llama.cpp

afmoe.cpp

graph: skip the LM head for pooled embeddings (#30445 )

2026-10-13 16:16:14 +02:00

apertus.cpp

graph: skip the LM head for pooled embeddings (#30445 )

2026-10-13 16:16:14 +02:00

arcee.cpp

graph: skip the LM head for pooled embeddings (#30445 )

2026-10-13 16:16:14 +02:00

arctic.cpp

graph: skip the LM head for pooled embeddings (#30445 )

2026-10-13 16:16:14 +02:00

arwkv7.cpp

graph: skip the LM head for pooled embeddings (#30445 )

2026-10-13 16:16:14 +02:00

baichuan.cpp

graph: skip the LM head for pooled embeddings (#30445 )

2026-10-13 16:16:14 +02:00

bailingmoe2.cpp

graph: skip the LM head for pooled embeddings (#30445 )

2026-10-13 16:16:14 +02:00

bailingmoe3.cpp

graph: skip the LM head for pooled embeddings (#30445 )

2026-10-13 16:16:14 +02:00

bailingmoe.cpp

graph: skip the LM head for pooled embeddings (#30445 )

2026-10-13 16:16:14 +02:00

beam.cpp

model: add Beam (Reflection) (#30670 )

2026-10-17 00:07:55 +02:00

bert.cpp

models: drop the t_embd expand of the embedding graphs (#30461 )

2026-10-14 08:15:08 +02:00

bitnet.cpp

graph: skip the LM head for pooled embeddings (#30445 )

2026-10-13 16:16:14 +02:00

bloom.cpp

graph: skip the LM head for pooled embeddings (#30445 )

2026-10-13 16:16:14 +02:00

chameleon.cpp

graph: skip the LM head for pooled embeddings (#30445 )

2026-10-13 16:16:14 +02:00

chatglm.cpp

graph: skip the LM head for pooled embeddings (#30445 )

2026-10-13 16:16:14 +02:00

clef.cpp

clef : include <algorithm> for std::find and std::copy (#30981 )

2026-10-22 18:14:47 +02:00

clip.cpp

llama: Restore quantization of mmprojs (#26818 )

2026-08-10 11:58:32 +02:00

codeshell.cpp

graph: skip the LM head for pooled embeddings (#30445 )

2026-10-13 16:16:14 +02:00

cogvlm.cpp

graph: skip the LM head for pooled embeddings (#30445 )

2026-10-13 16:16:14 +02:00

cohere2.cpp

graph: skip the LM head for pooled embeddings (#30445 )

2026-10-13 16:16:14 +02:00

cohere2moe.cpp

graph: skip the LM head for pooled embeddings (#30445 )

2026-10-13 16:16:14 +02:00

command-r.cpp

graph: skip the LM head for pooled embeddings (#30445 )

2026-10-13 16:16:14 +02:00

dbrx.cpp

graph: skip the LM head for pooled embeddings (#30445 )

2026-10-13 16:16:14 +02:00

deci.cpp

graph: skip the LM head for pooled embeddings (#30445 )

2026-10-13 16:16:14 +02:00

deepseek2.cpp

graph: skip the LM head for pooled embeddings (#30445 )

2026-10-13 16:16:14 +02:00

deepseek2ocr.cpp

convert : add --fuse-qkv flag to fuse Q/K/V into QKV during HF-to-GGUF conversion (#22780 )

2026-09-07 00:47:05 +08:00

deepseek4.cpp

llama: create new fused op for QSA mask (#30640 )

2026-10-22 01:02:22 +08:00

deepseek32.cpp

graph: skip the LM head for pooled embeddings (#30445 )

2026-10-13 16:16:14 +02:00

deepseek.cpp

graph: skip the LM head for pooled embeddings (#30445 )

2026-10-13 16:16:14 +02:00

delta-net-base.cpp

models : remove redundant ggml_cont calls now supported by all unary kernels (#30198 )

2026-10-13 09:26:52 +02:00

dflash.cpp

dflash : load optional FC input scale (#30945 )

2026-10-22 08:24:12 +03:00

dots1.cpp

graph: skip the LM head for pooled embeddings (#30445 )

2026-10-13 16:16:14 +02:00

dots3note.cpp

graph: skip the LM head for pooled embeddings (#30445 )

2026-10-13 16:16:14 +02:00

dream.cpp

graph: skip the LM head for pooled embeddings (#30445 )

