ggml-org/llama.cpp

Pull Requests Commits
context : explicit llama_context_i abstract interface
ggerganov committed 296 days ago
Verified 9cd78f11
gguf_convert_endian.py: implement byteswapping for q4_k and q6_k (#11349)
AlekseiNikiforovIBM committed 296 days ago
Verified 651adf4b
SYCL: Fix GGML_SYCL_DEBUG macro (#11995)
qnixsynapse committed 296 days ago
Verified 8303e8b0
context : reuse built_attn_mha
ggerganov committed 296 days ago
Verified 4a1054b5
context : fix recurrent reserve
ggerganov committed 296 days ago
Verified a5a85a3b
context : remove redundant virtual, protected -> private
ggerganov committed 297 days ago
Verified 0699a44c
graph : remove the build_kv_... API from llama_graph_i
ggerganov committed 297 days ago
Verified 6378112c
run: allow to customize prompt by env var LLAMA_PROMPT_PREFIX (#12041)
benoitf committed 297 days ago
Verified 7ad0779f
Some llama-run cleanups (#11973)
ericcurtin committed 297 days ago
Verified f777a73e
cont : enc should work now, next is dec
ggerganov committed 297 days ago
Verified 372fa3a8
ggml-cpu: Support s390x SIMD Instruction Set (#12019)
taronaeo committed 298 days ago
Verified af7747c9
CUDA: app option to compile without FlashAttention (#12025)
JohannesGaessler committed 298 days ago
Verified a28e0d5e
llava: build clip image from pixels (#11999)
tinglou committed 298 days ago
Verified 36c258ee
ci : fix arm upload artifacts (#12024)
ggerganov committed 298 days ago
Verified f3e64859
CUDA: optimize FA for GQA + large batches (#12014)
JohannesGaessler committed 298 days ago
Verified 5fa07c2f
ci : Build on Github-hosted arm64 runners (#12009)
Rohanjames1997 committed 298 days ago
Verified 335eb04a
server : disable Nagle's algorithm (#12020)
ggerganov committed 298 days ago
Verified cf756d6e
cuda: Add Q5_1, Q5_0, Q4_1 and Q4_0 to F32 conversion support. (#12000)
gcp committed 298 days ago
Verified d7090842
llama.swiftui : add "Done" dismiss button to help view (#11998)
danbev committed 298 days ago
Verified de8b5a36
wip enc-dec
ggerganov committed 299 days ago
Verified f5e80208
Merge branch 'master' into gg/llama-kv-cache
ggerganov committed 299 days ago
Verified c4c0a4d1
llama : skip loading unused tensors (#12004)
ggerganov committed 299 days ago
Verified 51f311e0
context : fix n_outputs init
ggerganov committed 299 days ago
Verified 3753b30d
context : wrap input tensors in struct
ggerganov committed 299 days ago
Verified f588a70d
context : add logs
ggerganov committed 299 days ago
Verified ebf1bdf9
doc: update contributing guidelines [no ci] (#11969)
JohannesGaessler committed 299 days ago
Verified 586d5fe6
graph : remove worst_case from the API
ggerganov committed 299 days ago
Verified 548c230d
CUDA: correct the lowest Maxwell supported by CUDA 12 (#11984)
PureJourney committed 299 days ago
Verified ecc8e3ae
context : add save/load for recurrent context
ggerganov committed 299 days ago
Verified 2645a7d9
MUSA: support ARM64 and enable dp4a .etc (#11843)
BodhiHu committed 299 days ago
Verified 0b3863ff
Terms Of Use Privacy Policy Refund Policy Impressum