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

model: add dots3-note (#27060)

Browse Source 
* text: conversion

* init impl

* address review comments

* fix rope

* move to a new llama_kv_cache_dsa_iswa
This commit is contained in:
Xuan-Son Nguyen
2026-08-21 19:52:34 +02:00
committed by GitHub
parent 873e5d8e39
commit 5a32f7b66e
20 changed files with 1412 additions and 9 deletions
Download Patch File Download Diff File Expand all files Collapse all files
tests/test-llama-archs.cpp
+21 -1
View File
@@ -104,6 +104,7 @@ static gguf_context_ptr get_gguf_ctx(const llm_arch arch, const bool moe) {
} else if (arch == LLM_ARCH_DEEPSEEK2
|| arch == LLM_ARCH_DEEPSEEK32
|| arch == LLM_ARCH_GLM_DSA
|| arch == LLM_ARCH_DOTS3NOTE
|| arch == LLM_ARCH_KIMI_LINEAR
|| arch == LLM_ARCH_BAILINGMOE3
|| arch == LLM_ARCH_KIMI_K3
@@ -166,6 +167,7 @@ static gguf_context_ptr get_gguf_ctx(const llm_arch arch, const bool moe) {
if (arch == LLM_ARCH_DEEPSEEK2
|| arch == LLM_ARCH_DEEPSEEK32
|| arch == LLM_ARCH_GLM_DSA
|| arch == LLM_ARCH_DOTS3NOTE
|| arch == LLM_ARCH_KIMI_LINEAR
|| arch == LLM_ARCH_BAILINGMOE3
|| arch == LLM_ARCH_KIMI_K3
@@ -175,6 +177,22 @@ static gguf_context_ptr get_gguf_ctx(const llm_arch arch, const bool moe) {
ms.add_kv(LLM_KV_ROPE_DIMENSION_COUNT, uint32_t(64));
ms.add_kv(LLM_KV_ATTENTION_KEY_LENGTH_MLA, uint32_t(192));
ms.add_kv(LLM_KV_ATTENTION_VALUE_LENGTH_MLA, uint32_t(128));
if (arch == LLM_ARCH_DOTS3NOTE) {
// SWA layers reuse the same MLA geometry as the full layers in this fixture
ms.add_kv(LLM_KV_ATTENTION_KV_LORA_RANK_SWA, uint32_t(512));
ms.add_kv(LLM_KV_ATTENTION_KEY_LENGTH_SWA, uint32_t(576));
ms.add_kv(LLM_KV_ATTENTION_VALUE_LENGTH_SWA, uint32_t(512));
ms.add_kv(LLM_KV_ATTENTION_KEY_LENGTH_MLA_SWA, uint32_t(192));
ms.add_kv(LLM_KV_ATTENTION_VALUE_LENGTH_MLA_SWA, uint32_t(128));
ms.add_kv(LLM_KV_ROPE_FREQ_BASE_SWA, 10000.0f);
// indexer on the full-attention layers (inverse of the swa pattern)
std::vector<uint32_t> indexer_types;
indexer_types.reserve(n_layer);
for (uint32_t il = 0; il < n_layer; il++) {
indexer_types.push_back(il % 2 ? 0 : 1);
}
ms.add_kv(LLM_KV_ATTENTION_INDEXER_TYPES, indexer_types);
}
} else if (arch == LLM_ARCH_MINIMAX_M3) {
// partial rotary: n_rot must not exceed the indexer key length (64)
ms.add_kv(LLM_KV_ROPE_DIMENSION_COUNT, uint32_t(64));
@@ -197,7 +215,8 @@ static gguf_context_ptr get_gguf_ctx(const llm_arch arch, const bool moe) {
ms.add_kv(LLM_KV_ROPE_FREQ_BASE_SWA, 10000.0f);
// SWA pattern: every 5th layer is full attention (matches E2B layer_types)
ms.add_kv(LLM_KV_ATTENTION_SLIDING_WINDOW_PATTERN, uint32_t(5));
} else if (arch == LLM_ARCH_COHERE2MOE || arch == LLM_ARCH_MIMO2 || arch == LLM_ARCH_STEP35 || arch == LLM_ARCH_MUSE_GLIMMER || arch == LLM_ARCH_GRANITE_SWA) {
} else if (arch == LLM_ARCH_COHERE2MOE || arch == LLM_ARCH_MIMO2 || arch == LLM_ARCH_STEP35 ||
arch == LLM_ARCH_MUSE_GLIMMER || arch == LLM_ARCH_GRANITE_SWA || arch == LLM_ARCH_DOTS3NOTE) {
std::vector<uint32_t> pattern;
pattern.reserve(n_layer);
for (uint32_t il = 0; il < n_layer; il++) {
@@ -365,6 +384,7 @@ static bool moe_mandatory(const llm_arch arch) {
case LLM_ARCH_DEEPSEEK:
case LLM_ARCH_DEEPSEEK2:
case LLM_ARCH_DEEPSEEK32:
case LLM_ARCH_DOTS3NOTE:
case LLM_ARCH_GLM4_MOE:
case LLM_ARCH_GLM_DSA:
case LLM_ARCH_EXAONE_MOE: