ggerganov/llama.cpp

Pull Requests Commits

ggerganov committed 133 days ago

Verified 37bdfbef

ggerganov committed 133 days ago

Verified efc33ea6

ggerganov committed 134 days ago

Verified 7664390b

kv-cache : utilize ggml_set_rows broadcast

ggerganov committed 134 days ago

Verified 36f8e20d

cont : support non-continuous slots

ggerganov committed 134 days ago

Verified 332f0735

cont : kv-cells cp/set for non-cont slots

ggerganov committed 134 days ago

Verified 39d0b1e8

cont : migrate to using set of indices instead of slot head

ggerganov committed 134 days ago

Verified f875d6cb

cont : gate the ggml_set_rows usage with env var

ggerganov committed 134 days ago

Verified db2bb378

kv-cache : use ggml_set_rows

ggerganov committed 134 days ago

Verified 79dac3c8

ggml : fix supports_op

rgerganov committed 134 days ago

Verified 1f647b59

ggml : simplify forward_dup_f32

rgerganov committed 134 days ago

Verified eba97574

metal : add ggml_set_rows implementation

ggerganov committed 134 days ago

Verified c0cfc2f7

tests : add ggml_set_rows

ggerganov committed 134 days ago

Verified 828e5d2f

ggml : ggml_set_rows update comment + better index name

ggerganov committed 134 days ago

Verified e73690a6

ggml : support GGML_TYPE_F32 ".from_float" trait

ggerganov committed 134 days ago

Verified e8970972

ggml : ggml_set_rows support quantized dst

ggerganov committed 134 days ago

Verified 630c84a2

ggml : ggml_set_rows support broadcast

ggerganov committed 134 days ago

Verified df71c803

ggml : add ggml_is_contiguous_rows

ggerganov committed 134 days ago

Verified 313a444b

ggml : add repeat impl for i64

ggerganov committed 134 days ago

Verified 695b6b70

use I64 for indices

rgerganov committed 134 days ago

Verified f2cd962f

ggml : add ggml_set_rows

rgerganov committed 134 days ago

Verified c1a581a1

kv-cells : fix tracking of seq_pos (#14339)

ggerganov committed 134 days ago

Verified 7b50d589

vulkan: update windows SDK in CI (#14334)

jeffbolznv committed 134 days ago

Verified 3a9457df

quantize : handle user-defined pruning of whole layers (blocks) (#13037)

EAddario committed 134 days ago

Verified fa4a9f2a

gguf-py : fix SpecialVocab parsing when post_processor is null (#14330)

CISC committed 135 days ago

Verified 238005c2

run : avoid double tokenization (#14327)

retr0reg committed 135 days ago

Verified 66aba7ac

examples : fix is_first logic for tokenization (#14329)

ggerganov committed 135 days ago

Verified f1f5e82d

HIP: enable vec fattn on RDNA4 (#14323)

IMbackK committed 135 days ago

Verified af3373f1

mtmd : fix Pixtral OOM with large images by capping image_size to 1024 (#14326)

yuiseki committed 135 days ago

Verified 5d5c066d

common : use std::string_view now that we target c++17 (#14319)

CISC committed 135 days ago

Verified 40bfa04c

Older