# Antarmuka: QuantizationConfig

## Hierarki

- **`QuantizationConfig`**

  ↳ [`MlxQuantizationConfig`](MlxQuantizationConfig)

## Properti

### bits

• `Optional` **bits**: `number`

#### Didefinisikan di[[bits.defined-in]]

[packages/hub/src/lib/parse-safetensors-metadata.ts:930](https://github.com/huggingface/huggingface.js/blob/main/packages/hub/src/lib/parse-safetensors-metadata.ts#L930)

___

### config\_groups

• `Optional` **config\_groups**: `Record`\<`string`, \{ `format?`: `string` ; `targets?`: `string`[] ; `weights?`: \{ `num_bits?`: `number`  }  }\>

#### Didefinisikan di[[configgroups.defined-in]]

[packages/hub/src/lib/parse-safetensors-metadata.ts:935](https://github.com/huggingface/huggingface.js/blob/main/packages/hub/src/lib/parse-safetensors-metadata.ts#L935)

___

### expert\_dtype

• `Optional` **expert\_dtype**: `string`

Presisi pakar yang di-routed ketika berbeda dari kuantizer utama (misalnya pakar FP4 dengan perhatian FP8).

#### Didefinisikan di[[expertdtype.defined-in]]

[packages/hub/src/lib/parse-safetensors-metadata.ts:920](https://github.com/huggingface/huggingface.js/blob/main/packages/hub/src/lib/parse-safetensors-metadata.ts#L920)

___

### format

• `Optional` **format**: `string`

#### Didefinisikan di[[format.defined-in]]

[packages/hub/src/lib/parse-safetensors-metadata.ts:934](https://github.com/huggingface/huggingface.js/blob/main/packages/hub/src/lib/parse-safetensors-metadata.ts#L934)

___

### group\_size

• **group\_size** `Optional`: `number`

#### Didefinisikan di[[groupsize.defined-in]]

[packages/hub/src/lib/parse-safetensors-metadata.ts:928](https://github.com/huggingface/huggingface.js/blob/main/packages/hub/src/lib/parse-safetensors-metadata.ts#L928)

___

### ignore

• `Optional` **abaikan**: `string`[]

compressed-tensors menamai daftar pengecualian sebagai `ignore` alih-alih `modules_to_not_convert`, menggunakan sintaks target berawalan `re:` yang sama dengan `config_groups[].targets`.

#### Didefinisikan di[[ignore.defined-in]]

[packages/hub/src/lib/parse-safetensors-metadata.ts:940](https://github.com/huggingface/huggingface.js/blob/main/packages/hub/src/lib/parse-safetensors-metadata.ts#L940)

___

### load\_in\_4bit

• `Optional` **load\_in\_4bit**: `boolean`

#### Didefinisikan di[[loadin4bit.defined-in]]

[packages/hub/src/lib/parse-safetensors-metadata.ts:931](https://github.com/huggingface/huggingface.js/blob/main/packages/hub/src/lib/parse-safetensors-metadata.ts#L931)

___

### load\_in\_8bit

• `Optional` **load\_in\_8bit**: `boolean`

#### Didefinisikan di[[loadin8bit.defined-in]]

[packages/hub/src/lib/parse-safetensors-metadata.ts:932](https://github.com/huggingface/huggingface.js/blob/main/packages/hub/src/lib/parse-safetensors-metadata.ts#L932)

___

### mode

• `Optional` **mode**: `string`

Mode kuantisasi MLX (misalnya `affine`); konfigurasi MLX tidak mendeklarasikan `quant_method`.

#### Didefinisikan di[[mode.defined-in]]

[packages/hub/src/lib/parse-safetensors-metadata.ts:927](https://github.com/huggingface/huggingface.js/blob/main/packages/hub/src/lib/parse-safetensors-metadata.ts#L927)

___

### modul\_ke\_tidak\_konversi

• `Optional` **modules\_to\_not\_convert**: `string`[]

#### Didefinisikan di[[modulestonotconvert.defined-in]]

[packages/hub/src/lib/parse-safetensors-metadata.ts:929](https://github.com/huggingface/huggingface.js/blob/main/packages/hub/src/lib/parse-safetensors-metadata.ts#L929)

___

### quant\_algo

• `Optional` **quant\_algo**: `string`

Algoritma NVIDIA ModelOpt (`quant_method: "modelopt"`), misalnya `"NVFP4"`, `"W4A16_NVFP4"`, `"FP8"`, atau `"MIXED_PRECISION"` — di mana algoritma per modul terdapat di `quantized_layers`.

#### Didefinisikan di[[quantalgo.defined-in]]

[packages/hub/src/lib/parse-safetensors-metadata.ts:945](https://github.com/huggingface/huggingface.js/blob/main/packages/hub/src/lib/parse-safetensors-metadata.ts#L945)

___

### quant\_method

• `Optional` **quant\_method**: `string`

#### Didefinisikan di[[quantmethod.defined-in]]

[packages/hub/src/lib/parse-safetensors-metadata.ts:918](https://github.com/huggingface/huggingface.js/blob/main/packages/hub/src/lib/parse-safetensors-metadata.ts#L918)

___

### quantized\_layers

• `Optional` **lapisan_kuantisasi\_**: `Record`\<`string`, \{ `quant_algo?`: `string`  }\>

ModelOpt mixed-precision: nama modul (atau modul induk, mis. `…mlp.experts`) -> algoritmanya.

#### Didefinisikan di[[quantizedlayers.defined-in]]

[packages/hub/src/lib/parse-safetensors-metadata.ts:947](https://github.com/huggingface/huggingface.js/blob/main/packages/hub/src/lib/parse-safetensors-metadata.ts#L947)

___

### store\_dtype

• `Optional` **store\_dtype**: `string`

Peran yang sama dengan `expert_dtype` di bawah nama lain: MiMo-V2.6 menggunakan `quant_method: "fp8"` untuk lapisan dense-nya, tetapi menyimpan expert yang di-rutekan sebagai `store_dtype: "mxfp4"`, dikemas dua per `U8`.

#### Didefinisikan di[[storedtype.defined-in]]

[packages/hub/src/lib/parse-safetensors-metadata.ts:925](https://github.com/huggingface/huggingface.js/blob/main/packages/hub/src/lib/parse-safetensors-metadata.ts#L925)