Skip to content

Add MXFP4 GGUF QuantizationType #1677

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
merged 4 commits into from
Aug 7, 2025
Merged

Add MXFP4 GGUF QuantizationType #1677

Show file tree
Hide file tree
Changes from 2 commits
Commits
Show all changes
4 commits
Select commit Hold shift + click to select a range
3887e3c
add MXFP4 QuantizationType
CISC Aug 7, 2025
5203864
add MXFP4 size and description
CISC Aug 7, 2025
f551091
fix test
CISC Aug 7, 2025
5af7077
Merge branch 'main' into gguf-mxfp4
ngxson Aug 7, 2025
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
5 changes: 5 additions & 0 deletions packages/gguf/src/quant-descriptions.ts
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -132,6 +132,10 @@ export const GGUF_QUANT_DESCRIPTIONS: Record<GGMLQuantizationType, { txt: string
txt: "Ternary quantization.",
src_url: "https://github.com/ggml-org/llama.cpp/pull/8151",
},
[GGMLQuantizationType.MXFP4]: {
txt: "4-bit Microscaling Block Floating Point.",
src_url: "https://github.com/ggml-org/llama.cpp/pull/15091",
},
};

const QK_K = 256;
Expand Down Expand Up @@ -173,4 +177,5 @@ export const GGML_QUANT_SIZES = {
[GGMLQuantizationType.BF16]: calcBPW(1, 2),
[GGMLQuantizationType.TQ1_0]: calcBPW(256, 2 + 4 * 13),
[GGMLQuantizationType.TQ2_0]: calcBPW(256, 2 + 64),
[GGMLQuantizationType.MXFP4]: calcBPW(32, 1 + 16),
};
3 changes: 3 additions & 0 deletions packages/tasks/src/gguf.ts
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -40,6 +40,7 @@ export enum GGMLFileQuantizationType {
Q4_0_8_8 = 35,
TQ1_0 = 36,
TQ2_0 = 37,
MXFP4_MOE = 38,

// custom quants used by unsloth
// they are not officially a scheme enum value in GGUF, but only here for naming
Expand Down Expand Up @@ -95,6 +96,7 @@ export const GGUF_QUANT_ORDER: GGMLFileQuantizationType[] = [
GGMLFileQuantizationType.Q4_1,
GGMLFileQuantizationType.Q4_2,
GGMLFileQuantizationType.Q4_3,
GGMLFileQuantizationType.MXFP4_MOE,

// 3-bit quantizations
GGMLFileQuantizationType.Q3_K_XL,
Expand Down Expand Up @@ -197,4 +199,5 @@ export enum GGMLQuantizationType {
BF16 = 30,
TQ1_0 = 34,
TQ2_0 = 35,
MXFP4 = 39,
}
Loading