d5 d2-d4 e7-e6 Ng1-f3 Nb8-c6 Nb1-c3 Ng8-f6 e2-e3 h7-h6 Bf1-d3 a7-a6 Ke1-g1 Bf8-d6 e3-e4
# 13    -14     542 Nb1-a3 e7-e5 e2-e4 d7-d5 d2-d4 Ng8-f6 d4xe5 Nf6xe4 Na3-b5 Bf8-c5 Nb5-d4 Nb8-c6 Bf1-b5 Bc5xd4
# 14    -17     481 Ng1-h3 d7-d5 d2-d4 Nb8-c6 Nh3-g5 e7-e5 d4xe5 Nc6xe5 e2-e4 Ng8-f6 Nb1-c3 Nf6xe4 Nc3xe4 d5xe4 Ng5xe4 Bf8-b4 c2-c3
# 15    -12     591 f2-f4 d7-d5 d2-d3 c7-c5 e2-e4 Ng8-f6 Nb1-c3 e7-e6 Bc1-e3 Nb8-c6 Ng1-f3 d5-d4 e4-e5 d4xe3 e5xf6
# 16    -10     850 c2-c4 d7-d5 Nb1-c3 e7-e6 Ng1-f3 Ng8-f6 d2-d4 Nb8-c6 e2-e3 h7-h6 Bf1-d3 d5xc4 Bd3xc4 Bf8-d6 Ke1-g1 Ke8-g8 Bc4-d3 a7-a6 a2-a3 Bc8-d7 h2-h3 Qd8-e7 Rf1-e1 Ra8-d8 e3-e4 e6-e5
# 17    -17     442 h2-h4 e7-e5 e2-e4 d7-d6 d2-d3 Ng8-f6 Ng1-f3 Nb8-c6 Nb1-c3 Bc8-e6 Bc1-e3
# 18    -13     578 a2-a4 d7-d5 d2-d4 e7-e5 d4xe5 Nb8-c6 Nb1-c3 Bc8-e6 Bc1-e3 Nc6xe5 Ng1-f3 f7-f6 b2-b3
# 19    -17     436 g2-g4 d7-d5 e2-e3 f7-f6 Nb1-c3 e7-e5 d2-d4 Nb8-c6
# 20    -19     373 b2-b4 e7-e5 e2-e4 Ng8-f6 Nb1-c3 d7-d5 Bf1-b5 Bc8-d7 Bb5xd7 Nb8xd7

nodes = 5479807 <89% qnodes> time = 11329ms nps = 483697 eps = 364376 nneps = 3931
Tree: nodes = 1336084 depth = 32 pps = 3981 visits = 45110
      qsearch_calls = 9994 search_calls = 0
move d2d4
Bye Bye

