Sign in
Join now
llama-cpp-bonsai
okami.mom
/
Star
Starred
Star
Something went wrong. Try again.
Fork
0
Eye filled
Watch
Watching
Watch
llama.cpp fork for Ternary Bonsai 2 27B: MTP speculation, suffix cache reuse (SCR), q8 KV, two-slot fairness, fused DeltaNet kernels
Star
Starred
Star
Something went wrong. Try again.
Fork
Eye filled
Watch
Watching
Watch
Overview
Overview
Issues
Issues
Pulls
Pulls
Pipelines
Pipelines
llama-cpp-bonsai
/
src
/
models
152 files
at
main
nekomimi.pet
·
bonsai: fused CUDA kernel for DeltaNet conv path (gdn-conv)
·
2w ago
6d191706
afmoe.cpp
hparams : refactor `hparams.n_layer` (#24060)
4mo ago
apertus.cpp
hparams : refactor `hparams.n_layer` (#24060)
4mo ago
arcee.cpp
hparams : refactor `hparams.n_layer` (#24060)
4mo ago
arctic.cpp
hparams : refactor `hparams.n_layer` (#24060)
4mo ago
arwkv7.cpp
hparams : refactor `hparams.n_layer` (#24060)
4mo ago
baichuan.cpp
hparams : refactor `hparams.n_layer` (#24060)
4mo ago
bailingmoe.cpp
hparams : refactor `hparams.n_layer` (#24060)
4mo ago
bailingmoe2.cpp
hparams : refactor `hparams.n_layer` (#24060)
4mo ago
bailingmoe3.cpp
model : support DSpark for bailingmoe3 (#27508)
2mo ago
bert.cpp
hparams : refactor `hparams.n_layer` (#24060)
4mo ago
bitnet.cpp
hparams : refactor `hparams.n_layer` (#24060)
4mo ago
bloom.cpp
hparams : refactor `hparams.n_layer` (#24060)
4mo ago
chameleon.cpp
hparams : refactor `hparams.n_layer` (#24060)
4mo ago
chatglm.cpp
hparams : refactor `hparams.n_layer` (#24060)
4mo ago
clip.cpp
llama: Restore quantization of mmprojs (#26818)
2mo ago
codeshell.cpp
hparams : refactor `hparams.n_layer` (#24060)
4mo ago
cogvlm.cpp
hparams : refactor `hparams.n_layer` (#24060)
4mo ago
cohere2.cpp
Add arch support for cohere2-MoE (#24260)
4mo ago
cohere2moe.cpp
llama : load MTP tensors only if they are really used (#26296)
2mo ago
command-r.cpp
hparams : refactor `hparams.n_layer` (#24060)
4mo ago
dbrx.cpp
hparams : refactor `hparams.n_layer` (#24060)
4mo ago
deci.cpp
hparams : refactor `hparams.n_layer` (#24060)
4mo ago
deepseek.cpp
model : NvFP4 quantized LM head support (#23046)
5mo ago
deepseek2.cpp
model: use ggml_rope_set_offset() (#27382)
2mo ago
deepseek2ocr.cpp
hparams : refactor `hparams.n_layer` (#24060)
4mo ago
deepseek32.cpp
models : remove duplicate metadata load (#27378)
2mo ago
deepseek4.cpp
DeepseekV4: fix rollback with multi-seq (#26756)
2mo ago
delta-net-base.cpp
bonsai: fused CUDA kernel for DeltaNet conv path (gdn-conv)
2w ago
dflash.cpp
spec : add DFlash2 support (local convolution + candidate selector) (#27342) (#27816) (#261)
1mo ago
dots1.cpp
hparams : refactor `hparams.n_layer` (#24060)
4mo ago
dots3note.cpp
model: add dots3-note (#27060)
2mo ago
dream.cpp
hparams : refactor `hparams.n_layer` (#24060)
4mo ago
dspark.cpp
dspark: clarify runtime terminology
1mo ago
eagle3.cpp
