Commits

History for

Commits on Jun 25, 2023

ggml : sync latest ggml (custom operators)
ggerganov
committedJun 25, 2023

Commits on Jun 24, 2023

ggml : improve ggml_graph_dump_dot, add ggml_format_name (ggerganov#1978 )
slaren
committedJun 24, 2023

Commits on Jun 19, 2023

ggml : sync latest ggml repo (ggerganov#1924 )
ggerganov
committedJun 19, 2023

Commits on Jun 18, 2023

metal : handle buffers larger than device's maxBufferLength (ggerganov#1826 )
ggerganov
committedJun 18, 2023

Commits on Jun 14, 2023

CUDA full GPU acceleration, KV cache in VRAM (ggerganov#1827 )
JohannesGaessler
committedJun 14, 2023

Commits on Jun 13, 2023

train : improved training-from-scratch example (ggerganov#1652 )

xaedes
and
ggerganov
committedJun 13, 2023

Commits on Jun 6, 2023

Multi GPU support, CUDA refactor, CUDA scratch buffer (ggerganov#1703 )
JohannesGaessler
committedJun 6, 2023

Commits on Jun 5, 2023

ggml : add SOTA 2,3,4,5,6 bit k-quantizations (ggerganov#1684 )

committedJun 5, 2023

Commits on Jun 4, 2023

llama : Metal inference (ggerganov#1642 )
ggerganov
committedJun 4, 2023

Commits on May 29, 2023

ggml : sync cgraph import / export API
ggerganov
committedMay 29, 2023

Commits on May 27, 2023

ggml : add ggml_tensor_overhead()
ggerganov
committedMay 27, 2023
ggml : sync ggml core (minor additions, e.g. ggml_get_tensor_by_name())
ggerganov
committedMay 27, 2023

Commits on May 22, 2023

OpenCL Token Generation Acceleration (ggerganov#1459 )

committedMay 22, 2023

Commits on May 20, 2023

ggml : add ggml_clamp() (ggerganov#1539 )
ggerganov
committedMay 20, 2023

Commits on May 19, 2023

ggml : use F16 instead of F32 in Q4_0, Q4_1, Q8_0 (ggerganov#1508 )
ggerganov
committedMay 19, 2023

Commits on May 14, 2023

ggml : various fixes (ggerganov#1450 )
ggerganov
committedMay 14, 2023
ggml : add GGML_QNT_VERSION to track quantization format changes
ggerganov
committedMay 14, 2023

Commits on May 13, 2023

ggml : GPU-accelerated token generation (ggerganov#1412 )

JohannesGaessler
and
ggerganov
committedMay 13, 2023
ggml : implement backward pass for llama + small training-llama-from-scratch example (ggerganov#1360 )

xaedes
and
ggerganov
committedMay 13, 2023

Commits on May 11, 2023

ggml : remove bit shuffling (ggerganov#1405 )

ggerganov
and
sw
committedMay 11, 2023

Commits on May 2, 2023

ggml: add names to tensors (ggerganov#1268 )
slaren
committedMay 2, 2023

Commits on May 1, 2023

cuBLAS: refactor and optimize f16 mat mul performance (ggerganov#1259 )
slaren
committedMay 1, 2023

Commits on Apr 30, 2023

ggml : add Q5 WASM SIMD + GGML_FTYPE
ggerganov
committedApr 30, 2023

Commits on Apr 29, 2023

ggml : fix visibility and unused warnings
ggerganov
committedApr 29, 2023

Commits on Apr 28, 2023

Remove Q4_3 which is no better than Q5 (ggerganov#1218 )
sw
committedApr 28, 2023
ggml : sync ggml (ggml_alibi)
ggerganov
committedApr 28, 2023
ggml : add CLBlast support (ggerganov#1164 )

committedApr 28, 2023

Commits on Apr 26, 2023

ggml : add Q5_0 and Q5_1 quantization (ggerganov#1187 )

ggerganov
and
sw
committedApr 26, 2023

Commits on Apr 25, 2023

ggml : add Q8_0 quantization format (rename the old one to Q8_1) (ARM NEON) (ggerganov#1179 )
ggerganov
committedApr 25, 2023

Commits on Apr 24, 2023

ggml : export symbols (ggerganov#1155 )
ggerganov
committedApr 24, 2023

Commits on Apr 20, 2023

ggml : sync ggml (add GPT-NeoX RoPE implementation)
ggerganov
committedApr 20, 2023
llama : multi-threaded quantization (ggerganov#1075 )

committedApr 20, 2023
ggml : add Q4_3 quantization (ggerganov#1082 )
ggerganov
committedApr 20, 2023

Commits on Apr 19, 2023

Add NVIDIA cuBLAS support (ggerganov#1044 )
slaren
committedApr 19, 2023

Commits on Apr 18, 2023

ggml : add new Q4_2 quantization (ARM only) (ggerganov#1046 )
ggerganov
committedApr 18, 2023