Trouble linking to Leela v0.11

[lukepharr]

Hi,

I am trying to run Leela v0.11 (leela_0110_macOS) on Sabaki, but am getting a "Connection Failed" error when I try to start the engine. I am using a MacBook running Catalina. I have KataGo running, so I've been able to get gtp working in the past. I am wondering if anyone knows what might be going wrong? I have attached photos of my engine setup on Sabaki.

Thanks in advance for your wonderful help.

Best

[kusumotonorio]

When I tried the same settings, I got the following message:

Nonsensical options: Playouts are restricted but thinking on the opponent's time is still allowed. Add --noponder if you want a weakened engine.

Adding a --noponder may give you good results.

[romton843]

I have the same issue with Leela GPU and Katago. I can have Elf CPU or CPU version working as well as Leela CPU version but not the GPU version of Leela nor Katago. Oh, and all of them work in command lines and Lizzie works with Leela GPU. No idea why, my posted issue got closed because no one was able to help. From my point of view, it might be a problem of how Sabaki and those engine are dealing with the graphic cards... I wish you luck and perhaps someone will figure this all out. Best regards. Romain

[kusumotonorio]

If you'd like, I'd like to see your settings in Sabaki and the instructions for launching it on the command line.

[romton843]

THank you Kusumotonorio, Here are my settings in Sabaki

Katago ‪C:\Go\katago\katago.exe gtp -model C:\Go\katago\g170e-b20c256x2-s2430231552-d525879064.bin.gz -config C:\Go\katago\gtp_custom.cfg

Leela Zero CPU C:\Go\leelacpu\leelaz.exe --gtp -w C:\Go\leelacpu\networks\98ae471207c26af3946880caf9daaaa2d3db77b1e212660f8ab6a2b2e11a21dd.gz --noponder time_settings 0 5 1

Leela Zero GPU ‪C:\Go\leelagpu\leelaz.exe --gtp -w C:\Go\leelagpu\networks\98ae471207c26af3946880caf9daaaa2d3db77b1e212660f8ab6a2b2e11a21dd.gz --noponder time_settings 0 5 1

Command line for Leela GPU: C:\Go\leelagpu\leelaz.exe --gtp -w C:\Go\leelagpu\networks\98ae471207c26af3946880caf9daaaa2d3db77b1e212660f8ab6a2b2e11a21dd.gz --noponder

Using OpenCL batch size of 5 Using 10 thread(s). RNG seed: 10321835993674782726 BLAS Core: Haswell Detecting residual layers...v1...256 channels...40 blocks. Initializing OpenCL (autodetecting precision). Detected 2 OpenCL platforms. Platform version: OpenCL 2.1 Platform profile: FULL_PROFILE Platform name: Intel(R) OpenCL HD Graphics Platform vendor: Intel(R) Corporation Device ID: 0 Device name: Intel(R) UHD Graphics 630 Device type: GPU Device vendor: Intel(R) Corporation Device driver: 26.20.100.7985 Device speed: 1100 MHz Device cores: 24 CU Device score: 621 Platform version: OpenCL 1.2 CUDA 11.0.228 Platform profile: FULL_PROFILE Platform name: NVIDIA CUDA Platform vendor: NVIDIA Corporation Device ID: 1 Device name: GeForce GTX 1050 Ti with Max-Q Design Device type: GPU Device vendor: NVIDIA Corporation Device driver: 452.06 Device speed: 1417 MHz Device cores: 6 CU Device score: 1112 Selected platform: NVIDIA CUDA Selected device: GeForce GTX 1050 Ti with Max-Q Design with OpenCL 1.2 capability. Half precision compute support: No. Tensor Core support: No. Detected 2 OpenCL platforms. Platform version: OpenCL 2.1 Platform profile: FULL_PROFILE Platform name: Intel(R) OpenCL HD Graphics Platform vendor: Intel(R) Corporation Device ID: 0 Device name: Intel(R) UHD Graphics 630 Device type: GPU Device vendor: Intel(R) Corporation Device driver: 26.20.100.7985 Device speed: 1100 MHz Device cores: 24 CU Device score: 621 Platform version: OpenCL 1.2 CUDA 11.0.228 Platform profile: FULL_PROFILE Platform name: NVIDIA CUDA Platform vendor: NVIDIA Corporation Device ID: 1 Device name: GeForce GTX 1050 Ti with Max-Q Design Device type: GPU Device vendor: NVIDIA Corporation Device driver: 452.06 Device speed: 1417 MHz Device cores: 6 CU Device score: 1112 Selected platform: NVIDIA CUDA Selected device: GeForce GTX 1050 Ti with Max-Q Design with OpenCL 1.2 capability. Half precision compute support: No. Tensor Core support: No. Loaded existing SGEMM tuning. Wavefront/Warp size: 32 Max workgroup size: 1024 Max workgroup dimensions: 1024 1024 64 Loaded existing SGEMM tuning. Wavefront/Warp size: 32 Max workgroup size: 1024 Max workgroup dimensions: 1024 1024 64 Using OpenCL single precision (less than 5% slower than half). Detected 2 OpenCL platforms. Platform version: OpenCL 2.1 Platform profile: FULL_PROFILE Platform name: Intel(R) OpenCL HD Graphics Platform vendor: Intel(R) Corporation Device ID: 0 Device name: Intel(R) UHD Graphics 630 Device type: GPU Device vendor: Intel(R) Corporation Device driver: 26.20.100.7985 Device speed: 1100 MHz Device cores: 24 CU Device score: 621 Platform version: OpenCL 1.2 CUDA 11.0.228 Platform profile: FULL_PROFILE Platform name: NVIDIA CUDA Platform vendor: NVIDIA Corporation Device ID: 1 Device name: GeForce GTX 1050 Ti with Max-Q Design Device type: GPU Device vendor: NVIDIA Corporation Device driver: 452.06 Device speed: 1417 MHz Device cores: 6 CU Device score: 1112 Selected platform: NVIDIA CUDA Selected device: GeForce GTX 1050 Ti with Max-Q Design with OpenCL 1.2 capability. Half precision compute support: No. Tensor Core support: No. Loaded existing SGEMM tuning. Wavefront/Warp size: 32 Max workgroup size: 1024 Max workgroup dimensions: 1024 1024 64 Setting max tree size to 3736 MiB and cache size to 415 MiB.

Here is my old thread with all the details: https://github.com/SabakiHQ/Sabaki/issues/629

[kusumotonorio]

Hmmm. I'm using Sabaki on a Mac, so I can't verify it on Windows, but the only other thing I can think of is about file execution permissions.

Is there a difference in file access rights between the actual CPU and GPU versions of Leela?