ggml-org/ggml

Pull Requests Commits

sync : whispper.cpp

ggerganov committed 311 days ago

Verified d579069b

git : remove kompute submodule (#1300)

ggerganov committed 311 days ago

Verified d62df60a

sync : resolve conflicts (#0)

ggerganov committed 311 days ago

9ae55b67

sync : llama.cpp

ggerganov committed 311 days ago

d3fa95e5

vulkan: support SET_ROWS (llama/14587)

jeffbolznv committed 311 days ago

3b7e745a

vulkan: optimizations for deepseek prompt processing (llama/14555)

jeffbolznv committed 311 days ago

5eb294f4

model : support LiquidAI LFM2 hybrid family (llama/14620)

tdakhran committed 311 days ago

5d6c37b4

HIP : Add HIP 7.0+ compatibility for hipBLAS compute types (llama/14634)

slojosic-amd committed 311 days ago

34869695

opencl: add tiled mul_mat_f16_f32 (llama/14535)

rmatif committed 311 days ago

43badec3

opencl: add `set_rows` for `f16` and `f32` (llama/14547)

lhez committed 311 days ago

e08f01d9

Docs: script to auto-generate ggml operations docs (llama/14598)

am17an committed 311 days ago

e16bef2f

SYCL: Initial set_rows kernel implementation (llama/14562)

qnixsynapse committed 311 days ago

e51c43c2

cuda : support Falcon-H1 state size for SSM_SCAN (llama/14602)

compilade committed 311 days ago

9b9fb591

ggml : add ggml_scale_bias (llama/14417)

ngxson committed 311 days ago

3ae98975

ggml : prevent integer overflow in gguf tensor size calculation (llama/14595)

Yuuoniy committed 311 days ago

19356a42

vulkan: optimize flash attention split_k_reduce (llama/14554)

jeffbolznv committed 311 days ago

2f52a6fa

vulkan : fix rope with partial rotation and non-cont src (llama/14582)

jeffbolznv committed 311 days ago

40376b71

cuda : fix rope with partial rotation and non-cont src (llama/14580)

ggerganov committed 311 days ago

14d30dfe

CUDA: add bilinear interpolation for upscale (llama/14563)

am17an committed 311 days ago

1f1f781b

musa: fix build warnings (unused variable) (llama/14561)

yeahdongcn committed 311 days ago

0b617425

CUDA: add bf16 and i32 to getrows (llama/14529)

am17an committed 311 days ago

157456dd

vulkan: increase LOAD_VEC_A to 8 (IQ1/IQ2) or 4 (IQ3) (llama/14485)

netrunnereve committed 311 days ago

8ce0b199

vulkan: fix rms_norm+mul fusion (llama/14545)

jeffbolznv committed 311 days ago

d2f0e767

vulkan: Handle updated FA dim2/3 definition (llama/14518)

jeffbolznv committed 311 days ago

26af1808

opencl: add GELU_ERF (llama/14476)

CISC committed 311 days ago

dadec190

test-backend-ops: add support for specifying output format (llama/14368)

yeahdongcn committed 311 days ago

553ae993

metal : disable fast math in all quantize kernels (llama/14528)

ggerganov committed 311 days ago

abedc006

CANN: Replace aclrtMemsetSync with aclnnInplaceZero operator (llama/14002)

luyhcsu committed 311 days ago

f204a652

ggml : implement GEGLU_ERF and GEGLU_QUICK ops (llama/14445)

CISC committed 311 days ago

6d599cd7

opencl : broadcast for soft_max (llama/14510)

lhez committed 311 days ago

c279888d

Older