ggerganov/llama.cpp

Pull Requests Commits

ggml : use F16C conversion when available

ggerganov committed 2 years ago

Verified 40ec4882

benchmark-matmul: Print the average of the test results (#1490)

rankaiyx committed 2 years ago

Verified c238b587

convert.py: Support models which are stored in a single pytorch_model.bin (#1469)

TheBloke committed 2 years ago

Verified 2b264693

~7% faster Q5_1 AVX2 code (#1477)

ilyakurdyukov committed 2 years ago

Verified 42627421

define default model path once, sync path with readme (#1366)

ott2 committed 2 years ago

Verified 95606554

Add alternate include path for openblas (#1476)

sandyiscool committed 2 years ago

Verified 2a5ee023

fix get_num_physical_cores() (#1436)

zrm committed 2 years ago

Verified 63d20469

benchmark-matmul: fix clang-tidy issues, report results in GFLOPS (#1458)

slaren committed 2 years ago

Verified b5c9295e

cuda : deduplicated dequantization code (#1453)

JohannesGaessler committed 2 years ago

Verified eb363627

ggml : alternative fix for race condition bug in non-inplace ggml_compute_forward_diag_mask_f32 (#1454)

xaedes committed 2 years ago

Verified 79b2d5b6

ggml : various fixes (#1450)

ggerganov committed 2 years ago

Verified 13c351ad

ggml : add AVX support based on AVX2 code (#1430)

katsu560 committed 2 years ago

Verified 60f8c361

ggml : add GGML_QNT_VERSION to track quantization format changes

ggerganov committed 2 years ago

Verified 601a0334

cuda : fix convert function (#1412)

ggerganov committed 2 years ago

08737ef7

make : fix PERF build with cuBLAS

ggerganov committed 2 years ago

bda4d7c2

llama : fix unused warning

ggerganov committed 2 years ago

Verified 5a5aeb1e

ggml : multi-thread mul and diag_mask ops (#1428)

ggerganov committed 2 years ago

Verified 66841fdb

ggml : GPU-accelerated token generation (#1412)

JohannesGaessler committed 2 years ago

Verified 905d87b7

ggml : implement backward pass for llama + small training-llama-from-scratch example (#1360)

xaedes committed 2 years ago

Verified f954edda

ggml : sync alibi fix from ggml repo

ggerganov committed 2 years ago

Verified f048af02

Adding SSE instructions to ggml_vec_dot_q4_0_q8_0 (#1413)

3ooabkhxtn committed 2 years ago

Verified ac0cd259

llama : fix various warnings

ggerganov committed 2 years ago

Verified 0cd22e19

embedding : remove unused code (#1426)

SanftMonster committed 2 years ago

Verified 6456a4eb

readme : update Q4_0 perplexities

ggerganov committed 2 years ago

Verified cdd53508

llama : free ggml context in set / copy state data (close #1425)

ggerganov committed 2 years ago

Verified 738ace39

opencl : fix kernels for the new formats (#1422)

SlyEcho committed 2 years ago

Verified 699b1ad7

llama : fix --mtest option (close #1414)

ggerganov committed 2 years ago

Verified fb62f924

CLI args use - instead of _, backwards compatible (#1416)

JohannesGaessler committed 2 years ago

Verified 773ee249

Add clang-tidy reviews to CI (#1407)

slaren committed 2 years ago

Verified 553fd4d4

readme : add C#/.NET bindings repo (#1409)

SanftMonster committed 2 years ago

Verified 089b1c93

Older