spec: add eagle3-v3 support for gpt-oss model (#25794)
3mo ago
ernie4-5-moe.cpp
model : NvFP4 quantized LM head support (#23046)
5mo ago
ernie4-5.cpp
hparams : refactor `hparams.n_layer` (#24060)
4mo ago
eurobert.cpp
hparams : refactor `hparams.n_layer` (#24060)
4mo ago
exaone-moe.cpp
minor : fix lint issues (#24165)
4mo ago
exaone.cpp
hparams : refactor `hparams.n_layer` (#24060)
4mo ago
exaone4.cpp
model : fix SWA not being enabled for EXAONE 4.5 (#26848)
2mo ago
falcon-h1.cpp
hparams : refactor `hparams.n_layer` (#24060)
4mo ago
falcon.cpp
hparams : refactor `hparams.n_layer` (#24060)
4mo ago
gemma-embedding.cpp
hparams : refactor `hparams.n_layer` (#24060)
4mo ago
gemma.cpp
hparams : refactor `hparams.n_layer` (#24060)
4mo ago
gemma2.cpp
hparams : refactor `hparams.n_layer` (#24060)
4mo ago
gemma3.cpp
hparams : refactor `hparams.n_layer` (#24060)
4mo ago
gemma3n.cpp
hparams : refactor `hparams.n_layer` (#24060)
4mo ago
gemma4-assistant.cpp
mtp: support for gemma-4 E2B and E4B assistants (#24282)
4mo ago
gemma4.cpp
llama: move suppress_tokens handling to common/sampling (#26276)
3mo ago
glm-dsa.cpp
models : remove duplicate metadata load (#27378)
2mo ago
glm4-moe.cpp
model : support MTP in GLM-4.5-Air (#26534)
2mo ago
glm4.cpp
hparams : refactor `hparams.n_layer` (#24060)
4mo ago
gpt2.cpp
hparams : refactor `hparams.n_layer` (#24060)
4mo ago
gptneox.cpp
hparams : refactor `hparams.n_layer` (#24060)
4mo ago
granite-hybrid.cpp
model : GraniteSWAForCausalLM / GraniteMoeSWAForCausalLM (#25505)
2mo ago
granite-moe.cpp
model : GraniteSWAForCausalLM / GraniteMoeSWAForCausalLM (#25505)
2mo ago
granite-swa.cpp
model : GraniteSWAForCausalLM / GraniteMoeSWAForCausalLM (#25505)
2mo ago
granite-switch.cpp
model : GraniteSWAForCausalLM / GraniteMoeSWAForCausalLM (#25505)
2mo ago
granite.cpp
model : GraniteSWAForCausalLM / GraniteMoeSWAForCausalLM (#25505)
2mo ago
grok.cpp
hparams : refactor `hparams.n_layer` (#24060)
4mo ago
grovemoe.cpp
hparams : refactor `hparams.n_layer` (#24060)
4mo ago
hunyuan-dense.cpp
model: move `load_hparams` and `load_tensors` to per-model definition (#22004)
5mo ago
hunyuan-moe.cpp
hparams : refactor `hparams.n_layer` (#24060)
4mo ago
hunyuan-vl.cpp
model : NvFP4 quantized LM head support (#23046)
5mo ago
hy-v3.cpp
llama : load MTP tensors only if they are really used (#26296)
2mo ago
internlm2.cpp
hparams : refactor `hparams.n_layer` (#24060)
4mo ago
jais.cpp
hparams : refactor `hparams.n_layer` (#24060)
4mo ago
jais2.cpp
hparams : refactor `hparams.n_layer` (#24060)
4mo ago
jamba.cpp
hparams : refactor `hparams.n_layer` (#24060)
4mo ago
jina-bert-v2.cpp
hparams : refactor `hparams.n_layer` (#24060)
4mo ago
jina-bert-v3.cpp
hparams : refactor `hparams.n_layer` (#24060)
4mo ago
kimi-k3.cpp
model: add Kimi-K3 text model (#26185)
2mo ago
kimi-linear.cpp
hparams : refactor `hparams.n_layer` (#24060)
4mo ago
