ggml-org/llama.cpp

Pull Requests Commits

ggml : rms_norm in chunks

ggerganov committed 2 years ago

a1cdd29c

ggml : process mul mat rows in chunks

ggerganov committed 2 years ago

5a317898

llama : fix compile warnings in llama_set_state_data()

ggerganov committed 2 years ago

8a203f9f

ggml : fix scalar implementation of Q4_1 dot

ggerganov committed 2 years ago

4fd3e292

ggml : use F16 instead of F32 in Q4_0, Q4_1, Q8_0 (#1508)

ggerganov committed 2 years ago

Verified 2d5db483

tests : add missing header

ggerganov committed 2 years ago

Verified 6986c783

examples : add persistent chat (#1495)

ejones committed 2 years ago

Verified 943e6081

main : make reverse prompt option act as a stop token in non-interactive mode (#1032)

data-angel committed 2 years ago

Verified 7694b52b

readme : adds WizardLM to the list of supported models (#1485)

dakennedyd committed 2 years ago

Verified 79e3efb0

minor : fix compile warnings

ggerganov committed 2 years ago

Verified 4b7e245a

make kv_f16 the default for api users (#1517)

Green-Sky committed 2 years ago

Verified 5ea43392

Fixes #1511 lambda issue for w64devkit (mingw) (#1513)

DannyDaemonic committed 2 years ago

Verified ee965413

Remove unused n_parts parameter (#1509)

sw committed 2 years ago

Verified dc271c52

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

Older