D:\Chess Engines\ScorpioNN>

Code: Select all

D:\Chess Engines\ScorpioNN>scorpio use_nn 1 go quit
feature done=0
ht 4194304 X 16 = 64.0 MB
eht 524288 X 8 = 8.0 MB
pht 32768 X 24 = 0.8 MB
treeht 335539200 X 40 = 12799.8 MB
processors [128]
EgbbProbe 4.1 by Daniel Shawul
509 egbbs loaded !
Loading neural network ...
2018-09-08 15:20:45.528777: I c:\users\user\source\repos\tensorflow\tensorflow\core\platform\cpu_feature_guard.cc:141] Your CPU supports instructions that this TensorFlow binary was not compiled to use: AVX AVX2
2018-09-08 15:20:45.835804: I c:\users\user\source\repos\tensorflow\tensorflow\core\common_runtime\gpu\gpu_device.cc:1392] Found device 0 with properties:
name: GeForce GTX 1060 6GB major: 6 minor: 1 memoryClockRate(GHz): 1.835
pciBusID: 0000:01:00.0
totalMemory: 6.00GiB freeMemory: 4.97GiB
2018-09-08 15:20:45.978576: I c:\users\user\source\repos\tensorflow\tensorflow\core\common_runtime\gpu\gpu_device.cc:1392] Found device 1 with properties:
name: GeForce GTX 1060 6GB major: 6 minor: 1 memoryClockRate(GHz): 1.835
pciBusID: 0000:05:00.0
totalMemory: 6.00GiB freeMemory: 4.97GiB
2018-09-08 15:20:45.989946: I c:\users\user\source\repos\tensorflow\tensorflow\core\common_runtime\gpu\gpu_device.cc:1471] Adding visible gpu devices: 0, 1
2018-09-08 15:20:47.592534: I c:\users\user\source\repos\tensorflow\tensorflow\core\common_runtime\gpu\gpu_device.cc:952] Device interconnect StreamExecutor with strength 1 edge matrix:
2018-09-08 15:20:47.597685: I c:\users\user\source\repos\tensorflow\tensorflow\core\common_runtime\gpu\gpu_device.cc:958]      0 1
2018-09-08 15:20:47.602585: I c:\users\user\source\repos\tensorflow\tensorflow\core\common_runtime\gpu\gpu_device.cc:971] 0:   N N
2018-09-08 15:20:47.606590: I c:\users\user\source\repos\tensorflow\tensorflow\core\common_runtime\gpu\gpu_device.cc:971] 1:   N N
2018-09-08 15:20:47.610950: I c:\users\user\source\repos\tensorflow\tensorflow\core\common_runtime\gpu\gpu_device.cc:1084] Created TensorFlow device (/job:localhost/replica:0/task:0/device:GPU:0 with 4734 MB memory) -> physical GPU (device: 0, name: GeForce GTX 1060 6GB, pci bus id: 0000:01:00.0, compute capability: 6.1)
2018-09-08 15:20:47.877256: I c:\users\user\source\repos\tensorflow\tensorflow\core\common_runtime\gpu\gpu_device.cc:1084] Created TensorFlow device (/job:localhost/replica:0/task:0/device:GPU:1 with 4734 MB memory) -> physical GPU (device: 1, name: GeForce GTX 1060 6GB, pci bus id: 0000:05:00.0, compute capability: 6.1)
2018-09-08 15:20:48.172619: I c:\users\user\source\repos\tensorflow\tensorflow\core\common_runtime\gpu\gpu_device.cc:1471] Adding visible gpu devices: 0, 1
2018-09-08 15:20:48.177133: I c:\users\user\source\repos\tensorflow\tensorflow\core\common_runtime\gpu\gpu_device.cc:952] Device interconnect StreamExecutor with strength 1 edge matrix:
2018-09-08 15:20:48.185003: I c:\users\user\source\repos\tensorflow\tensorflow\core\common_runtime\gpu\gpu_device.cc:958]      0 1
2018-09-08 15:20:48.188497: I c:\users\user\source\repos\tensorflow\tensorflow\core\common_runtime\gpu\gpu_device.cc:971] 0:   N N
2018-09-08 15:20:48.192012: I c:\users\user\source\repos\tensorflow\tensorflow\core\common_runtime\gpu\gpu_device.cc:971] 1:   N N
2018-09-08 15:20:48.197520: I c:\users\user\source\repos\tensorflow\tensorflow\core\common_runtime\gpu\gpu_device.cc:1084] Created TensorFlow device (/job:localhost/replica:0/task:0/device:GPU:0 with 4734 MB memory) -> physical GPU (device: 0, name: GeForce GTX 1060 6GB, pci bus id: 0000:01:00.0, compute capability: 6.1)
2018-09-08 15:20:48.209015: I c:\users\user\source\repos\tensorflow\tensorflow\core\common_runtime\gpu\gpu_device.cc:1084] Created TensorFlow device (/job:localhost/replica:0/task:0/device:GPU:1 with 4734 MB memory) -> physical GPU (device: 1, name: GeForce GTX 1060 6GB, pci bus id: 0000:05:00.0, compute capability: 6.1)
Neural network loaded !
loading_time = 5s
[st = 11114ms, mt = 29250ms , hply = 0 , moves_left 10]
63 18 111 1403  e2-e4 e7-e5 d2-d4 e5xd4 Qd1xd4
64 15 226 4366  d2-d4 d7-d5 Nb1-c3 Nb8-c6 e2-e3
65 13 338 6921  d2-d4 d7-d5 Ng1-f3 Ng8-f6 Nb1-c3
66 13 454 9319  d2-d4 d7-d5 Nb1-c3 Ng8-f6 e2-e3 Nb8-c6
67 13 569 12084  d2-d4 d7-d5 Nb1-c3 Nb8-c6 e2-e3 Ng8-f6 Ng1-f3
68 13 681 15319  d2-d4 d7-d5 Nb1-c3 Ng8-f6 f2-f3 Nb8-c6 e2-e3 e7-e5 d4xe5 Nc6xe5
69 13 796 19304  d2-d4 d7-d5 Nb1-c3 Nb8-c6 Bc1-f4 Bc8-f5 Ng1-f3 e7-e6
70 13 911 25742  d2-d4 d7-d5 Nb1-c3 Nb8-c6 Bc1-f4 Bc8-f5 f2-f3 a7-a6 e2-e3 e7-e6
71 13 1026 31703  d2-d4 d7-d5 Nb1-c3 Nb8-c6 Bc1-f4 Bc8-f5 f2-f3 Ng8-f6 e2-e3 e7-e6 a2-a3 Bf8-d6 Ng1-e2 Ke8-g8 Bf4xd6 c7xd6

