ggml-org/llama.cpp

Pull Requests Commits

Add llama enum for IQ2_XS

Iwan Kawrakow committed 1 year ago

9bfcb16f

iq2_xs: had forgotten to delete iq2-data.h

Iwan Kawrakow committed 1 year ago

a1610b05

iq2_xs: faster AVX2 dit product

Iwan Kawrakow committed 1 year ago

8299b03a

iq2_xs: AVX2 dot product - 19.5 t/s

Iwan Kawrakow committed 1 year ago

3198e94f

iq2_xs: better ARM_NEON dot product

Iwan Kawrakow committed 1 year ago

52ea3f79

iq2_xs: working, but dog slow, ARM_NEON dot product

Iwan Kawrakow committed 1 year ago

ff49d876

iq2_xs: Metal now works

Iwan Kawrakow committed 1 year ago

55e2cae8

iq2_xs: WIP Metal

Iwan Kawrakow committed 1 year ago

0aacd551

iq2_xs: CUDA and scalar CPU works

Iwan Kawrakow committed 1 year ago

9b6e38d8

iq2_xs: this should have been in the basics

Iwan Kawrakow committed 1 year ago

9f21b82e

Iwan Kawrakow committed 1 year ago

3569fa3f

scripts : script to get Paul Graham essays in txt format (#4838)

ggerganov committed 1 year ago

Verified d9653894

server : update readme about token probs (#4777)

ibehnam committed 1 year ago

Verified 128de358

server : add api-key flag to documentation (#4832)

Zsapi committed 1 year ago

Verified 8c583303

ggml : fix vld1q_s8_x4 32-bit compat (#4828)

ggerganov committed 1 year ago

Verified 18c2e175

CUDA: faster softmax via shared memory + fp16 math (#4742)

JohannesGaessler committed 1 year ago

Verified 8f900abf

common : fix the short form of `--grp-attn-w`, not `-gat` (#4825)

howlger committed 1 year ago

Verified 1fc2f265

readme : add link to SOTA models

ggerganov committed 1 year ago

Verified a9a8c5de

SOTA 2-bit quants (#4773)

ikawrakow committed 1 year ago

Verified dd5ae064

swift : exclude ggml-metal.metal from the package (#4822)

ggerganov committed 1 year ago

Verified 668b31fc

llama.swiftui : update readme

ggerganov committed 1 year ago

Verified 42ea63c5

main : add self-extend support (#4815)

ggerganov committed 1 year ago

Verified 52531fdf

examples : add passkey test (#3856)

ggerganov committed 1 year ago

Verified b0034d93

readme : add lgrammel/modelfusion JS/TS client for llama.cpp (#4814)

lgrammel committed 1 year ago

Verified b7e79829

llama-bench : add no-kv-offload parameter (#4812)

slaren committed 1 year ago

Verified 226460cc

CUDA: fixed redundant value dequantization (#4809)

JohannesGaessler committed 1 year ago

Verified d5a410e8

llama : remove unused vars (#4796)

ggerganov committed 1 year ago

Verified 9dede37d

llama : remove redundant GQA check (#4796)

ggerganov committed 1 year ago

Verified 3c36213d

llama.swiftui : use llama.cpp as SPM package (#4804)

azarovalex committed 1 year ago

Verified 72d8407b

llama : print tensor meta for debugging

ggerganov committed 1 year ago

Verified d117d4dc

Older