Lumpiasty/llama.cpp
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity

model: Add MiniMax-M3 (MSA: MiniMax Sparse Attention) support (#24908)

Browse Source 
* Add preliminary MiniMax-M3 support

Text-only port that re-uses existing components: MiniMax-M2 style GQA with
per-head QK-norm and partial rotary, DeepSeek-V3 style leading-dense and
routed/shared experts, and swigluoai activation. Sparse attention is not
yet supported (dense fallback); vision tower and MTP heads are dropped.

* MiniMax-M3 vision tower (mmproj + clip graph)

* Delete m3_vision_ref.py

* Update clip.cpp

* MSA

* Update constants.py

* Update minimax.py

* Cache creation. Working withotu flash attention

* Added flash attention for sparse layers

* Decomposed slow cpu OP into GPU + CPU ops. Massive speedup over long ctx

* Rewrote indexer op to be cuda native. Modified flash attention to match per group block picking

* Implement sparse attention calc out of stock ops.

* Fix a cache allocation and cont issue

* Fixed -fa auto crash, flagged debug spots

* Delete vocab.json

* Delete model.safetensors.index.json

* Delete generation_config.json

* Delete Minimax directory

* Handled multi stream case to fall back on Dense Attention

* Development scaffolding cleanup. No functional change to the decode or
4-way paths. Full debug harness remains at <8136a9c68ed7a5eb009aa67bba3fda8062f4648f> for reproducing the
selection-parity validation.

* Remove redundant comment from minimax-m3.cpp

* Changed 3 Gelu Ops for vision into Gelu_erf ops

* Assert that n_kv is multiple of 128

* Rename MSA index tensors to indexer convention

Note: All GGUFs generated before this change will need to be regenerated.

* Fix incorrect Assert

* Review driven changes (#3)

* Remove comment from conversion minimax.py

Co-authored-by: Sigbjørn Skjæret <1629204+CISC@users.noreply.github.com>

* Remove whitespaces from constants.py

Co-authored-by: Sigbjørn Skjæret <1629204+CISC@users.noreply.github.com>

* Tighten comment in minimax.py

Co-authored-by: Sigbjørn Skjæret <1629204+CISC@users.noreply.github.com>

* inherit MiniMax-M3 from MiniMax-M2

* drop dead text_config fallbacks

* Add indexer writer methods

* Reuse LLM_FFN_SWIGLU_OAI_MOE

* Remove duplicate  indexer setters, add only block_size/local_blocks, follow value naming convention

* Fix conversion error /gguf_writer.py

Co-authored-by: Sigbjørn Skjæret <1629204+CISC@users.noreply.github.com>

* Update gguf-py/gguf/gguf_writer.py

Co-authored-by: Sigbjørn Skjæret <1629204+CISC@users.noreply.github.com>

* Update gguf-py/gguf/tensor_mapping.py

Co-authored-by: Sigbjørn Skjæret <1629204+CISC@users.noreply.github.com>

* Update conversion/minimax.py

Co-authored-by: Sigbjørn Skjæret <1629204+CISC@users.noreply.github.com>

* Update conversion/minimax.py

Co-authored-by: Sigbjørn Skjæret <1629204+CISC@users.noreply.github.com>

* Remove whitespace in src/llama-kv-cache.cpp

Co-authored-by: Sigbjørn Skjæret <1629204+CISC@users.noreply.github.com>

* Remove Whitespace in Update src/llama-model.h

Co-authored-by: Sigbjørn Skjæret <1629204+CISC@users.noreply.github.com>

* Remove whitespace in src/llama-hparams.h

Co-authored-by: Sigbjørn Skjæret <1629204+CISC@users.noreply.github.com>

* remove multimodal code upon maintainer request. Will be made as a separate PR

* Whitespace clean in tensor_mapping.py

* Log cache size on launch, block ctx shift, support prompt caching

Log indexer cache size on launch

Disallow ctx shift

Support prompt caching

* Update minimax-m3.cpp

* Optimize implementation, add multi stream support. 

Fully rewrote minimax-m3.cpp for speed and buffer size gains:

Unified the 4-way + decode, 1 FA call per layer instead of 4, with the groups mapped onto ne[3]

Custom CPU op now emits block-level mask, expanded on GPU, which causes CPU to GPU transfer to shrinks at prefill

Decode: ~25 nodes/layer vs ~50, no per-group concats/conts

Unified selection semantics, so both regimes rank bs + local bias (position-anchored local force), which means prefill/decode can no longer disagree on selection

can_reuse on the MSA bias input. Graph reuse at decode restored (was rebuilding the full graph every token)

In-place mask adds, shrinking compute buffer ~6.8 to ~4.2 GiB at ub2048/62k

Multi-stream: MSA now runs with -np N when kv_unified=false. Decode stays batched across streams (still 1 FA call), prefill loops per stream. dense fallback only for --kv-unified + multi-seq

Measured effect on expert offload bound setup: decode 6.2(4WAY)–7.15(MSA_decode) -> 7.7~7.8 t/s, flat from 5k to 60k+. prefill around 10% faster. buffer about 20% smaller, multi-user support.

