mikhail/llama.cpp-mtp-turboquant
1
0
Fork 0
Code Issues Pull Requests Actions 34 Packages Projects Releases Wiki Activity

Kimi-Linear support (backend agnostic + MLA KV cache) (#18755)

Browse Source 
* kimi linear model implementation

* kimi linear convert_hf_to_gguf

* kimi linear constants.py tensor_mapping.py

* Kimi Linear ggml.h

* kimi linear ggml-cpu

* Kimi Linear ggml-cuda

* Kimi Linear ggml.c

* kimi linear src/llama

* remove "const int64_t n_seq_tokens = q->ne[2];" to get rid of unused variable warning

* remove type mismatch warning

* read MoE params

* removed some hard coded code

* removed all hard code

* use DeepseekV2 tokenizer

* removed unnecessary internal methods called by the old set_vocab of KimiLinear

* rewrite get_vocab for KimiLinear. Removed all kda_scan code

* removed all traces of kda_scan

* reduce OP count by 1 due to removal of kda_scan

* Move KIMI_LINEAR to llm_arch_is_hybrid to enable KV cache

* set n_embd_head_k/v to ensure kv cache works

* don't quantize conv1d of Kimi Linear

* Kimi Linear backend agnostic

* removed LOG_INFO

* naive chunking form implemented

* fixed some comments

* add Kimi-K2 specific tokens to be recognized as EOG

* build_kda_autoregressive is implemented to replace build_kda_recurrent for faster inference. sync'd to b7682

* replaced Akk and Aqk with mul_mat and clamp

* no clamp version

* Moved Aqk computation out of the loop

* fixed typo and split wkv_b into wk_b and wv_b

* MLA KV cache support

* fix trailing spaces

* moved const llama_model & model; around to follow qwen3next format and see if it cna pass the -Wunused-private-field error

* fix trailing whitespace

* removed traling whitespaces in empty line + make sure indentation is multiple of 4

* try to make lint happy

* remove blank lines to make lint happy

* removed at least blank line containing white space

* fixed flake8 complaints locally

* return ggml_tensor * pair in kda_autoregressive and kda_chunking as in ngxson's Qwen3Next improvement

* removed Kimi-Linear specific change that causes failure at server-windows

* removed private: from kimi_linear to make build checks happy

* removed unnecessary ggml_cont before ggml_reshape

* created static function causal_conv1d to abtract similar code for q/k/v

* merged dt_bias to SSM_DT. Do -exp(log_A) in convert_hf_to_gguf.py.

* reverted to original

* fixed find_hparam calls. Fixed e_score_correction_bias to use bias instead of weight. Removed all ssm_conv bias terms.

* remove DT_B from constants.py. remove one comment line in llama-model.cpp

* new class llm_graph_input_mem_hybrid_k to get around the new MLA change. switch the concat order of ggml_concat calls in kimi-linear.cpp to accommodate MLA changes. Removed support for exp_probs_b.weight

* remove ssm_o_norm_b

* remove ssm_o_norm_b

* changed hparams.kda_head_dim to hparams.n_embd_head_kda. added TODO comment for class llama_graph_mem_hybrid_k

* removed all ggml_cont b4 ggml_reshape_4d

* Whitespace

* replaced all hparams.get with find_hparams

* added new names for n_experts, n_experts_used and score_func in TextModel and removed their code in KimiLinear in convert_hf_to_gguf.py. Removed unnecessary ggml_cont and GGML_ASSERT in kimi-linear.cpp

* use is_mla to switch between different mem_hybrid types

* fixed logical errors in convert_hf_to_gguf.py pointed out by CISC

* removed if else for required parameters kv_lora_rank and qk_rope_head_dim

* add back ggml_cont for Vcur

* minor changes

* removed extra line in llama-vocab.cpp. Added back the comment in llama-graph.cpp