#  1      0    1673 e2-e4 e7-e5 d2-d4 d7-d5 Nb1-c3 e5xd4 Qd1xd4 Ng8-f6 Qd4-e5 Bc8-e6 Nc3xd5 Bf8-d6 Bf1-b5 c7-c6 Nd5xf6 g7xf6
#  2     19   25819 d2-d4 d7-d5 Nb1-c3 Nb8-c6 Bc1-f4 Bc8-f5 f2-f3 Ng8-f6 e2-e3 e7-e6 a2-a3 Bf8-d6 Bf1-b5 Ke8-g8 Bf4xd6 Qd8xd6 Ng1-e2 h7-h6 Ke1-g1 a7-a6 Bb5xc6 b7xc6 Ne2-f4 e6-e5
#  3      4    2227 Nb1-c3 d7-d5 d2-d4 Ng8-f6 a2-a3 Nb8-c6 Ng1-f3 e7-e6 e2-e3 h7-h6 Bf1-d3 a7-a6 Ke1-g1 Bf8-d6 Bd3-e2
#  4      6    3231 Ng1-f3 Ng8-f6 d2-d4 d7-d5 a2-a3 Nb8-c6 Nb1-c3 e7-e6 e2-e3 h7-h6 Bf1-d3 a7-a6 Ke1-g1 Bf8-d6 Qd1-e2
#  5     -5    1220 e2-e3 d7-d5 d2-d4 Ng8-f6 Ng1-e2 Nb8-c6 Nb1-c3 e7-e6 Ne2-f4 Bf8-d6 a2-a3 h7-h6 Bf1-e2 a7-a6 Ke1-g1 Ke8-g8 Be2-f3 Bc8-d7 Nf4-d3 e6-e5 d4xe5
#  6      0    1565 d2-d3 Nb8-c6 Bc1-d2 d7-d5 Ng1-f3 Ng8-f6 Nb1-c3 e7-e5 e2-e4 Bf8-c5 Nc3xd5 Ke8-g8 Bd2-e3 Nc6-d4 Nf3xe5 Nf6xd5
#  7     -5     972 g2-g3 e7-e5 e2-e4 d7-d5 e4xd5 Qd8xd5 f2-f3 Nb8-c6 Nb1-c3 Qd5-d4 d2-d3 f7-f6 Ng1-e2 Qd4-b6 Bf1-g2 Ke8-f7 Nc3-d5 Qb6-a5 Ne2-c3 Bc8-e6 Nd5-e3 a7-a6 Ke1-g1 Bf8-c5
#  8    -11     612 b2-b3 d7-d5 d2-d4 e7-e5 e2-e4 Ng8-f6 d4xe5 Nf6xe4 Bc1-b2 Bf8-b4 c2-c3 Bb4-c5 Bf1-b5 Nb8-c6 f2-f3 Qd8-h4 g2-g3 Ne4xg3
#  9    -17     487 f2-f3 e7-e5 d2-d3 d7-d5 e2-e4 Ng8-f6 Nb1-c3 d5-d4 Nc3-b5 Nb8-c6 Bc1-d2 a7-a6 Nb5-a3 h7-h6
# 10    -16     484 h2-h3 d7-d5 d2-d4 Ng8-f6 Ng1-f3 Nb8-c6 Nb1-c3 e7-e6 e2-e3 h7-h6 Bf1-d3 a7-a6 Ke1-g1 Bf8-d6 Rf1-e1
# 11    -13     556 c2-c3 d7-d5 d2-d4 Nb8-c6 e2-e4 e7-e6 e4-e5 Bc8-d7 Ng1-f3 a7-a6 Bf1-d3 f7-f6 Ke1-g1 f6xe5 Nf3xe5 Ng8-f6 Bc1-f4
# 12    -12     556 a2-a3 d7-d5 d2-d4 e7-e6 Ng1-f3 Nb8-c6 Nb1-c3 Ng8-f6 e2-e3 h7-h6 Bf1-d3 a7-a6 Ke1-g1 Bf8-d6 e3-e4
# 13    -16     530 Nb1-a3 d7-d5 d2-d4 e7-e5 d4xe5 Nb8-c6 Na3-b5 Nc6xe5 Qd1xd5 Bf8-d6 Nb5xd6 c7xd6 e2-e4 Ng8-f6 Qd5-b5 Ne5-c6 Bf1-d3
# 14    -17     503 Ng1-h3 d7-d5 d2-d4 Nb8-c6 Nh3-g5 Ng8-f6 Nb1-c3 a7-a6 e2-e3 e7-e5 d4xe5 Nc6xe5 Bf1-e2 Bc8-g4 Ke1-g1 h7-h6
# 15    -13     621 f2-f4 d7-d5 d2-d3 c7-c5 e2-e4 Ng8-f6 Nb1-c3 e7-e6 Bc1-e3 Nb8-c6 Ng1-f3 d5-d4 e4-e5 d4xe3 e5xf6 Qd8xf6 g2-g3
# 16     -8     787 c2-c4 d7-d5 c4xd5 e7-e6 d2-d4 e6xd5 Ng1-f3 Nb8-c6 Nb1-c3 Ng8-f6 e2-e3 h7-h6 Bf1-d3 a7-a6 Ke1-g1 Bf8-d6 e3-e4 Bc8-g4 Nc3xd5 Nc6xd4 Qd1-a4 Nd4-c6
# 17    -18     407 h2-h4 e7-e5 e2-e4 d7-d6 d2-d3 Nb8-c6 Ng1-f3 Bc8-e6 Nb1-c3 Ng8-f6 Bc1-e3 a7-a6 a2-a3 h7-h6 g2-g3 Bf8-e7 Bf1-g2 Ke8-g8 Ke1-g1
# 18    -13     566 a2-a4 d7-d5 d2-d4 c7-c6 e2-e3 Ng8-f6 Ng1-f3 h7-h6 Bf1-d3 e7-e6 Ke1-g1 Bf8-e7 h2-h3
# 19    -19     421 g2-g4 d7-d5 e2-e3 Ng8-f6 g4-g5 Nf6-e4 d2-d4 e7-e5 d4xe5 Ne4xg5 Bf1-g2 Bf8-b4 c2-c3 Bb4-e7 Qd1xd5
# 20    -20     365 b2-b4 e7-e5 Bc1-b2 Nb8-c6 Ng1-f3 Ng8-f6 e2-e3 d7-d5 Nf3xe5 Bf8xb4 Bf1-e2 Bc8-e6 Ne5xc6 b7xc6 Ke1-g1 Ke8-g8

