Skip to content

Annier.rocmfp4 support #86

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
jimw567 merged 3 commits into from
Aug 16, 2026
Merged

Annier.rocmfp4 support #86

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
3 commits
Select commit Hold shift + click to select a range
31796c3
rocmfp4: add Q4_0_ROCMFP4 quant type (M1 dequant + hipBLAS path)
Jul 31, 2026
01d4e4e
ggml-cuda: rocmfp4 MMVQ decode path (M2)
Jul 31, 2026
f81c5b0
Integarted rocmfp4 quantization from https://github.com/charlie12345/…
Aug 13, 2026
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
11 changes: 10 additions & 1 deletion ggml/include/ggml.h
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -430,7 +430,9 @@ extern "C" {
GGML_TYPE_NVFP4 = 40, // NVFP4 (4 blocks, E4M3 scale)
GGML_TYPE_Q1_0 = 41,
GGML_TYPE_Q2_0 = 42,
GGML_TYPE_COUNT = 43,
GGML_TYPE_Q4_0_ROCMFP4 = 100, // ROCmFP4 experimental UE4M3 scales + packed AMD FP4 blocks
GGML_TYPE_Q4_0_ROCMFP4_FAST = 101, // ROCmFP4 single-scale speed layout
GGML_TYPE_COUNT = 102,
};

// precision
Expand Down Expand Up @@ -475,6 +477,13 @@ extern "C" {
GGML_FTYPE_MOSTLY_NVFP4 = 26, // except 1d tensors
GGML_FTYPE_MOSTLY_Q1_0 = 27, // except 1d tensors
GGML_FTYPE_MOSTLY_Q2_0 = 28, // except 1d tensors
GGML_FTYPE_MOSTLY_Q4_0_ROCMFP4 = 100, // except 1d tensors
GGML_FTYPE_MOSTLY_Q4_0_ROCMFP4_LEAN = 101, // ROCmFP4 with Q5_K token embeddings
GGML_FTYPE_MOSTLY_Q4_0_ROCMFP4_COHERENT = 102, // ROCmFP4 with Q6_K token embeddings
GGML_FTYPE_MOSTLY_Q4_0_ROCMFP4_FAST = 103, // ROCmFP4 single-scale speed layout
GGML_FTYPE_MOSTLY_Q4_0_ROCMFP4_FAST_COHERENT = 104, // ROCmFP4 fast with Q6_K token embeddings
GGML_FTYPE_MOSTLY_Q4_0_ROCMFP4_STRIX = 105, // ROCmFP4 Strix Halo quality/speed recipe
GGML_FTYPE_MOSTLY_Q4_0_ROCMFP4_STRIX_LEAN = 106, // ROCmFP4 Strix Halo size-biased K/V recipe
};

// available tensor operations:
Expand Down
Loading
Loading