* f16 gguf cannot run without context length

* made a mistake of adding back n_ctx parsing

---------

Co-authored-by: Piotr Wilkin (ilintar) <piotr.wilkin@syndatis.com>
This commit is contained in:
ymcki
2026-02-06 18:39:58 +08:00
committed by GitHub
parent 1946e46f4c
commit 3688c4f504
18 changed files with 1521 additions and 23 deletions
Download Patch File Download Diff File Expand all files Collapse all files
src/llama-arch.cpp
+70
View File
@@ -120,6 +120,7 @@ static const std::map<llm_arch, const char *> LLM_ARCH_NAMES = {
{ LLM_ARCH_MIMO2, "mimo2" },
{ LLM_ARCH_LLAMA_EMBED, "llama-embed" },
{ LLM_ARCH_MAINCODER, "maincoder" },
{ LLM_ARCH_KIMI_LINEAR, "kimi-linear" },
{ LLM_ARCH_UNKNOWN, "(unknown)" },
};
@@ -246,6 +247,8 @@ static const std::map<llm_kv, const char *> LLM_KV_NAMES = {
{ LLM_KV_SSM_GROUP_COUNT, "%s.ssm.group_count" },
{ LLM_KV_SSM_DT_B_C_RMS, "%s.ssm.dt_b_c_rms" },
{ LLM_KV_KDA_HEAD_DIM, "%s.kda.head_dim" },
{ LLM_KV_WKV_HEAD_SIZE, "%s.wkv.head_size" },
{ LLM_KV_POSNET_EMBEDDING_LENGTH, "%s.posnet.embedding_length" },
@@ -371,6 +374,15 @@ static const std::map<llm_tensor, const char *> LLM_TENSOR_NAMES = {
{ LLM_TENSOR_SSM_DT_NORM, "blk.%d.ssm_dt_norm" },
{ LLM_TENSOR_SSM_B_NORM, "blk.%d.ssm_b_norm" },
{ LLM_TENSOR_SSM_C_NORM, "blk.%d.ssm_c_norm" },
{ LLM_TENSOR_SSM_CONV1D_Q, "blk.%d.ssm_conv1d_q" },
{ LLM_TENSOR_SSM_CONV1D_K, "blk.%d.ssm_conv1d_k" },
{ LLM_TENSOR_SSM_CONV1D_V, "blk.%d.ssm_conv1d_v" },
{ LLM_TENSOR_SSM_F_A, "blk.%d.ssm_f_a" },
{ LLM_TENSOR_SSM_F_B, "blk.%d.ssm_f_b" },
{ LLM_TENSOR_SSM_BETA, "blk.%d.ssm_beta" },
{ LLM_TENSOR_SSM_G_A, "blk.%d.ssm_g_a" },
{ LLM_TENSOR_SSM_G_B, "blk.%d.ssm_g_b" },
{ LLM_TENSOR_SSM_NORM, "blk.%d.ssm_norm" },
{ LLM_TENSOR_ATTN_Q_A_NORM, "blk.%d.attn_q_a_norm" },
{ LLM_TENSOR_ATTN_KV_A_NORM, "blk.%d.attn_kv_a_norm" },
{ LLM_TENSOR_ATTN_Q_A, "blk.%d.attn_q_a" },
@@ -2289,6 +2301,54 @@ static std::set<llm_tensor> llm_get_tensor_names(llm_arch arch) {
LLM_TENSOR_FFN_DOWN,
LLM_TENSOR_FFN_UP,
};
case LLM_ARCH_KIMI_LINEAR:
return {
LLM_TENSOR_TOKEN_EMBD,
LLM_TENSOR_OUTPUT_NORM,
LLM_TENSOR_OUTPUT,
LLM_TENSOR_ROPE_FREQS,
LLM_TENSOR_ATTN_NORM,
LLM_TENSOR_ATTN_Q,
LLM_TENSOR_ATTN_K,
LLM_TENSOR_ATTN_V,
LLM_TENSOR_ATTN_OUT,
LLM_TENSOR_FFN_NORM,
// Dense FFN (layer 0 only)