2026-10-13 16:16:14 +02:00

eagle3.cpp

graph: skip the LM head for pooled embeddings (#30445 )

2026-10-13 16:16:14 +02:00

ernie4-5-moe.cpp

graph: skip the LM head for pooled embeddings (#30445 )

2026-10-13 16:16:14 +02:00

ernie4-5.cpp

graph: skip the LM head for pooled embeddings (#30445 )

2026-10-13 16:16:14 +02:00

eurobert.cpp

models: drop the t_embd expand of the embedding graphs (#30461 )

2026-10-14 08:15:08 +02:00

exaone4.cpp

graph: skip the LM head for pooled embeddings (#30445 )

2026-10-13 16:16:14 +02:00

exaone-moe.cpp

graph: skip the LM head for pooled embeddings (#30445 )

2026-10-13 16:16:14 +02:00

exaone.cpp

graph: skip the LM head for pooled embeddings (#30445 )

2026-10-13 16:16:14 +02:00

falcon-h1.cpp

graph: skip the LM head for pooled embeddings (#30445 )

2026-10-13 16:16:14 +02:00

falcon.cpp

graph: skip the LM head for pooled embeddings (#30445 )

2026-10-13 16:16:14 +02:00

flint.cpp

models : add Flint model (FlintForCausalLM) (#30928 )

2026-10-21 23:26:22 +02:00

gemma2.cpp

graph: skip the LM head for pooled embeddings (#30445 )

2026-10-13 16:16:14 +02:00

gemma3.cpp

graph: skip the LM head for pooled embeddings (#30445 )

2026-10-13 16:16:14 +02:00

gemma3n.cpp

graph: skip the LM head for pooled embeddings (#30445 )

2026-10-13 16:16:14 +02:00

gemma4-assistant.cpp

graph: skip the LM head for pooled embeddings (#30445 )

2026-10-13 16:16:14 +02:00

gemma4.cpp

graph: skip the LM head for pooled embeddings (#30445 )

2026-10-13 16:16:14 +02:00

gemma-embedding2.cpp

models: drop the t_embd expand of the embedding graphs (#30461 )

2026-10-14 08:15:08 +02:00

gemma-embedding.cpp

models: drop the t_embd expand of the embedding graphs (#30461 )

2026-10-14 08:15:08 +02:00

gemma.cpp

graph: skip the LM head for pooled embeddings (#30445 )

2026-10-13 16:16:14 +02:00

glm4-moe.cpp

graph: skip the LM head for pooled embeddings (#30445 )

2026-10-13 16:16:14 +02:00

glm4.cpp

graph: skip the LM head for pooled embeddings (#30445 )

2026-10-13 16:16:14 +02:00

glm5-next.cpp

glm5 : optimize inputs (#30932 )

2026-10-22 16:59:29 +03:00

glm-dsa.cpp

graph: skip the LM head for pooled embeddings (#30445 )

2026-10-13 16:16:14 +02:00

gpt2.cpp

graph: skip the LM head for pooled embeddings (#30445 )

2026-10-13 16:16:14 +02:00

gptneox.cpp

graph: skip the LM head for pooled embeddings (#30445 )

2026-10-13 16:16:14 +02:00

granite-hybrid.cpp

models : Record layer inputs for all granite models (#30957 )

2026-10-22 10:02:41 +03:00

granite-moe.cpp

model: fix all granite family parameter counts (#28643 )

2026-09-10 09:49:56 +03:00

granite-swa.cpp

models : Record layer inputs for all granite models (#30957 )

2026-10-22 10:02:41 +03:00

granite-switch.cpp

models : Record layer inputs for all granite models (#30957 )

2026-10-22 10:02:41 +03:00

granite.cpp

models : Record layer inputs for all granite models (#30957 )

2026-10-22 10:02:41 +03:00

grok.cpp

graph: skip the LM head for pooled embeddings (#30445 )

2026-10-13 16:16:14 +02:00

grovemoe.cpp

graph: skip the LM head for pooled embeddings (#30445 )

2026-10-13 16:16:14 +02:00

hrm-text.cpp

graph: skip the LM head for pooled embeddings (#30445 )

2026-10-13 16:16:14 +02:00

hunyuan-dense.cpp

model: move load_hparams and load_tensors to per-model definition (#22004 )

2026-05-04 12:36:59 +02:00

hunyuan-moe.cpp

graph: skip the LM head for pooled embeddings (#30445 )

2026-10-13 16:16:14 +02:00

hunyuan-vl.cpp

graph: skip the LM head for pooled embeddings (#30445 )