laguna.cpp
model: Add Laguna-S-2.1 LLM_TYPE (#26233)
3mo ago
lfm2.cpp
model : support DSpark for LFM2 models (#27383)
2mo ago
lfm2moe.cpp
hparams : refactor `hparams.n_layer` (#24060)
4mo ago
llada-moe.cpp
hparams : refactor `hparams.n_layer` (#24060)
4mo ago
llada.cpp
hparams : refactor `hparams.n_layer` (#24060)
4mo ago
llama-embed.cpp
model: move `load_hparams` and `load_tensors` to per-model definition (#22004)
5mo ago
llama.cpp
spec: add EAGLE3 speculative decoding support (#18039)
4mo ago
llama4.cpp
hparams : refactor `hparams.n_layer` (#24060)
4mo ago
maincoder.cpp
hparams : refactor `hparams.n_layer` (#24060)
4mo ago
mamba-base.cpp
mamba2 : Flatten in/out projections to dispatch GEMM instead of GEMV (#27513)
2mo ago
mamba.cpp
hparams : refactor `hparams.n_layer` (#24060)
4mo ago
mamba2.cpp
mamba2: remove hardcoded 2x expansion factor and invalid d_inner % d_state check (#23082)
4mo ago
mellum.cpp
hparams : refactor `hparams.n_layer` (#24060)
4mo ago
mimo2.cpp
model : load MiMo V2 MTP tensors only if used (#26412)
2mo ago
minicpm.cpp
hparams : refactor `hparams.n_layer` (#24060)
4mo ago
minicpm3.cpp
model: use ggml_rope_set_offset() (#27382)
2mo ago
minimax-01.cpp
model : add support for MiniMaxText01ForCausalLM and MiniMaxM1ForCausalLM (#27018)
2mo ago
minimax-m2.cpp
spec: add Minimax2 eagle3 support
3mo ago
minimax-m3.cpp
mtmd, common: various fixes (#27071)
2mo ago
mistral3.cpp
hparams : refactor `hparams.n_layer` (#24060)
4mo ago
mistral4.cpp
model: move `load_hparams` and `load_tensors` to per-model definition (#22004)
5mo ago
models.h
bonsai: fused CUDA kernel for DeltaNet conv path (gdn-conv)
2w ago
modern-bert.cpp
hparams : refactor `hparams.n_layer` (#24060)
4mo ago
mpt.cpp
hparams : refactor `hparams.n_layer` (#24060)
4mo ago
muse-glimmer.cpp
model: Muse Glimmer Support (#26841)
2mo ago
nanbeige.cpp
model: Add support for Nanbeige4.2 (#25994)
3mo ago
nemotron-h-moe.cpp
model: add MTP support for Nemotron model (#26725)
2mo ago
nemotron-h.cpp
Dflash support for nemotron-3.5 (#26905)
2mo ago
nemotron.cpp
hparams : refactor `hparams.n_layer` (#24060)
4mo ago
neo-bert.cpp
hparams : refactor `hparams.n_layer` (#24060)
4mo ago
nomic-bert-moe.cpp
hparams : refactor `hparams.n_layer` (#24060)
4mo ago
nomic-bert.cpp
hparams : refactor `hparams.n_layer` (#24060)
4mo ago
olmo.cpp
hparams : refactor `hparams.n_layer` (#24060)
4mo ago
olmo2.cpp
hparams : refactor `hparams.n_layer` (#24060)
4mo ago
olmoe.cpp
hparams : refactor `hparams.n_layer` (#24060)
4mo ago
openai-moe.cpp
spec: add eagle3-v3 support for gpt-oss model (#25794)
3mo ago
openelm.cpp
hparams : refactor `hparams.n_layer` (#24060)
4mo ago
orion.cpp
hparams : refactor `hparams.n_layer` (#24060)
4mo ago
paddleocr.cpp
model : NvFP4 quantized LM head support (#23046)
5mo ago
pangu-embed.cpp
hparams : refactor `hparams.n_layer` (#24060)
