ggerganov/llama.cpp

Pull Requests Commits

CUDA: determine FA parallel blocks at runtime

JohannesGaessler committed 1 year ago

f6711cef

readme : update bindings (#12229)

lmbelo committed 1 year ago

Verified 3d652bfd

CUDA: fix FA logic for PTX 7.0 and CC >= 7.5 (#12222)

JohannesGaessler committed 1 year ago

Verified 5220a16d

HIP: rocWMMA documentation and enabling in workflow builds (#12179)

hjc4869 committed 1 year ago

Verified 3ffbbd5c

update function-calling.md w/ template override for functionary-small-v3.2 (#12214)

ochafik committed 1 year ago

Verified 42994048

llava: add big-endian conversion for image encoder (#12218)

taronaeo committed 1 year ago

Verified e9b2f84f

HIP/CUDA: set the paramerter value in maintain_cuda_graph instead of replaceing it. (#12209)

IMbackK committed 1 year ago

Verified e721c05c

android : fix KV cache log message condition (#12212)

hanyin-arm committed 1 year ago

Verified 57b6abf8

opencl : fix buffer alignment (#12197)

linehill committed 1 year ago

Verified 94bb63e4

opencl : fix `ulong` kernel args were set from `int` variables (#12174)

linehill committed 1 year ago

Verified f7924399

opencl : fix profile-related errors (#12095)

simon886212 committed 1 year ago

Verified ed4ce0dd

ggml-cpu: Faster IQ1 mul_mat_vec on AVX2 using BMI2 instructions (#12154)

remyoudompheng committed 1 year ago

Verified 07d15723

SYCL: Disable f16 Unary OPs as not supported by the kernels (#12201)

qnixsynapse committed 1 year ago

Verified 5e43f104

ggml : fix GGMLMetalClass ODR (#12200)

pminev committed 1 year ago

Verified 16e4b22c

ci : add fetch-depth to xcframework upload (#12195)

danbev committed 1 year ago

Verified 074c4fd3

`tool-call`: fix Qwen 2.5 Coder support, add micro benchmarks, support trigger patterns for lazy grammars (#12034)

ochafik committed 1 year ago

Verified 669912d9

ci : fix xcframework artifact tag (#12191)

danbev committed 1 year ago

Verified fa31c438

ci : remove xframework upload (#12190)

danbev committed 1 year ago

Verified 3ccbfe5a

server : fix cache reuse logic (#12161)

Clauszy committed 1 year ago

Verified 06a92a19

llama : add xcframework build script (#11996)

danbev committed 1 year ago

Verified a057897a

ggml : portability fixes for VS 2017 (#12150)

mgroeber9110 committed 1 year ago

Verified 5bbe6a9f

readme : fix roadmap link (#12185)

ggerganov committed 1 year ago

Verified 20a9b8f5

main: allow preloading conversation with -p and add -st / --single-turn (#12145)

CISC committed 1 year ago

Verified 56d7a9f8

`server`: fix deadly typo in response_format.json_schema.schema handling (#12168)

ochafik committed 1 year ago

Verified 1a24c462

HIP: implement FlashAttention via rocWMMA for CDNA and RDNA3+ (#12032)

hjc4869 committed 1 year ago

Verified becade5d

ggerganov committed 1 year ago

dfd6b2c0

cuda: unary ops as float + de-duplicate (ggml/1130)

cmdr2 committed 1 year ago

b64d7cc2

ggerganov committed 1 year ago

3d1cf3cf

cuda/vulkan: specify fp32-only support for some operations in supports_op (ggml/1129)

cmdr2 committed 1 year ago

0cbee131

ggerganov committed 1 year ago

8371d445

Older