ggml-org/llama.cpp

Pull Requests Commits

test zero max buffer size

slaren committed 1 year ago

1e6e363d

Clarify default MMQ for CUDA and LLAMA_CUDA_FORCE_MMQ flag (#8115)

isaac-mcfadyen committed 1 year ago

Verified 88540445

CUDA: fix misaligned shared memory read (#8123)

JohannesGaessler committed 1 year ago

Verified c8771ab5

llama : extend llm_build_ffn() to support _scale tensors (#8103)

Eddie-Wang1120 committed 1 year ago

Verified 494165f3

`json`: better support for "type" unions (e.g. nullable arrays w/ typed items) (#7863)

ochafik committed 1 year ago

Verified 9b2f16f8

`json`: fix additionalProperties, allow space after enum/const (#7840)

ochafik committed 1 year ago

Verified 6777c544

fixes #7999 (adds control vectors to all `build_XXX()` functions in `llama.cpp` [needs testing] (#8060)

jukofyork committed 1 year ago

Verified 163d50ad

llama : implement Unigram tokenizer needed by T5 and FLAN-T5 model families (#5763)

fairydreaming committed 1 year ago

Verified 6fcbf682

llama : return nullptr from llama_grammar_init (#8093)

danbev committed 1 year ago

Verified e6bf0077

`json`: support integer minimum, maximum, exclusiveMinimum, exclusiveMaximum (#7797)

ochafik committed 1 year ago

Verified 84631fe1

disable docker CI on pull requests (#8110)

slaren committed 1 year ago

Verified dd047b47

Add healthchecks to llama-server containers (#8081)

codearranger committed 1 year ago

Verified 925c3095

Gguf dump start data offset via --data-offset and some extra refactor (#8054)

mofosyne committed 1 year ago

Verified c8ad3595

cvector: better prompt handling, add "mean vector" method (#8069)

ngxson committed 1 year ago

Verified 49c03c79

Add chat template support for llama-cli (#8068)

ngxson committed 1 year ago

Verified 48e6b92c

SimpleChat v3.1: Boolean chat request options in Settings UI, cache_prompt (#7950)

hanishkvc committed 1 year ago

Verified 3791ad21

Update control vector help (#8104)

HatsuneMikuUwU33 committed 1 year ago

Verified f702a90e

[SYCL] Re-enabled mul_mat_batched_sycl (#8095)

airMeng committed 1 year ago

Verified 083bacce

CUDA: fix matrix multiplication algorithm choice (#8102)

JohannesGaessler committed 1 year ago

Verified 2df373ac

CUDA: fix MMQ writeback for int8 tensor cores (#8100)

JohannesGaessler committed 1 year ago

Verified 3b099bcd

CUDA: use MMQ instead of cuBLAS by default (#8075)

JohannesGaessler committed 1 year ago

Verified a818f302

gguf-py : fix tensor groups for encoder-decoder models in gguf-dump.py (#8090)

fairydreaming committed 1 year ago

Verified d62e4aaa

CUDA: optimize MMQ int8 tensor core performance (#8062)

JohannesGaessler committed 1 year ago

Verified 9a590c82

Option to split during conversion (#6942)

christianazinn committed 1 year ago

Verified 52fc8705

disable publishing the full-rocm docker image (#8083)

slaren committed 1 year ago

Verified 8cb508d0

embedding : more cli arguments (#7458)

YannFollet committed 1 year ago

Verified 646ef4a9

gguf-py, convert-hf : model conversion support for T5 and FLAN-T5 model variants (#5763)

fairydreaming committed 1 year ago

Verified de0d6a68

ggml : remove ggml_task_type and GGML_PERF (#8017)

slaren committed 1 year ago

Verified 95f57bb5

llama : add support for BitnetForCausalLM (#7931)

Eddie-Wang1120 committed 1 year ago

Verified e112b610

server : fix JSON-Scheme typo (#7975)

akx committed 1 year ago

Verified 6a2f298b

Older