2026-10-13 16:16:14 +02:00

hy-v3.cpp

graph: skip the LM head for pooled embeddings (#30445 )

2026-10-13 16:16:14 +02:00

hy-v4.cpp

graph: skip the LM head for pooled embeddings (#30445 )

2026-10-13 16:16:14 +02:00

internlm2.cpp

graph: skip the LM head for pooled embeddings (#30445 )

2026-10-13 16:16:14 +02:00

jais2.cpp

graph: skip the LM head for pooled embeddings (#30445 )

2026-10-13 16:16:14 +02:00

jais.cpp

graph: skip the LM head for pooled embeddings (#30445 )

2026-10-13 16:16:14 +02:00

jamba.cpp

graph: skip the LM head for pooled embeddings (#30445 )

2026-10-13 16:16:14 +02:00

jina-bert-v2.cpp

hparams : refactor hparams.n_layer (#24060 )

2026-06-05 11:09:36 +03:00

jina-bert-v3.cpp

models: clean up some dead switch branches in old models (#28669 )

2026-09-10 16:09:35 +03:00

k2-horizon.cpp

graph: skip the LM head for pooled embeddings (#30445 )

2026-10-13 16:16:14 +02:00

kimi-k3.cpp

graph: skip the LM head for pooled embeddings (#30445 )

2026-10-13 16:16:14 +02:00

kimi-linear.cpp

graph: skip the LM head for pooled embeddings (#30445 )

2026-10-13 16:16:14 +02:00

kolibri1.cpp

model : drop the redundant precision overrides in the Kolibri-1 router (#30865 )

2026-10-20 15:58:09 +03:00

laguna.cpp

model: load YaRN beta metadata for laguna models (#30331 )

2026-10-13 22:18:01 +08:00

lfm2.cpp

refactor : move LFM2 shortconv and FFN blocks into a shared LFM base (#30874 )

2026-10-20 23:17:13 +02:00

lfm2moe.cpp

model: add NVIDIA Nemotron-3-Puzzle-75B-A9B (NemotronHPuzzle) support (#25444 )

2026-09-03 08:53:08 +02:00

lfm-base.cpp

refactor : move LFM2 shortconv and FFN blocks into a shared LFM base (#30874 )

2026-10-20 23:17:13 +02:00

llada-moe.cpp

graph: skip the LM head for pooled embeddings (#30445 )

2026-10-13 16:16:14 +02:00

llada.cpp

graph: skip the LM head for pooled embeddings (#30445 )

2026-10-13 16:16:14 +02:00

llama4.cpp

graph: skip the LM head for pooled embeddings (#30445 )

2026-10-13 16:16:14 +02:00

llama-embed.cpp

model: move load_hparams and load_tensors to per-model definition (#22004 )

2026-05-04 12:36:59 +02:00

llama.cpp

graph: skip the LM head for pooled embeddings (#30445 )

2026-10-13 16:16:14 +02:00

maincoder.cpp

graph: skip the LM head for pooled embeddings (#30445 )

2026-10-13 16:16:14 +02:00

maion-coder.cpp

models : add maion-coder (#29778 )

2026-10-15 12:18:46 +03:00

mamba2.cpp

mamba2: remove hardcoded 2x expansion factor and invalid d_inner % d_state check (#23082 )

2026-06-26 08:50:54 +03:00

mamba-base.cpp

fix(mamba) : make time-step projection input contiguous (#28832 )

2026-09-20 07:58:14 +03:00

mamba.cpp

graph: skip the LM head for pooled embeddings (#30445 )

2026-10-13 16:16:14 +02:00

maple.cpp

graph: skip the LM head for pooled embeddings (#30445 )

2026-10-13 16:16:14 +02:00

mellum.cpp

model : MTP support for Mellum2.1 (#30612 )

2026-10-16 10:20:03 +02:00

mimo2.cpp

graph: skip the LM head for pooled embeddings (#30445 )

2026-10-13 16:16:14 +02:00

minicpm3.cpp

graph: skip the LM head for pooled embeddings (#30445 )

2026-10-13 16:16:14 +02:00

minicpm.cpp

hparams : refactor hparams.n_layer (#24060 )

2026-06-05 11:09:36 +03:00

minimax-01.cpp

graph: skip the LM head for pooled embeddings (#30445 )

2026-10-13 16:16:14 +02:00

minimax-m2.cpp

graph: skip the LM head for pooled embeddings (#30445 )

