ngxson/llama.cpp

Pull Requests Commits

correct a small bug

ngxson committed 1 year ago

1609be01

rename build fn

ngxson committed 1 year ago

c282e408

ngxson committed 1 year ago

1fe63313

refactor resize_position_embeddings

ngxson committed 1 year ago

1326bf80

fix calc number of output tokens

ngxson committed 1 year ago

526c5aa0

ngxson committed 1 year ago

9bba547e

ngxson committed 1 year ago

afd43a91

support kimi vision model

ngxson committed 1 year ago

56cf4cad

ngxson committed 1 year ago

a7dcaa94

Merge branch 'master' into xsn/kimi-vl

ngxson committed 1 year ago

131cd43b

musa: handle __hgt2_mask, available starting from MUSA SDK rc4.3.0 (#15413)

yeahdongcn committed 1 year ago

Verified 67f09a3a

ggml-cpu: add mxfp4 VSX intrinsics for Power9+ (ppc64le) hardware (#15385)

mgiessing committed 1 year ago

Verified 6424594c

chat : clarify the meaning of reasoning_format (#15408)

ngxson committed 1 year ago

Verified e9288e88

server : remove swa_full warning (#15399)

ggerganov committed 1 year ago

Verified 9d262f4b

batched-bench : use rand tokens (#15398)

ggerganov committed 1 year ago

Verified f0d3c740

mtmd : clean up clip_n_output_tokens (#15391)

ngxson committed 1 year ago

Verified f08c4c0d

codeowners : remove mmv.*

ggerganov committed 1 year ago

6d7f1117

ggerganov committed 1 year ago

60212f1e

scripts : update sync scripts

ggerganov committed 1 year ago

f0c541d3

llama : merge conts and reshapes and remove unnecessary cont (#15380)

CISC committed 1 year ago

Verified baa9255a

readme : update hot topics (#15397)

ggerganov committed 1 year ago

Verified 3007baf2

server : fix incoming tasks not process in order (#15395)

davidef committed 1 year ago

Verified d1d82416

Fix broken build: require updated pip to support --break-system-packages (#15357)

danchev committed 1 year ago

Verified 618575c5

ggml-quants : fix make_qp_quants NANs and IQ1 assertion errors (#15379)

compilade committed 1 year ago

Verified f44f7931

vulkan: disable spirv-opt for bfloat16 shaders (#15352)

jeffbolznv committed 1 year ago

Verified ae532eac

server : export max observed n_past value (#15361)

okuvshynov committed 1 year ago

Verified e5155e69

vulkan: Use larger workgroups for mul_mat_vec when M is small (#15355)

jeffbolznv committed 1 year ago

Verified 21c17b5b

vulkan: support sqrt (#15370)

ddwkim committed 1 year ago

Verified 19f4deca

convert : force patch_embd weights to F16 or F32 to avoid broken GGUFs (#15367)

CISC committed 1 year ago

Verified 4d196981

ci : fix hang in windows-hip build/release (#15365)

CISC committed 1 year ago

Verified b143fbc8

Older