matteo
|
caf5681fcb
server : support jinja extra template kwargs (Qwen3 enable_thinking feature), from command line and from client (#13196)
|
6 mesiacov pred |
Renat
|
83790b0e7e
server : fix appearance of the chats list context menu for Safari (#14322)
|
6 mesiacov pred |
Akarshan Biswas
|
f47c1d7106
SYCL: disable faulty fp16 exp kernel (#14395)
|
6 mesiacov pred |
Sigbjørn Skjæret
|
a5d1fb6212
ggml : fix unmerged GGML_FPxx_TO_FPxx refactoring (#14443)
|
6 mesiacov pred |
Sigbjørn Skjæret
|
a0535ffa0d
ggml : implement REGLU/GEGLU/SWIGLU ops (#14158)
|
6 mesiacov pred |
Jeff Bolz
|
bd9c981d72
vulkan: Add fusion support for RMS_NORM+MUL (#14366)
|
6 mesiacov pred |
Aman Gupta
|
27208bf657
CUDA: add bf16 and f32 support to cublas_mul_mat_batched (#14361)
|
6 mesiacov pred |
Jeff Bolz
|
63a7bb3c7e
vulkan: handle noncontig in the final case of ggml_vk_get_cpy_pipeline (#14378)
|
6 mesiacov pred |
Jeff Bolz
|
00d5282c7f
vulkan: lock accesses of pinned_memory vector (#14333)
|
6 mesiacov pred |
Weizhao Ouyang
|
566c16fcce
model : add support for ERNIE 4.5 0.3B model (#14408)
|
6 mesiacov pred |
Xinpeng Dou
|
b25e92774e
fix async_mode bug (#14432)
|
6 mesiacov pred |
Sigbjørn Skjæret
|
6609507a91
ci : fix windows build and release (#14431)
|
6 mesiacov pred |
Jeff Bolz
|
ceb1bf5a34
vulkan: Fix GGML_VULKAN_SHADER_DEBUG_INFO (#14427)
|
6 mesiacov pred |
Georgi Gerganov
|
72babea5de
graph : make llm_graph_context destructor virtual (#14410)
|
6 mesiacov pred |
Georgi Gerganov
|
43678060c1
recurrent : call balloc split_reset() in init_batch() (#14414)
|
6 mesiacov pred |
Radoslav Gerganov
|
8d94219a4a
ggml : add ggml_set_rows (#14274)
|
6 mesiacov pred |
Sigbjørn Skjæret
|
f667f1e624
convert : fix broken sentencepiece vocab (#14416)
|
6 mesiacov pred |
Xuan-Son Nguyen
|
8846aace49
model : gemma3n text-only (#14400)
|
6 mesiacov pred |
bandoti
|
a01047b041
cmake: regen vulkan shaders when shaders-gen sources change (#14398)
|
6 mesiacov pred |
Sigbjørn Skjæret
|
b25346221d
llama : return mistral-v7-tekken as default template only (#14390)
|
6 mesiacov pred |
Georgi Gerganov
|
e8215dbb96
metal : add special-case mat-vec mul for ne00 == 4 (#14385)
|
6 mesiacov pred |
Georgi Gerganov
|
5783ae4359
metal : batch rows copy in a single threadgroup (#14384)
|
6 mesiacov pred |
Aaron Teo
|
bf5bcd0b85
docs: update s390x documentation + add faq (#14389)
|
6 mesiacov pred |
R0CKSTAR
|
716301d1b0
musa: enable fp16 mma (all) and cublas on qy2 (#13842)
|
6 mesiacov pred |
Aaron Teo
|
60ef23d6c1
ggml-cpu: enable IBM NNPA Vector Intrinsics (#14317)
|
6 mesiacov pred |
Sigbjørn Skjæret
|
b193d53069
ggml : do not output unprintable characters on GGUF load failure (#14381)
|
6 mesiacov pred |
Anton Mitkov
|
2bf9d539dd
sycl: GGML_SYCL_DISABLE_OPT on by default for all Intel Devices (#13973)
|
6 mesiacov pred |
lhez
|
73e53dc834
opencl: ref count `ggml_backend_opencl_context` and refactor profiling (#14254)
|
6 mesiacov pred |
Georgi Gerganov
|
62af464227
batch : fix check for empty sequences in memory (#14364)
|
6 mesiacov pred |
Mathieu Baudier
|
c148cf1946
cmake : use LLAMA_BUILD_NUMBER when defining LLAMA_INSTALL_VERSION (#14362)
|
6 mesiacov pred |