ggerganov/llama.cpp

Pull Requests Commits

cuda : re-add q4_0

ggerganov committed 1 year ago

Verified 62d7b6c8

cuda : build only necessary templates

ggerganov committed 1 year ago

Verified 503983a6

make : force CPU extensions [no ci]

ggerganov committed 1 year ago

Verified ae41fd2e

server : fix segfault on long system prompt (#8987)

compilade committed 1 year ago

Verified 98a532d4

cmake : remove unused option GGML_CURL (#9011)

ggerganov committed 1 year ago

Verified 43bdd3ce

ggml : move rope type enum to ggml.h (#8949)

danbev committed 1 year ago

Verified 06943a69

export-lora : throw error if lora is quantized (#9002)

ngxson committed 1 year ago

Verified 828d6ff7

ci : fix github workflow vulnerable to script injection (#9008)

diogoteles08 committed 1 year ago

Verified fc4ca27b

ci : enable RPC in all of the released builds (#9006)

rgerganov committed 1 year ago

Verified 1f67436c

llama : model-based max number of graph nodes calculation (#8970)

nicoboss committed 1 year ago

Verified 0fd93cde

docs: introduce gpustack and gguf-parser (#8873)

thxCode committed 1 year ago

Verified 84eb2f4f

grammar-parser : fix possible null-deref (#9004)

DavidKorczynski committed 1 year ago

Verified 1262e7ed

ggml: fix div-by-zero (#9003)

DavidKorczynski committed 1 year ago

Verified df5478fb

Fix a spelling mistake (#9001)

Septa2112 committed 1 year ago

Verified 2589292c

py : fix requirements check '==' -> '~=' (#8982)

ggerganov committed 1 year ago

Verified d3ae0ee8

server : handle models with missing EOS token (#8997)

ggerganov committed 1 year ago

Verified 5ef07e25

gguf-py : Numpy dequantization for most types (#8939)

compilade committed 1 year ago

Verified 4134999e

flake.lock: Update (#8979)

ggerganov committed 1 year ago

Verified 8cd1bcfd

update guide (#8909)

arthw committed 1 year ago

Verified a21c6fd4

llama : check all graph nodes when searching for result_embd_pooled (#8956)

fairydreaming committed 1 year ago

Verified 33309f66

Optimize Vulkan backend for better CPU performance and less GPU synchronization overhead. (#8943)

mtavenrath committed 1 year ago

Verified 7c5bfd57

metal : fix uninitialized abort_callback (#8968)

slaren committed 1 year ago

Verified 6e02327e

llama : default n_swa for phi-3 (#8931)

ngxson committed 1 year ago

Verified 7eb23840

Add support for encoder-only T5 models (#8900)

fairydreaming committed 1 year ago

Verified 7c3f55c1

gguf-py : fix double call to add_architecture() (#8952)

tarilabs committed 1 year ago

Verified 911b437f

Merge commit from fork

ggerganov committed 1 year ago

Verified b72942fa

llama : add support for lora adapters in T5 model (#8938)

fairydreaming committed 1 year ago

Verified 6afd1a99

make : fix llava obj file race (#8946)

ggerganov committed 1 year ago

Verified 272e3bd9

llama : better replace_all (cont) (#8926)

ggerganov committed 1 year ago

Verified 45a55b91

llava : support MiniCPM-V-2.5 (#7599)

tc-mb committed 1 year ago

Verified 3071c0a5

Older