* set default cache type to F32

* Fix potential DSA double indexer cache  allocation bug, only allocate in-cache k_idx for archs that opt in

* remove F16 downcasts in MSA attention, force F32 indexer score accum

* Add Minimax eos to llama vocab

* Guard edge case where idx cache can become stale after a tail trim

* Update llama-kv-cache.h

* Update llama-kv-cache.cpp

* Update llama-kv-cache.cpp

* Update llama-kv-cache.h

* Update llama-kv-cache.cpp

* Review driven changes

* style fix

* indexer hparams are required

* fix tests

* fix lint

---------

Co-authored-by: Daniel Han <danielhanchen@gmail.com>
Co-authored-by: Sigbjørn Skjæret <1629204+CISC@users.noreply.github.com>
Co-authored-by: Xuan Son Nguyen <son@huggingface.co>
This commit is contained in:
timkhronos
2026-07-26 19:43:45 +02:00
committed by GitHub
co-authored by Sigbjørn Skjæret Daniel Han Xuan Son Nguyen
parent 42fc243060
commit b1d4c65524
21 changed files with 1044 additions and 23 deletions
Download Patch File Download Diff File Expand all files Collapse all files
src/llama-kv-cache.h
+10
View File
@@ -173,10 +173,12 @@ public:
// get views of the current state of the cache
ggml_tensor * get_k(ggml_context * ctx, int32_t il, uint32_t n_kv, const slot_info & sinfo) const;
ggml_tensor * get_v(ggml_context * ctx, int32_t il, uint32_t n_kv, const slot_info & sinfo) const;
ggml_tensor * get_k_idx(ggml_context * ctx, int32_t il, uint32_t n_kv, const slot_info & sinfo) const;
// store k_cur and v_cur in the cache based on the provided head location
ggml_tensor * cpy_k(ggml_context * ctx, ggml_tensor * k_cur, ggml_tensor * k_idxs, int32_t il, const slot_info & sinfo) const;
ggml_tensor * cpy_v(ggml_context * ctx, ggml_tensor * v_cur, ggml_tensor * v_idxs, int32_t il, const slot_info & sinfo) const;
ggml_tensor * cpy_k_idx(ggml_context * ctx, ggml_tensor * k_idx_cur, ggml_tensor * k_idxs, int32_t il, const slot_info & sinfo) const;
//
// preparation API
@@ -228,9 +230,11 @@ private:
ggml_tensor * k;
ggml_tensor * v;
ggml_tensor * k_idx; // MSA single-head indexer keys, F32
std::vector<ggml_tensor *> k_stream;
std::vector<ggml_tensor *> v_stream;
std::vector<ggml_tensor *> k_idx_stream;
};
bool v_trans = true; // the value tensor is transposed
@@ -259,6 +263,9 @@ private:
// env: LLAMA_KV_CACHE_DEBUG
int debug = 0;
// set when a k_idx (indexer) cache exists and the stream layout supports MSA (single seq, or one stream per seq)
bool msa_strict_slots = false;
// this is the SWA type of the cache - not to be confused with the model SWA type
const llama_swa_type swa_type = LLAMA_SWA_TYPE_NONE;
@@ -291,6 +298,7 @@ private:
size_t size_k_bytes() const;
size_t size_v_bytes() const;
size_t size_k_idx_bytes() const;
ggml_tensor * build_rope_shift(
const llama_cparams & cparams,
@@ -370,6 +378,7 @@ public:
// get views of the current state of the cache
ggml_tensor * get_k(ggml_context * ctx, int32_t il) const;
ggml_tensor * get_v(ggml_context * ctx, int32_t il) const;
ggml_tensor * get_k_idx(ggml_context * ctx, int32_t il) const;
// store k_cur and v_cur in the cache based on the provided head location
// note: the heads in k_cur and v_cur should be laid out contiguously in memory
@@ -379,6 +388,7 @@ public:
// - v_idxs [n_tokens] or [n_tokens*n_embd_v_gqa] depending if V cache is transposed
ggml_tensor * cpy_k(ggml_context * ctx, ggml_tensor * k_cur, ggml_tensor * k_idxs, int32_t il) const;
ggml_tensor * cpy_v(ggml_context * ctx, ggml_tensor * v_cur, ggml_tensor * v_idxs, int32_t il) const;
ggml_tensor * cpy_k_idx(ggml_context * ctx, ggml_tensor * k_idx_cur, ggml_tensor * k_idxs, int32_t il) const;
// create destination indices for each head of the current batch for where it would be written in the KV cache
// the indices address the global KV cache (not per stream) - this is not relevant for the user of this API, but