ggerganov/llama.cpp

Pull Requests Commits

make : clean-up [no ci]

ggerganov committed 1 year ago

Verified a2385da5

metal : fix BF16 requirement for FA kernels

ggerganov committed 1 year ago

Verified b89e71b1

cuda : disable BF16 FA

ggerganov committed 1 year ago

Verified bc143ecf

metal : prevent int overflows [no ci]

ggerganov committed 1 year ago

Verified 5d1a10d2

build : remove obsolete compile flag [no ci]

ggerganov committed 1 year ago

Verified 486a5eb8

metal : compile-guard bf16 FA kernels

ggerganov committed 1 year ago

Verified 120d5128

metal : minor clean-up

ggerganov committed 1 year ago

Verified 2fccc8ac

metal : use F16 precision in FA kernels

ggerganov committed 1 year ago

Verified 7facc29d

ggml : add ggml_flash_attn_ext_get_prec

ggerganov committed 1 year ago

Verified 25e87730

server : minor UI fix (#10207)

ngxson committed 1 year ago

Verified 76c6e7f1

server : revamp chat UI with vuejs and daisyui (#10175)

ngxson committed 1 year ago

Verified a71d81cf

scripts : add amx to sync-ggml.sh [no ci]

ggerganov committed 1 year ago

Verified eec4d717

ggerganov committed 1 year ago

Verified 3b088286

scripts : sync update

ggerganov committed 1 year ago

Verified a2c6fd74

ggml : add ggml-cpu.h to the public headers (#10204)

slaren committed 1 year ago

Verified 97404c4a

Remove identical wte/etw logic for jais (#10203)

fmz committed 1 year ago

Verified 60e17ce2

DRY: Fixes clone functionality (#10192)

wwoodsTM committed 1 year ago

Verified 5107e8ce

fix q4_0_8_8 format for corrupted tokens issue (#10198)

snadampal committed 1 year ago

Verified 2319126a

Optimize RWKV6 Operator Naming and Implement Multi-core CPU/ SYCL Acceleration (#10133)

Zhiyuan Li committed 1 year ago

Verified 3bcd40b3

metal : add BF16 support (#8439)

ggerganov committed 1 year ago

Verified 5c333e01

server : remove hack for extra parallel slot (#10187)

ggerganov committed 1 year ago

Verified b11f9ba9

metal : fix from ptr buffer name (#10189)

slaren committed 1 year ago

Verified 94d8cb8b

ggml : adjust is_first_call init value (#10193)

ggerganov committed 1 year ago

Verified 1dc04b2d

metal : add quantized FA support (#10149)

ggerganov committed 1 year ago

Verified a1eaf6a9

llama : add <|tool_call|> formatting to Granite template (#10177)

gabe-l-hart committed 1 year ago

Verified b8deef0e

ggml : fix arch check in bf16_to_fp32 (#10164)

slaren committed 1 year ago

Verified a9e8a9a0

Q6_K AVX improvements (#10118)

netrunnereve committed 1 year ago

Verified 34073647

ggml : fix gelu tables initialization (#10172)

slaren committed 1 year ago

Verified d5a409e5

ggml : fix q4xx mat mul, increase ggml_aligned_malloc alignment (#10167)

slaren committed 1 year ago

Verified 401558b7

server : clarify /slots endpoint, add is_processing (#10162)

ngxson committed 1 year ago

Verified 9e0ecfb6

Older