github/gpt4all
1
0
Fork 0
mirror of https://github.com/nomic-ai/gpt4all.git synced 2025-09-17 16:28:20 +00:00
Code Issues Packages Projects Releases Wiki Activity

expose n_gpu_layers parameter of llama.cpp (#1890)

Browse Source 
Also dynamically limit the GPU layers and context length fields to the maximum supported by the model.

Signed-off-by: Jared Van Bortel <jared@nomic.ai>
This commit is contained in:
Jared Van Bortel
2024-01-31 14:17:44 -05:00
committed by GitHub
parent f549d5a70a
commit 061d1969f8
31 changed files with 381 additions and 157 deletions
Download Patch File Download Diff File Expand all files Collapse all files

2
gpt4all-bindings/csharp/Gpt4All/Bindings/LLModel.cs
View File

@@ -183,7 +183,7 @@ public class LLModel : ILLModel
/// <returns>true if the model was loaded successfully, false otherwise.</returns>
public bool Load(string modelPath)
{
return NativeMethods.llmodel_loadModel(_handle, modelPath, 2048);
return NativeMethods.llmodel_loadModel(_handle, modelPath, 2048, 100);
}
protected void Destroy()

3
gpt4all-bindings/csharp/Gpt4All/Bindings/NativeMethods.cs
View File

@@ -71,7 +71,8 @@ internal static unsafe partial class NativeMethods
public static extern bool llmodel_loadModel(
[NativeTypeName("llmodel_model")] IntPtr model,
[NativeTypeName("const char *")][MarshalAs(UnmanagedType.LPUTF8Str)] string model_path,
[NativeTypeName("int32_t")] int n_ctx);
[NativeTypeName("int32_t")] int n_ctx,
[NativeTypeName("int32_t")] int ngl);
[DllImport("libllmodel", CallingConvention = CallingConvention.Cdecl, ExactSpelling = true)]

2
gpt4all-bindings/csharp/Gpt4All/Model/Gpt4AllModelFactory.cs
View File

@@ -43,7 +43,7 @@ public class Gpt4AllModelFactory : IGpt4AllModelFactory
}
_logger.LogDebug("Model created handle=0x{ModelHandle:X8}", handle);
_logger.LogInformation("Model loading started");
var loadedSuccessfully = NativeMethods.llmodel_loadModel(handle, modelPath, 2048);
var loadedSuccessfully = NativeMethods.llmodel_loadModel(handle, modelPath, 2048, 100);
_logger.LogInformation("Model loading completed success={ModelLoadSuccess}", loadedSuccessfully);
if (!loadedSuccessfully)
{