nodes = 5289015 <89% qnodes> time = 11296ms nps = 468220 eps = 353303 nneps = 3807
Tree: nodes = 1284366 depth = 30 pps = 3858 visits = 43583
      qsearch_calls = 9354 search_calls = 0
move d2d4
Bye Bye
"Good decisions come from experience, and experience comes from bad decisions."
__________________________________________________________________
Ted Summers

Daniel Shawul
Posts: 3810
Joined: Tue Mar 14, 2006 10:34 am
Location: Ethiopia
Contact:

Re: Scorpio 2.8.7 MCTS+NN windows version

Post by Daniel Shawul » Sat Sep 08, 2018 7:29 pm

Fantastic!!

That is another thing off my shoulders.

Now I can go back to my linux machine and fix multi-gpu issues, as well as making the cpu to
use more cores efficiently.

Thank you very much !

Daniel

User avatar
Werner
Posts: 2438
Joined: Wed Mar 08, 2006 9:09 pm

Re: Scorpio 2.8.7 MCTS+NN windows version

Post by Werner » Sat Sep 08, 2018 7:41 pm

Hi Daniel,
I played some games against lc0 on 6x64 net too:

Lc0-1 9154 1CPU - ScorpioNet : 14,0/14 14-0-0 (11111111111111) 100% +1200 :(

by the way: can Scorpio read weights_9154.txt.gz renamed to net-6x64.pb ?
Werner

Daniel Shawul
Posts: 3810
Joined: Tue Mar 14, 2006 10:34 am
Location: Ethiopia
Contact:

Re: Scorpio 2.8.7 MCTS+NN windows version

Post by Daniel Shawul » Sat Sep 08, 2018 7:57 pm

Werner wrote:
Sat Sep 08, 2018 7:41 pm
Hi Daniel,
I played some games against lc0 on 6x64 net too:

Lc0-1 9154 1CPU - ScorpioNet : 14,0/14 14-0-0 (11111111111111) 100% +1200 :(

by the way: can Scorpio read weights_9154.txt.gz renamed to net-6x64.pb ?
Lol ok -- i guess i shoulda expected that.
Turn off neural network usage (use_nn=0) and use montecarlo search with those options I gave
before and it will probably beat Lc0 on CPU.

If you have to use NN on CPU, use the smallest one ( 2x32.pb net ) that should be much faster
than 6x64. Btw all my nets are trained with not more than 0.5 million games since I am mostly focused
on getting things working at the moment.

Daniel

Daniel Shawul
Posts: 3810
Joined: Tue Mar 14, 2006 10:34 am
Location: Ethiopia
Contact:

Re: Scorpio 2.8.7 MCTS+NN windows version

Post by Daniel Shawul » Sat Sep 08, 2018 8:43 pm

There maybe an issue with networks trained from PGN games (CCRL games) .

I have mostly been using another set of networks that I trained from set of EPD postions.
This set was also used for training standard scorpio's eval.

Could you try the net-2x32.pb file contained in https://github.com/dshawul/Scorpio/rele ... ts-epd.zip
It beat tscp three times in a row here on the cpu though it has some issues converting endgames.

These are trained from much less games (maybe 25000 games) but they should be able to capture atleast piece values.
My training script needs a lot more work for sure.
Here is the last game with the 2x32 net from nets-epd nor nets-ccrl. I figure this engine should be around 2400 from looking
at the way it plays but I am not chess player. It plays badly in the endgame though.



DAniel

User avatar
Werner
Posts: 2438
Joined: Wed Mar 08, 2006 9:09 pm

Re: Scorpio 2.8.7 MCTS+NN windows version

Post by Werner » Sun Sep 09, 2018 5:47 am

Daniel Shawul wrote:
Sat Sep 08, 2018 8:43 pm
Could you try the net-2x32.pb file contained in https://github.com/dshawul/Scorpio/rele ... ts-epd.zip
It beat tscp three times in a row here on the cpu though it has some issues converting endgames.
...sorry - same result.
Werner

User avatar
Werner
Posts: 2438
Joined: Wed Mar 08, 2006 9:09 pm

Re: Scorpio 2.8.7 MCTS+NN windows version

Post by Werner » Sun Sep 09, 2018 1:58 pm

I did not know, how strong these 6x64 net with LC0 is:

1 Lc0-1 9154 1CPU 3205 +436 111½1111111111101111 18.5/20
2 Scorpio 2.8 x64 1CPU 2769 -436 000½0000000000010000 1.5/20
perhaps I will make some more games with it...
Werner

User avatar
Werner
Posts: 2438
Joined: Wed Mar 08, 2006 9:09 pm

Re: Scorpio 2.8.7 MCTS+NN windows version

Post by Werner » Mon Sep 10, 2018 8:14 am

Werner wrote:
Sun Sep 09, 2018 1:58 pm
I did not know, how strong these 6x64 net with LC0 is:

1 Lc0-1 9154 1CPU 3205 +436 111½1111111111101111 18.5/20
2 Scorpio 2.8 x64 1CPU 2769 -436 000½0000000000010000 1.5/20
perhaps I will make some more games with it...
sorry - wrong configuration with Scorpio here - I have to repeat this match.
Werner

Post Reply