LLamaSharp/docs/xmldocs/llama.native.llamamodelquan...

92 lines
1.6 KiB
Markdown

# LLamaModelQuantizeParams
Namespace: LLama.Native
Quantizer parameters used in the native API
```csharp
public struct LLamaModelQuantizeParams
```
Inheritance [Object](https://docs.microsoft.com/en-us/dotnet/api/system.object) → [ValueType](https://docs.microsoft.com/en-us/dotnet/api/system.valuetype) → [LLamaModelQuantizeParams](./llama.native.llamamodelquantizeparams.md)
**Remarks:**
llama_model_quantize_params
## Fields
### **nthread**
number of threads to use for quantizing, if <=0 will use std::thread::hardware_concurrency()
```csharp
public int nthread;
```
### **ftype**
quantize to this llama_ftype
```csharp
public LLamaFtype ftype;
```
### **imatrix**
pointer to importance matrix data
```csharp
public IntPtr imatrix;
```
## Properties
### **allow_requantize**
allow quantizing non-f32/f16 tensors
```csharp
public bool allow_requantize { get; set; }
```
#### Property Value
[Boolean](https://docs.microsoft.com/en-us/dotnet/api/system.boolean)<br>
### **quantize_output_tensor**
quantize output.weight
```csharp
public bool quantize_output_tensor { get; set; }
```
#### Property Value
[Boolean](https://docs.microsoft.com/en-us/dotnet/api/system.boolean)<br>
### **only_copy**
only copy tensors - ftype, allow_requantize and quantize_output_tensor are ignored
```csharp
public bool only_copy { get; set; }
```
#### Property Value
[Boolean](https://docs.microsoft.com/en-us/dotnet/api/system.boolean)<br>
### **pure**
disable k-quant mixtures and quantize all tensors to the same type
```csharp
public bool pure { get; set; }
```
#### Property Value
[Boolean](https://docs.microsoft.com/en-us/dotnet/api/system.boolean)<br>