2026-10-13 16:16:14 +02:00

minimax-m3.cpp

graph: skip the LM head for pooled embeddings (#30445 )

2026-10-13 16:16:14 +02:00

mistral3.cpp

mtmd: add mtmd_audio_stream + support voxtral-realtime (#30629 )

2026-10-16 01:46:42 +02:00

mistral4.cpp

model: move load_hparams and load_tensors to per-model definition (#22004 )

2026-05-04 12:36:59 +02:00

models.h

models : add Flint model (FlintForCausalLM) (#30928 )

2026-10-21 23:26:22 +02:00

modern-bert.cpp

model : use exact GELU for ModernBERT encoders (#30108 )

2026-10-09 23:26:56 +08:00

mpt.cpp

graph: skip the LM head for pooled embeddings (#30445 )

2026-10-13 16:16:14 +02:00

muse-glimmer.cpp

graph: skip the LM head for pooled embeddings (#30445 )

2026-10-13 16:16:14 +02:00

muse-spark.cpp

model: add Muse Spark Open (text, vision and audio) (#30616 )

2026-10-15 23:27:12 +02:00

nanbeige.cpp

graph: skip the LM head for pooled embeddings (#30445 )

2026-10-13 16:16:14 +02:00

nemotron-h-moe.cpp

graph: skip the LM head for pooled embeddings (#30445 )

2026-10-13 16:16:14 +02:00

nemotron-h.cpp

models : add Nemotron Super 3.5 VL and Nemotron Bolt 74B A9B support (#30520 )

2026-10-15 16:34:56 +03:00

nemotron.cpp

nemotron : support non-standard GQA, BPE tokenizer and domyn-small pre-type (#23343 )

2026-10-22 13:34:47 +02:00

neo-bert.cpp

models: drop the t_embd expand of the embedding graphs (#30461 )

2026-10-14 08:15:08 +02:00

nomic-bert-moe.cpp

models: clean up some dead switch branches in old models (#28669 )

2026-09-10 16:09:35 +03:00

nomic-bert.cpp

models: clean up some dead switch branches in old models (#28669 )

2026-09-10 16:09:35 +03:00

olmo2.cpp

graph: skip the LM head for pooled embeddings (#30445 )

2026-10-13 16:16:14 +02:00

olmo.cpp

graph: skip the LM head for pooled embeddings (#30445 )

2026-10-13 16:16:14 +02:00

olmoe.cpp

graph: skip the LM head for pooled embeddings (#30445 )

2026-10-13 16:16:14 +02:00

openai-moe.cpp

graph: skip the LM head for pooled embeddings (#30445 )

2026-10-13 16:16:14 +02:00

openelm.cpp

graph: skip the LM head for pooled embeddings (#30445 )

2026-10-13 16:16:14 +02:00

orion.cpp

graph: skip the LM head for pooled embeddings (#30445 )

2026-10-13 16:16:14 +02:00

paddleocr.cpp

graph: skip the LM head for pooled embeddings (#30445 )

2026-10-13 16:16:14 +02:00

pangu-embed.cpp

graph: skip the LM head for pooled embeddings (#30445 )

2026-10-13 16:16:14 +02:00

phi2.cpp

graph: skip the LM head for pooled embeddings (#30445 )

2026-10-13 16:16:14 +02:00

phi3.cpp

graph: skip the LM head for pooled embeddings (#30445 )

2026-10-13 16:16:14 +02:00

phimoe.cpp

hparams : refactor hparams.n_layer (#24060 )

2026-06-05 11:09:36 +03:00

plamo2.cpp

graph: skip the LM head for pooled embeddings (#30445 )

2026-10-13 16:16:14 +02:00

plamo3.cpp

graph: skip the LM head for pooled embeddings (#30445 )

2026-10-13 16:16:14 +02:00

plamo.cpp

graph: skip the LM head for pooled embeddings (#30445 )

2026-10-13 16:16:14 +02:00

plm.cpp

graph: skip the LM head for pooled embeddings (#30445 )

2026-10-13 16:16:14 +02:00

pockettts.cpp

graph: skip the LM head for pooled embeddings (#30445 )

2026-10-13 16:16:14 +02:00

qwen2.cpp

graph: skip the LM head for pooled embeddings (#30445 )

2026-10-13 16:16:14 +02:00

qwen2moe.cpp

graph: skip the LM head for pooled embeddings (#30445 )

