ngxson/llama.cpp

Pull Requests Commits

ngxson committed 39 days ago

bbe95e69

ngxson committed 41 days ago

3efd047b

tests : move save-load-state from examples to tests (#23336)

ggerganov committed 41 days ago

Verified 40d5358d

server: expose prompt token counts in /slots endpoint (#23454)

ScrewTSW committed 41 days ago

Verified b65bb4ba

metal : optimize concat kernel and fix set kernel threads (#23411)

ggerganov committed 41 days ago

Verified a1a69f77

server : free draft/MTP resources on sleep to fix VRAM leak (#23461)

am17an committed 41 days ago

Verified 52fb93a2

server: re-inject subcommand when router spawns children under unified binary (#23442)

ServeurpersoCom committed 41 days ago

Verified c9021714

app : add batched-bench, fit-params, quantize & perplexity (#23459)

angt committed 41 days ago

Verified 1d7ab2b9

mtp: use inp_out_ids for skipping logit computation (#23433)

am17an committed 41 days ago

Verified 12e5d990

vocab : add Carbon-3B (HybridDNATokenizer) support (#23410)

kashif committed 41 days ago

Verified 7ea23ddf

doc: fix spec mtp typo (#23435)

ruixiang63 committed 41 days ago

Verified 2fc8d185

ui: Improve Git Hooks for UI development (#23403)

allozaur committed 41 days ago

Verified 5e932a1c

ggml : Check the right iface method before using the fallback 2d get (#23306)

TheBlueMatt committed 41 days ago

Verified 2754ce1b

llama-graph: fix null-buffer crash in llm_graph_input_attn_kv_iswa for SWA-only models (#23131)

ssfdre38 committed 41 days ago

Verified eeeaf618

hexagon: ssm-conv fix for large prompts (#23307)

tboinovski1 committed 41 days ago

Verified 0be84685

app : show version (#23426)

angt committed 41 days ago

Verified ce02093f

mtmd, model : merge HunyuanOCR into HunyuanVL and fix OCR vision precision (#23329)

wendadawen committed 41 days ago

Verified 6a257d44

ui: Add max image size option (#22849)

stduhpf committed 41 days ago

Verified 3a479c91

Move to backend sampling for MTP draft path (#23287)

gaugarg-nv committed 42 days ago

Verified ad277572

opencl: refactor backend initilization (#23318)

lhez committed 42 days ago

Verified 3a6db741

common/speculative : fix nullptr crash in get_devices_str (#23386)

ggerganov committed 42 days ago

Verified 510b5c2a

mtmd : DeepSeek-OCR image processing fixes, img_tool::resize padding refactor (#23345)

sfallah committed 42 days ago

Verified a8681a0e

vulkan: optimize operations in the IM2COL shader (#22685)

daniandtheweb committed 42 days ago

Verified acd604fb

feat: Add WAV MIME type variants and improve audio format detection (#23396)

allozaur committed 42 days ago

Verified 6ce96713

hexagon: HMX quantized matmul rework (#23368)

max-krasnyansky committed 42 days ago

Verified c9872a25

Programmatic Dependent Launch (PDL) for more performance on newer NVIDIA GPUs (Hopper+) (#22522)

aendk committed 42 days ago

Verified e9472282

app : introduce the llama unified executable (#23296)

angt committed 42 days ago

Verified 29f14822

refactor: Move text attachments up before the message content in chat completions payload (#23406)

allozaur committed 42 days ago

Verified e6b4acfe

mtmd: fit_params now take into account mmproj (#21489)

ngxson committed 42 days ago

Verified e2b129e1

docker : copy conversion files (#23370)

CISC committed 42 days ago

Verified 7e50ef7d

Older