LLM_TENSOR_FFN_GATE,
LLM_TENSOR_FFN_DOWN,
LLM_TENSOR_FFN_UP,
// MoE FFN (layers 1+)
LLM_TENSOR_FFN_GATE_INP,
LLM_TENSOR_FFN_GATE_EXPS,
LLM_TENSOR_FFN_DOWN_EXPS,
LLM_TENSOR_FFN_UP_EXPS,
LLM_TENSOR_FFN_EXP_PROBS_B,
// Shared experts
LLM_TENSOR_FFN_GATE_SHEXP,
LLM_TENSOR_FFN_DOWN_SHEXP,
LLM_TENSOR_FFN_UP_SHEXP,
// KDA (using SSM_ enum prefix, keeping GGUF names for backward compat)
LLM_TENSOR_SSM_CONV1D_Q,
LLM_TENSOR_SSM_CONV1D_K,
LLM_TENSOR_SSM_CONV1D_V,
LLM_TENSOR_SSM_F_A,
LLM_TENSOR_SSM_F_B,
LLM_TENSOR_SSM_BETA,
LLM_TENSOR_SSM_A,
LLM_TENSOR_SSM_G_A,
LLM_TENSOR_SSM_G_B,
LLM_TENSOR_SSM_DT,
LLM_TENSOR_SSM_NORM,
// MLA
LLM_TENSOR_ATTN_Q_A,
LLM_TENSOR_ATTN_Q_B,
LLM_TENSOR_ATTN_Q_A_NORM,
LLM_TENSOR_ATTN_KV_A_MQA,
LLM_TENSOR_ATTN_KV_B,
LLM_TENSOR_ATTN_K_B,
LLM_TENSOR_ATTN_V_B,
LLM_TENSOR_ATTN_KV_A_NORM,
};
default:
GGML_ABORT("unknown architecture for tensor mapping");
}
@@ -2392,6 +2452,15 @@ static const std::map<llm_tensor, llm_tensor_info> LLM_TENSOR_INFOS = {
{LLM_TENSOR_SSM_C_NORM, {LLM_TENSOR_LAYER_REPEATING, GGML_OP_MUL}},
{LLM_TENSOR_SSM_D, {LLM_TENSOR_LAYER_REPEATING, GGML_OP_MUL}},
{LLM_TENSOR_SSM_NORM, {LLM_TENSOR_LAYER_REPEATING, GGML_OP_MUL}},
// Kimi KDA - Conv tensors are 4D [d_conv, 1, d_inner, 1], reshaped to 2D at runtime
{LLM_TENSOR_SSM_CONV1D_Q, {LLM_TENSOR_LAYER_REPEATING, GGML_OP_MUL}},
{LLM_TENSOR_SSM_CONV1D_K, {LLM_TENSOR_LAYER_REPEATING, GGML_OP_MUL}},
{LLM_TENSOR_SSM_CONV1D_V, {LLM_TENSOR_LAYER_REPEATING, GGML_OP_MUL}},
{LLM_TENSOR_SSM_F_A, {LLM_TENSOR_LAYER_REPEATING, GGML_OP_MUL_MAT}},
{LLM_TENSOR_SSM_F_B, {LLM_TENSOR_LAYER_REPEATING, GGML_OP_MUL_MAT}},
{LLM_TENSOR_SSM_BETA, {LLM_TENSOR_LAYER_REPEATING, GGML_OP_MUL_MAT}},
{LLM_TENSOR_SSM_G_A, {LLM_TENSOR_LAYER_REPEATING, GGML_OP_MUL_MAT}},
{LLM_TENSOR_SSM_G_B, {LLM_TENSOR_LAYER_REPEATING, GGML_OP_MUL_MAT}},
{LLM_TENSOR_TIME_MIX_LERP_X, {LLM_TENSOR_LAYER_REPEATING, GGML_OP_MUL}},
{LLM_TENSOR_TIME_MIX_LN, {LLM_TENSOR_LAYER_REPEATING, GGML_OP_MUL}},
{LLM_TENSOR_CHANNEL_MIX_LERP_K, {LLM_TENSOR_LAYER_REPEATING, GGML_OP_MUL}},
@@ -2573,6 +2642,7 @@ bool llm_arch_is_hybrid(const llm_arch & arch) {
case LLM_ARCH_NEMOTRON_H:
case LLM_ARCH_NEMOTRON_H_MOE:
case LLM_ARCH_QWEN3NEXT:
case LLM_ARCH_KIMI_LINEAR:
return true;
default:
return false;