 leejet
|
bbac6a26b2
ggml: fix cuda kernel launch configuration for k_compute_batched_ptrs to support large batch (#16744)
|
3 months ago |
 Sigbjørn Skjæret
|
73a48c9790
convert : enable expert group selection for all models with it (#16691)
|
3 months ago |
|
Sigbjørn Skjæret
|
f696428ce8
graph : add clamping to ffn_moe_weights_sum to avoid div-by-zero (#16655)
|
3 months ago |
|
Sigbjørn Skjæret
|
7cce4f8158
model : set res->t_embd in SmallThinker models (#16782)
|
3 months ago |
 amirai21
|
8d8862829c
docs : add Jamba to Text-only models list (#16778)
|
3 months ago |
 Aman Gupta
|
f77c13b91f
CUDA: General GEMV fusion (#16715)
|
3 months ago |
 Gilad S.
|
3cfa9c3f12
vulkan: deduplicate Microsoft Direct3D12 devices (#16689)
|
3 months ago |
 Galunid
|
5d195f17bc
convert : handle mmproj filename/path properly (#16760)
|
3 months ago |
 Shunta Saito
|
226f295f4d
model : set res->t_embd in PLaMo2 models (#16766)
|
3 months ago |
 Giuseppe Scrivano
|
f90b4a8efe
vulkan: delete dead code (#16732)
|
3 months ago |
 Jeff Bolz
|
8423d01931
vulkan: Optimize SSM_SCAN (#16645)
|
3 months ago |
 compilade
|
5cca2542ac
convert : avoid dequantizing mxfp4 for GPT-OSS (#16756)
|
3 months ago |
|
leejet
|
55945d2ef5
ggml: fix CUDA grid launch condition for large block_nums.y in binbcast (#16742)
|
3 months ago |
|
Aman Gupta
|
0bcb40b48c
CUDA: use CUB for arbitary size argsort (#16754)
|
3 months ago |
 Florian Badie
|
69e9ff0103
webui: support q URL parameter (#16728)
|
3 months ago |
 Daniel Bevenius
|
5a91109a5d
model-conversion : add trust_remote_code for orig model run [no ci] (#16751)
|
3 months ago |
|
compilade
|
f8f071fadd
convert : handle pre-quantized models (#14810)
|
3 months ago |
 Johannes Gäßler
|
0bf47a1dbb
server: add memory breakdown print (#16740)
|
3 months ago |
 Julien Denize
|
dd62dcfab9
convert : Make mistral-common dependency optional (#16738)
|
3 months ago |
 Xuan-Son Nguyen
|
d0660f237a
mtmd-cli : allow using --jinja (#16718)
|
3 months ago |
 Prajwal B Mehendarkar
|
fe6a9882ac
Manually link -lbsd to resolve flock symbol on AIX (#16610)
|
3 months ago |
|
Aman Gupta
|
061f0eff02
ggml-cuda: use passed ops instead of hardcoded ops (#16712)
|
3 months ago |
 matteo
|
8cf6b42d46
server : send partial stop string when <EOG> is reached (#15007)
|
3 months ago |
 Matthew Michel
|
9de9672adb
sycl: use async memory allocation to fix crashes during graph recording (#16644)
|
3 months ago |
 Max Krasnyansky
|
63d2fc46e1
Add experimental ggml-hexagon backend for the Hexagon NPU (#16547)
|
3 months ago |
 Diego Devesa
|
a2e0088d92
Revert "ggml : Leverage the existing GGML_F32_VEC helpers to vectorize ggml_v…" (#16723)
|
3 months ago |
 Pascal
|
9b9201f65a
webui: introduce OpenAI-compatible model selector in JSON payload (#16562)
|
3 months ago |
 sirus20x6
|
19a5a3edfd
ggml : Leverage the existing GGML_F32_VEC helpers to vectorize ggml_vec_set_f32 for faster fills (#16522)
|
3 months ago |
 Acly
|
d8eaa26e4d
tests : fix test-thread-safety when compiling with multiple backends (#16699)
|
3 months ago |
|
Aman Gupta
|
9285325ce0
CUDA: fix bug in topk-moe softmax (#16711)
|
3 months ago |
