Open clmnsllwrmnn opened 6 years ago
Most likely you need to download VC Redistributable 2015. See instructions here: https://github.com/glinscott/leela-chess/wiki/Getting-Started
If that doesn't work run client -debug
for more info.
VC Redistributable already installed (forgot to mention this earlier).
C:\lczero\gpu-win>client -debug Args: [C:\lczero\gpu-win/lczero --weights=networks\0da6631265b9f9a69fde0f05e807aeaa0a459d6785fd9cbccd8bfc5590ec1880 -t1 --randomize -n -lC:\lczero\gpu-win/logs-8960/20180401170430.log --start=train 8960-0 1] Logging to C:\lczero\gpu-win/logs-8960/20180401170430.log. Using 1 thread(s). Generated 1924 moves Detecting residual layers...v1...64 channels...6 blocks. Initializing OpenCL. Detected 1 OpenCL platforms. Platform version: OpenCL 1.1 CUDA 6.5.51 Platform profile: FULL_PROFILE Platform name: NVIDIA CUDA Platform vendor: NVIDIA Corporation Device ID: 0 Device name: GeForce 210 Device type: GPU Device vendor: NVIDIA Corporation Device driver: 342.01 Device speed: 1402 MHz Device cores: 2 CU Device score: 1111 Selected platform: NVIDIA CUDA Selected device: GeForce 210 with OpenCL 1.1 capability. Loaded existing SGEMM tuning. Wavefront/Warp size: 32 Max workgroup size: 512 Max workgroup dimensions: 512 512 64 BLAS Core: Barcelona Created dirs data-8960-0 Found 0 existing chunks in data-8960-0/training 2018/04/01 17:04:59 exit status 3221226505
log-file is empty.
Try running: lczero -w networks\networkfilename
. And when it finishes loading, type go
a few times. If that works try typing train 1 1
.
C:\lczero\gpu-win>lczero -w networks\0da6631265b9f9a69fde0f05e807aeaa0a459d6785fd9cbccd8bfc5590ec1880 Using 2 thread(s). Generated 1924 moves Detecting residual layers...v1...64 channels...6 blocks. Initializing OpenCL. Detected 1 OpenCL platforms. Platform version: OpenCL 1.1 CUDA 6.5.51 Platform profile: FULL_PROFILE Platform name: NVIDIA CUDA Platform vendor: NVIDIA Corporation Device ID: 0 Device name: GeForce 210 Device type: GPU Device vendor: NVIDIA Corporation Device driver: 342.01 Device speed: 1402 MHz Device cores: 2 CU Device score: 1111 Selected platform: NVIDIA CUDA Selected device: GeForce 210 with OpenCL 1.1 capability. Loaded existing SGEMM tuning. Wavefront/Warp size: 32 Max workgroup size: 512 Max workgroup dimensions: 512 512 64 BLAS Core: Barcelona
nothing happens so far, cursor blinking. Typing...
go
System window opens with "lczero.exe funktioniert nicht mehr" (lczero.exe doesn't work anymore), only button: "Programm schließen" (Close Program)
Hm not sure what the problem is. Maybe update your video drivers? Or just use the cpu-win version instead of the gpu-win version.
Drivers are the most recent (from 2016 - no fancy new stuff); reinstalled twice with no effect.
cpu-win version works with one core (no idea how to change this) but is ridiculously slow (about 10 minutes/game on AMD Phenom II X4, 3,2 Ghz)
You can run multiple instances of the client to get it faster. Not sure how many depends on the machine.
im getting same problem,
Update your GPU drivers or reduce the amount of games played simultaneously.
2018/04/10 00:52:38 exit status 3221226505
im only running 1 client on gpu, it plays a couple games then throws that error
Checked the same issue for the same OpenCL platform version.
PS C:\Users\Igor\Documents\leela-chess\v0.6\gpu-win> .\client.exe -debug
Args: [C:\Users\Igor\Documents\leela-chess\v0.6\gpu-win/lczero --weights=networks\66916be98835ced4bb7023b94b2fae578eb2089b1b270f68e117ff7a7111bb44 -t1 --randomize -n -v800 -lC:\Users\Igor\Documents\leela-chess\v0.6\gpu-win/logs-7612/20180415183618.log --start=train 7612-0 1]
Logging to C:\Users\Igor\Documents\leela-chess\v0.6\gpu-win/logs-7612/20180415183618.log.
Using 1 thread(s).
Detecting residual layers...v2...128 channels...10 blocks.
Initializing OpenCL.
Detected 1 OpenCL platforms.
Platform version: OpenCL 1.1 CUDA 6.5.51
Platform profile: FULL_PROFILE
Platform name: NVIDIA CUDA
Platform vendor: NVIDIA Corporation
Device ID: 0
Device name: GeForce 8600 GT
Device type: GPU
Device vendor: NVIDIA Corporation
Device driver: 342.01
Device speed: 1188 MHz
Device cores: 4 CU
Device score: 1111
Selected platform: NVIDIA CUDA
Selected device: GeForce 8600 GT
with OpenCL 1.1 capability.
Loaded existing SGEMM tuning.
Wavefront/Warp size: 32
Max workgroup size: 512
Max workgroup dimensions: 512 512 64
BLAS Core: Core2
Found 0 existing chunks in data-7612-0/training
id name lczero v0.6
id author The LCZero Authors
uciok
Created dirs data-7612-0
2018/04/15 18:36:47 exit status 3221226505
PS C:\Users\Igor\Documents\leela-chess\v0.6\gpu-win> cat .\leelaz_opencl_tuning
0;XgemmBatched;128;16;128;16; -DKWG=32 -DKWI=2 -DMDIMA=32 -DMDIMC=32 -DMWG=32 -DNDIMB=8 -DNDIMC=8 -DNWG=16 -DSA=1 -DSB=1
-DSTRM=0 -DSTRN=0 -DVWM=1 -DVWN=2;OpenCL: NVIDIA Corporation GeForce 8600 GT @ 1188MHz
More details about GPU: gpu_caps_viewer_report.txt
The same error occurs for 0.7 version
Any ideas what went wrong?
Win 10, GeForce 210 OpenCL 1.1 client 0.4 Running Windows Defender but no difference if disabled and without any alerts; no other anti-virus/anti-malware/anti-something-SW installed.
(Leelaz_opencl_tuning: 0;XgemmBatched;64;16;64;16; -DKWG=32 -DKWI=2 -DMDIMA=32 -DMDIMC=32 -DMWG=32 -DNDIMB=8 -DNDIMC=8 -DNWG=16 -DSA=1 -DSB=1 -DSTRM=0 -DSTRN=0 -DVWM=1 -DVWN=2;OpenCL: NVIDIA Corporation GeForce 210 @ 1402MHz)