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

spec : add DFlash2 support (local convolution + candidate selector) (#27342) (#27816)

Browse Source 
* spec : add DFlash2 support (local convolution + candidate selector) (#27342)

* support DFlash2

* Add p_min in DFlash2

Assisted-by: Claude Opus 5

* Revert unnecessary changes

Assisted-by: Claude Opus 5

* Revert draft sampling in rejection sampling

Assisted-by: Claude Opus 5

* Refactor code structure

Assisted-by: Claude Opus 5

* Delete embedding scaling

Assisted-by: Claude Opus 5

* Gate output transforms on DFlash2

Assisted-by: Claude Opus 5

* Optimize Dflash 2 cost

Assisted-by: Claude Opus 5

* Avoid using atoi

Assisted-by: Claude Opus 5

* Modify comments

Assisted-by: Claude Opus 5

* Move llama_model_dflash_selector_top_k to llama-ext.h

Assisted-by: Claude Opus 5

* Formatting

Assisted-by: Claude Opus 5

* Apply patch to fix the mrope bug

Assisted-by: Claude Opus 5

* fix ci

Assisted-by: Claude Opus 5

* Fix graph number calculation

Assisted-by: Claude Opus 5

* rename hid and unary

Assisted-by: Claude Opus 5

---------

Co-authored-by: Jian Chen <jianchen0311@gmail.com>
Co-authored-by: Xuan-Son Nguyen <son@huggingface.co>

* revert top-k.cu changes

---------

Co-authored-by: Zihan Zhang <tiancaizhangdaxian@sjtu.edu.cn>
Co-authored-by: Jian Chen <jianchen0311@gmail.com>
This commit is contained in:
Xuan-Son Nguyen
2026-08-27 19:17:07 +02:00
committed by GitHub
co-authored by Jian Chen Zihan Zhang
parent 58546250cf
commit b10f9ca58c
15 changed files with 538 additions and 24 deletions
Download Patch File Download Diff File Expand all files Collapse all files
src/llama-arch.cpp
+20
View File
@@ -344,6 +344,12 @@ static const std::map<llm_kv, const char *> LLM_KV_NAMES = {
{ LLM_KV_NORM_BEFORE_RESIDUAL, "%s.norm_before_residual" },
{ LLM_KV_NORM_BEFORE_FC, "%s.norm_before_fc" },
{ LLM_KV_DFLASH_BLOCK_SIZE, "%s.block_size" },
{ LLM_KV_DFLASH_CONV_KERNEL_SIZE, "%s.conv_kernel_size" },
{ LLM_KV_DFLASH_CONV_GROUP_SIZE, "%s.conv_group_size" },
{ LLM_KV_DFLASH_SELECTOR_RANK, "%s.selector_rank" },
{ LLM_KV_DFLASH_SELECTOR_TOP_K, "%s.selector_top_k" },
{ LLM_KV_SHORTCONV_L_CACHE, "%s.shortconv.l_cache" },
// sentence-transformers dense modules feature dims
{ LLM_KV_DENSE_2_FEAT_IN, "%s.dense_2_feat_in" },
@@ -651,6 +657,13 @@ static const std::map<llm_tensor, const char *> LLM_TENSOR_NAMES = {
{ LLM_TENSOR_DSPARK_MARKOV_W1, "markov_w1" },
{ LLM_TENSOR_DSPARK_MARKOV_W2, "markov_w2" },
{ LLM_TENSOR_DSPARK_CONF_PROJ, "conf_proj" },
{ LLM_TENSOR_DFLASH_ATTN_CONV_BASE, "blk.%d.attn_conv_base" },
{ LLM_TENSOR_DFLASH_ATTN_CONV_PROJ, "blk.%d.attn_conv_proj" },
{ LLM_TENSOR_DFLASH_FFN_CONV_BASE, "blk.%d.ffn_conv_base" },
{ LLM_TENSOR_DFLASH_FFN_CONV_PROJ, "blk.%d.ffn_conv_proj" },
{ LLM_TENSOR_DFLASH_SELECTOR_PREV, "selector_predecessor" },
{ LLM_TENSOR_DFLASH_SELECTOR_NEXT, "selector_successor" },
{ LLM_TENSOR_DFLASH_SELECTOR_HIDDEN, "selector_hidden" },
};
// declare information about the model weight tensors:
@@ -916,6 +929,13 @@ static const std::map<llm_tensor, llm_tensor_info> LLM_TENSOR_INFOS = {
{LLM_TENSOR_DSPARK_MARKOV_W1, {LLM_TENSOR_LAYER_OUTPUT, GGML_OP_GET_ROWS}},
{LLM_TENSOR_DSPARK_MARKOV_W2, {LLM_TENSOR_LAYER_OUTPUT, GGML_OP_MUL_MAT}},
{LLM_TENSOR_DSPARK_CONF_PROJ, {LLM_TENSOR_LAYER_OUTPUT, GGML_OP_MUL_MAT}},
{LLM_TENSOR_DFLASH_ATTN_CONV_BASE, {LLM_TENSOR_LAYER_REPEATING, GGML_OP_MUL}},
{LLM_TENSOR_DFLASH_ATTN_CONV_PROJ, {LLM_TENSOR_LAYER_REPEATING, GGML_OP_MUL_MAT}},
{LLM_TENSOR_DFLASH_FFN_CONV_BASE, {LLM_TENSOR_LAYER_REPEATING, GGML_OP_MUL}},
{LLM_TENSOR_DFLASH_FFN_CONV_PROJ, {LLM_TENSOR_LAYER_REPEATING, GGML_OP_MUL_MAT}},
{LLM_TENSOR_DFLASH_SELECTOR_PREV, {LLM_TENSOR_LAYER_OUTPUT, GGML_OP_GET_ROWS}},
{LLM_TENSOR_DFLASH_SELECTOR_NEXT, {LLM_TENSOR_LAYER_OUTPUT, GGML_OP_GET_ROWS}},
{LLM_TENSOR_DFLASH_SELECTOR_HIDDEN, {LLM_TENSOR_LAYER_OUTPUT, GGML_OP_MUL_MAT}},
};
LLM_KV::LLM_KV(llm_arch arch, const char * suffix) : arch(arch), suffix(suffix) {}