Ewan Crawford
|
c61285e739
SYCL: Bump oneMath commit (#14152)
|
7 месяцев назад |
Christian Kastner
|
09cf2c7c65
cmake : Improve build-info.cpp generation (#14156)
|
7 месяцев назад |
Georgi Gerganov
|
c33fe8b8c4
vocab : prevent heap overflow when vocab is too small (#14145)
|
7 месяцев назад |
Anton Mitkov
|
ed52f3668e
sycl: Remove not needed copy f16->f32 for dnnl mul mat (#14125)
|
7 месяцев назад |
Georgi Gerganov
|
a681b4ba83
readme : remove project status link (#14149)
|
7 месяцев назад |
Georgi Gerganov
|
7d516443dd
server : re-enable SWA speculative decoding (#14131)
|
7 месяцев назад |
Georgi Gerganov
|
f6e1a7aa87
context : simplify output counting logic during decode (#14142)
|
7 месяцев назад |
Georgi Gerganov
|
c3ee46fab4
batch : remove logits_all flag (#14141)
|
7 месяцев назад |
Georgi Gerganov
|
e2c0b6e46a
cmake : handle whitepsaces in path during metal build (#14126)
|
7 месяцев назад |
Georgi Gerganov
|
9596506965
kv-cache : fix split_equal handling in unified implementation (#14130)
|
7 месяцев назад |
compilade
|
a20b2b05bc
context : round n_tokens to next multiple of n_seqs when reserving (#14140)
|
7 месяцев назад |
bandoti
|
2e89f76b7a
common: fix issue with regex_escape routine on windows (#14133)
|
7 месяцев назад |
Christian Kastner
|
532802f938
Implement GGML_CPU_ALL_VARIANTS for ARM (#14080)
|
7 месяцев назад |
Sigbjørn Skjæret
|
d4e0d95cf5
chore : clean up relative source dir paths (#14128)
|
7 месяцев назад |
Sigbjørn Skjæret
|
cc66a7f78f
tests : add test-tokenizers-repo (#14017)
|
7 месяцев назад |
Jeff Bolz
|
bd248d4dc7
vulkan: Better thread-safety for command pools/buffers (#14116)
|
7 месяцев назад |
Aman
|
7781e5fe99
webui: Wrap long numbers instead of infinite horizontal scroll (#14062)
|
7 месяцев назад |
Georgi Gerganov
|
89a184fa71
kv-cache : relax SWA masking condition (#14119)
|
7 месяцев назад |
Taylor
|
2baf07727f
server : pass default --keep argument (#14120)
|
7 месяцев назад |
Georgi Gerganov
|
7ae2932116
kv-cache : add LLAMA_KV_CACHE_DEBUG environment variable (#14121)
|
7 месяцев назад |
Jeff Bolz
|
1f7d50b293
vulkan: Track descriptor pools/sets per-context (#14109)
|
7 месяцев назад |
lhez
|
4c763c8d1b
opencl: add `mul_mv_id_q4_0_f32_8x_flat` (#14003)
|
7 месяцев назад |
compilade
|
dad5c44398
kv-cache : avoid modifying recurrent cells when setting inputs (#13834)
|
7 месяцев назад |
Sigbjørn Skjæret
|
55f6b9fa65
convert : fix duplicate key DeepSeek-R1 conversion error (#14103)
|
7 месяцев назад |
Sigbjørn Skjæret
|
3678b838bb
llama : support GEGLU for jina-bert-v2 (#14090)
|
7 месяцев назад |
Jeff Bolz
|
652b70e667
vulkan: force device 0 in CI (#14106)
|
7 месяцев назад |
Juk Armstrong
|
3a12db23b6
Fixed spec timings to: accepted/tested instead of accepted/drafted (#14104)
|
7 месяцев назад |
Georgi Gerganov
|
ae92c1855b
sync : ggml
|
7 месяцев назад |
Georgi Gerganov
|
b7ce1ad1e3
ggml : fix weak alias win32 (whisper/0)
|
7 месяцев назад |
0cc4m
|
97340b4c99
Vulkan: Don't default to CPU device (like llvmpipe), even if no other device is available, to allow fallback to CPU backend (#14099)
|
7 месяцев назад |