2026-10-13 16:16:14 +02:00

qwen2vl.cpp

graph: skip the LM head for pooled embeddings (#30445 )

2026-10-13 16:16:14 +02:00

qwen3.cpp

qwen3 : bidirectional attention and dense_2 output projection (#30607 )

2026-10-19 09:29:02 +03:00

qwen3moe.cpp

graph: skip the LM head for pooled embeddings (#30445 )

2026-10-13 16:16:14 +02:00

qwen3next.cpp

qwen: reshape output to 2d instead of 3d (#30823 )

2026-10-20 22:09:56 +03:00

qwen3tts.cpp

mtmd: support Qwen3-TTS (note: breaking change to llama-tts binary) (#26254 )

2026-08-04 17:26:15 +02:00

qwen3vl.cpp

graph: skip the LM head for pooled embeddings (#30445 )

2026-10-13 16:16:14 +02:00

qwen3vlmoe.cpp

graph: skip the LM head for pooled embeddings (#30445 )

2026-10-13 16:16:14 +02:00

qwen4exp.cpp

model : fix qwen4exp layer inputs for DFlash drafters (#30927 )

2026-10-22 16:49:38 +03:00

qwen35.cpp

qwen: reshape output to 2d instead of 3d (#30823 )

2026-10-20 22:09:56 +03:00

qwen35moe.cpp

qwen: reshape output to 2d instead of 3d (#30823 )

2026-10-20 22:09:56 +03:00

qwen.cpp

graph: skip the LM head for pooled embeddings (#30445 )

2026-10-13 16:16:14 +02:00

refact.cpp

graph: skip the LM head for pooled embeddings (#30445 )

2026-10-13 16:16:14 +02:00

rnd1.cpp

graph: skip the LM head for pooled embeddings (#30445 )

2026-10-13 16:16:14 +02:00

rwkv6-base.cpp

models : deduplicate delta-net graphs for Qwen family (#19597 )

2026-02-16 14:35:04 +02:00

rwkv6.cpp

graph: skip the LM head for pooled embeddings (#30445 )

2026-10-13 16:16:14 +02:00

rwkv6qwen2.cpp

graph: skip the LM head for pooled embeddings (#30445 )

2026-10-13 16:16:14 +02:00

rwkv7-base.cpp

models : deduplicate delta-net graphs for Qwen family (#19597 )

2026-02-16 14:35:04 +02:00

rwkv7.cpp

graph: skip the LM head for pooled embeddings (#30445 )

2026-10-13 16:16:14 +02:00

seed-oss.cpp

graph: skip the LM head for pooled embeddings (#30445 )

2026-10-13 16:16:14 +02:00

sensenova-u1.cpp

model: Add support for SenseNova U1 model (#28919 )

2026-10-21 21:54:16 +02:00

smallthinker.cpp

graph: skip the LM head for pooled embeddings (#30445 )

2026-10-13 16:16:14 +02:00

smollm3.cpp

graph: skip the LM head for pooled embeddings (#30445 )

2026-10-13 16:16:14 +02:00

spark2-5.cpp

graph: skip the LM head for pooled embeddings (#30445 )

2026-10-13 16:16:14 +02:00

stablelm.cpp

graph: skip the LM head for pooled embeddings (#30445 )

2026-10-13 16:16:14 +02:00

starcoder2.cpp

graph: skip the LM head for pooled embeddings (#30445 )

2026-10-13 16:16:14 +02:00

starcoder.cpp

graph: skip the LM head for pooled embeddings (#30445 )

2026-10-13 16:16:14 +02:00

step35.cpp

graph: skip the LM head for pooled embeddings (#30445 )

2026-10-13 16:16:14 +02:00

t5.cpp

graph: skip the LM head for pooled embeddings (#30445 )

2026-10-13 16:16:14 +02:00

t5encoder.cpp

model: move load_hparams and load_tensors to per-model definition (#22004 )

2026-05-04 12:36:59 +02:00

talkie.cpp

graph: skip the LM head for pooled embeddings (#30445 )

2026-10-13 16:16:14 +02:00

vllm-decision2.cpp

model: support vllm decision 2 model (#30378 )

2026-10-12 13:39:55 +02:00

wavtokenizer-dec.cpp

models: drop the t_embd expand of the embedding graphs (#30461 )

2026-10-14 08:15:08 +02:00

xverse.cpp

graph: skip the LM head for pooled embeddings (#30445 )

2026-10-13 16:16:14 +02:00