4mo ago
phi2.cpp
hparams : refactor `hparams.n_layer` (#24060)
4mo ago
phi3.cpp
hparams : refactor `hparams.n_layer` (#24060)
4mo ago
phimoe.cpp
hparams : refactor `hparams.n_layer` (#24060)
4mo ago
plamo.cpp
hparams : refactor `hparams.n_layer` (#24060)
4mo ago
plamo2.cpp
ggml : recurrent state rollback for ggml_ssm_scan (#26623)
2mo ago
plamo3.cpp
hparams : refactor `hparams.n_layer` (#24060)
4mo ago
plm.cpp
model: use ggml_rope_set_offset() (#27382)
2mo ago
pockettts.cpp
mtmd: support pocket-tts (#26871)
2mo ago
qwen.cpp
hparams : refactor `hparams.n_layer` (#24060)
4mo ago
qwen2.cpp
hparams : refactor `hparams.n_layer` (#24060)
4mo ago
qwen2moe.cpp
hparams : refactor `hparams.n_layer` (#24060)
4mo ago
qwen2vl.cpp
model : NvFP4 quantized LM head support (#23046)
5mo ago
qwen3.cpp
spec: add EAGLE3 speculative decoding support (#18039)
4mo ago
qwen35.cpp
bonsai: fused CUDA kernel for DeltaNet conv path (gdn-conv)
2w ago
qwen35moe.cpp
graph: six fusions for the hybrid-attention decode graph (-25% ops/token; Metal 2B +7.7%, 27B +2.6% decode; CUDA 9B +2.5%) (#165)
1mo ago
qwen3moe.cpp
spec: add EAGLE3 speculative decoding support (#18039)
4mo ago
qwen3next.cpp
model: MTP support for Qwen3-Next (#25589)
2mo ago
qwen3tts.cpp
mtmd: support Qwen3-TTS (note: breaking change to llama-tts binary) (#26254)
2mo ago
qwen3vl.cpp
mtmd: support Qwen3-TTS (note: breaking change to llama-tts binary) (#26254)
2mo ago
qwen3vlmoe.cpp
hparams : refactor `hparams.n_layer` (#24060)
4mo ago
refact.cpp
hparams : refactor `hparams.n_layer` (#24060)
4mo ago
rnd1.cpp
hparams : refactor `hparams.n_layer` (#24060)
4mo ago
rwkv6-base.cpp
models : deduplicate delta-net graphs for Qwen family (#19597)
8mo ago
rwkv6.cpp
hparams : refactor `hparams.n_layer` (#24060)
4mo ago
rwkv6qwen2.cpp
hparams : refactor `hparams.n_layer` (#24060)
4mo ago
rwkv7-base.cpp
models : deduplicate delta-net graphs for Qwen family (#19597)
8mo ago
rwkv7.cpp
hparams : refactor `hparams.n_layer` (#24060)
4mo ago
seed-oss.cpp
hparams : refactor `hparams.n_layer` (#24060)
4mo ago
smallthinker.cpp
hparams : refactor `hparams.n_layer` (#24060)
4mo ago
smollm3.cpp
hparams : refactor `hparams.n_layer` (#24060)
4mo ago
stablelm.cpp
hparams : refactor `hparams.n_layer` (#24060)
4mo ago
starcoder.cpp
hparams : refactor `hparams.n_layer` (#24060)
4mo ago
starcoder2.cpp
hparams : refactor `hparams.n_layer` (#24060)
4mo ago
step35.cpp
llama : load MTP tensors only if they are really used (#26296)
2mo ago
t5.cpp
hparams : refactor `hparams.n_layer` (#24060)
4mo ago
t5encoder.cpp
model: move `load_hparams` and `load_tensors` to per-model definition (#22004)
5mo ago
talkie.cpp
hparams : refactor `hparams.n_layer` (#24060)
4mo ago
wavtokenizer-dec.cpp
model : NvFP4 quantized LM head support (#23046)
5mo ago
xverse.cpp
hparams : refactor `hparams.n_layer` (#24060)
4mo ago