ggml-org/llama.cpp

Pull Requests Commits
ggml : add CLBlast q5_0, q5_1, q8_0 dequant kernels (#1225)
0cc4m committed 2 years ago
Verified 76a88492
ggml : add Q5 WASM SIMD + GGML_FTYPE
ggerganov committed 2 years ago
Verified 6bc4400e
Various fixes to mat_mul benchmark (#1253)
sw committed 2 years ago
Verified f0d70f14
ggml : fix labels for GGML_OP_ALIBI
ggerganov committed 2 years ago
Verified 3e5aa8a1
ggml : fix 32-bit ARM NEON
ggerganov committed 2 years ago
Verified c3ca7a5f
ggml : use vzip instead of vuzp for consistency
ggerganov committed 2 years ago
Verified e8c05161
ggml : fix visibility and unused warnings
ggerganov committed 2 years ago
Verified 0b5a9350
ggml : fix #if for f32_f32 mul_mat (CLBlast) (#1229)
ggerganov committed 2 years ago
Verified ec728e44
ggml : adjust mul_mat_f16 work memory (#1226)
ggerganov committed 2 years ago
Verified 214b6a35
build : fix reference to old llama_util.h
ggerganov committed 2 years ago
Verified 305eb5af
examples : fix save-load-state + rename llama-util.h
ggerganov committed 2 years ago
Verified 84ca9c2e
common : change default parameters to pre-#1126 (#1223)
ggerganov committed 2 years ago
Verified 334637e4
llama : new sampling algorithms (#1126)
ivanstepanovftw committed 2 years ago
Verified dd7eff57
cuBLAS: use host pinned memory and dequantize while copying (#1207)
slaren committed 2 years ago
Verified 7fc50c05
cuBLAS: non-contiguous tensor support (#1215)
SlyEcho committed 2 years ago
Verified b1ee8f59
Remove Q4_3 which is no better than Q5 (#1218)
sw committed 2 years ago
Verified 36d19a60
readme : update hot topics
ggerganov committed 2 years ago
Verified 7f15c5c4
ggml : sync ggml (ggml_alibi)
ggerganov committed 2 years ago
Verified 55390bca
examples : add Jeopardy example (#1168)
CRD716 committed 2 years ago
Verified 5fba3c01
llama : add session file format and saved sessions in main (#1169)
ejones committed 2 years ago
Verified 1481a9cf
ggml : add helper debug printf in soft_max
ggerganov committed 2 years ago
Verified 11d90236
ggml : add CLBlast support (#1164)
0cc4m committed 2 years ago
Verified 7296c961
Correcting link to w64devkit (#1214)
Folko-Ven committed 2 years ago
Verified 78ec5437
Add Manjaro CUDA include and lib dirs to Makefile (#1212)
JohannesGaessler committed 2 years ago
Verified 92a6e13a
add avx2 for dot_q8_0_q8_0, 2x faster than scalar (#1211)
YannFollet committed 2 years ago
Verified 04aaae1d
ggml : slightly faster AVX2 implementation for Q5 (#1197)
sw committed 2 years ago
Verified 0b2da205
readme : add quantization info
ggerganov committed 2 years ago
Verified f9be42ad
ggml : add Q5_0 and Q5_1 quantization (#1187)
ggerganov committed 2 years ago
Verified 574406dc
Allow setting the rng seed after initialization. (#1184)
asgeir committed 2 years ago
Verified 87a6f846
Updating build instructions to include BLAS support (#1183)
daniandtheweb committed 2 years ago
Verified ea3ad7eb
Terms Of Use Privacy Policy Refund Policy Impressum