ggerganov/llama.cpp

Pull Requests Commits
metal: fix regression when no metal devices are present (#15531)
booxter committed 159 days ago
Verified 111f8d06
CUDA: MoE helper in device code, better tile sizes (#15525)
JohannesGaessler committed 159 days ago
Verified 5eff6ec9
model-conversion : set pooling type to none in logits.cpp (#15564)
danbev committed 159 days ago
Verified dfd9b5f6
model-conversion : add model card template for embeddings [no ci] (#15557)
danbev committed 159 days ago
Verified 5a6bc6b1
batched-bench : fix unified KV cache handling + pp timing (#15562)
ggerganov committed 159 days ago
Verified 6b64f74b
convert : update Ernie 4.5 dense architecture name (#15555)
ownia committed 159 days ago
Verified 0d5a4702
metal : add FA kernels for HS=40 (#15559)
ggerganov committed 159 days ago
Verified b0ba31f5
convert : support interns1-mini (#15412)
RunningLeon committed 159 days ago
Verified 7da9fed0
CANN: ROPE cache sin/cos repeat (#15501)
noemotiovon committed 160 days ago
Verified c247d06f
vulkan: apply MUL_MAT_ID subgroup optimization to non-coopmat devices (#15524)
0cc4m committed 160 days ago
Verified 043fb27d
kv-cache : support layer reuse (#15504)
ggerganov committed 160 days ago
Verified b730706a
vulkan: Support FA with any multiple of 8 head sizes (#15537)
jeffbolznv committed 160 days ago
Verified c9a24fb9
vulkan: enable Conv2D for Apple after MoltenVK fixed the bug (#15526)
0cc4m committed 160 days ago
Verified a9c6ffcb
vulkan: workaround MoltenVK compile failure in multi_add (#15506)
jeffbolznv committed 160 days ago
Verified e78cf0d4
CUDA: fix half2 -> half conversion for HIP (#15529)
JohannesGaessler committed 161 days ago
Verified 710dfc46
vulkan: optimize rms_norm, and allow the work to spread across multiple SMs (#15281)
jeffbolznv committed 161 days ago
Verified 611f419c
model : add support for Seed-OSS (#15490)
pwilkin committed 161 days ago
Verified b1afcab8
scripts: fix compare-llama-bench.py (#15521)
JohannesGaessler committed 161 days ago
Verified 9ef53690
chat : fix debug build assertion in trim function (#15520)
LaffeyNyaa committed 161 days ago
Verified 21dc4dda
vulkan: Rewrite synchronization to allow some overlap between nodes (#15489)
jeffbolznv committed 161 days ago
Verified 289bf411
vulkan.Dockerfile: install vulkan SDK using tarball (#15282)
yeahdongcn committed 161 days ago
Verified b55f06e1
vulkan : support ggml_mean (#15393)
Acly committed 161 days ago
Verified 0a9b43e5
vulkan: optimize mul_mat_id loading row ids into shared memory (#15427)
jeffbolznv committed 161 days ago
Verified 330c3d2d
test-opt: allow slight inprecision (#15503)
JohannesGaessler committed 162 days ago
Verified e92734d5
ggml WebGPU: add support for quantization types (#15440)
reeselevine committed 162 days ago
Verified 45363632
model : gpt-oss add response_format support (#15494)
aldehir committed 162 days ago
Verified 32732f24
ggml: add `conv3d` op (#15182)
rmatif committed 162 days ago
Verified 92f7f0a5
cuda : add Pad Reflect 1D support (#14659)
YavorGIvanov committed 162 days ago
Verified b1ab9182
llama : remove KV cache defragmentation logic (#15473)
ggerganov committed 162 days ago
Verified 9ebebef6
ggml-cpu: Support Q5_0 and Q5_1 on s390x (#15486)
taronaeo committed 162 days ago
Verified ad5c975c
Terms Of Use Privacy Policy Refund Policy Impressum