ggml-org/llama.cpp

Pull Requests Commits
batch : add optional for sequential equal split (#14511)
ggerganov committed 1 year ago
Verified 67d1ef23
graph : prepare for 4D mask (#14515)
ggerganov committed 1 year ago
Verified 7b50f7c0
batch : add n_used count (#14512)
ggerganov committed 1 year ago
Verified c79184d2
CANN: Replace aclrtMemsetSync with aclnnInplaceZero operator (#14002)
luyhcsu committed 1 year ago
Verified 499a8f5a
ggml : implement GEGLU_ERF and GEGLU_QUICK ops (#14445)
CISC committed 1 year ago
Verified 28657a82
opencl : broadcast for soft_max (#14510)
lhez committed 1 year ago
Verified bee28421
vulkan: support mixed/deepseekR1 FA head sizes (#14509)
jeffbolznv committed 1 year ago
Verified 2b72bede
ggml: backward pass for split swiglu (#14483)
JohannesGaessler committed 1 year ago
Verified c8c4495b
Fix conditional enabling following arch checks for ggml-sycl (#14504)
s-Nick committed 1 year ago
Verified 7b63a71a
convert : correct gemma 3n conversion (#14450)
ngxson committed 1 year ago
Verified 0c2ee38a
kv-cache : use ggml_set_rows (#14285)
ggerganov committed 1 year ago
Verified a70c8a0c
ggml : fix FA mask dim 2 and 3 (#14505)
ggerganov committed 1 year ago
Verified 9067487c
ggml : remove kompute backend (#14501)
ggerganov committed 1 year ago
Verified d4cdd9c1
CUDA: add dynamic shared mem to softmax, refactor general usage (#14497)
am17an committed 1 year ago
Verified 55c2646b
gguf-py : add support for chat template jinja files (#14508)
CISC committed 1 year ago
Verified e75ba4c0
llama : initial Mamba-2 support (#9126)
compilade committed 1 year ago
Verified 5d46babd
sync : ggml
ggerganov committed 1 year ago
e17991c4
ggml : add version function to get lib version (ggml/1286)
danbev committed 1 year ago
c46944aa
Set RPATH to "@loader_path" / "$ORIGIN" to ensure executables and dynamic libraries search for dependencies in their origin directory. (#14309)
rotemdan committed 1 year ago
Verified f3ed38d7
CUDA: add softmax broadcast (#14475)
am17an committed 1 year ago
55a1c5a5
CUDA: broadcasting for FlashAttention mask (#14500)
JohannesGaessler committed 1 year ago
12a81af4
vulkan: support softmax/FA batch and broadcast (#14449)
jeffbolznv committed 1 year ago
8875523e
ggml : support bcast ggml_soft_max_ext, ggml_flash_attn_ext (#14435)
ggerganov committed 1 year ago
ec68e84c
opencl : fix possible buffer overflow in dump_tensor (#14490)
zhouwg committed 1 year ago
Verified 307e79d3
simple-chat : fix context-exceeded condition (#14494)
ggerganov committed 1 year ago
Verified d7f5f4e5
opencl : skip empty nodes on cgraph compute (#14491)
EZForever committed 1 year ago
Verified c8a4e470
opencl : update upscale to support align corners (#14488)
lhez committed 1 year ago
Verified 603e43dc
ci : add OpenCL to labeler workflow (#14496)
CISC committed 1 year ago
Verified 611ba4b2
github : add OpenCL backend to issue templates (#14492)
EZForever committed 1 year ago
Verified 85841e12
ggml : Callback before abort (#14481)
ScaledLizard committed 1 year ago
Verified 68b3cd65
Terms Of Use Privacy Policy Refund Policy Impressum