ggerganov/llama.cpp

Pull Requests Commits

ggerganov committed 1 year ago

Verified 4ec0e9ab

llama : add pooling switch

ggerganov committed 1 year ago

Verified e66da356

embeddings : fix llama_batch_init arg

ggerganov committed 1 year ago

Verified 9bbeb0f1

llama : do not use KV cache for non-causal models

ggerganov committed 1 year ago

Verified eb425962

llama : fix embeddings

ggerganov committed 1 year ago

Verified d0347840

ggerganov committed 1 year ago

Verified a0fc6266

add some new ops, fix some operators and add batch operations to certain operators. (ggml/747)

leejet committed 1 year ago

Verified 7d43c585

common : use LLAMA_DEFAULT_SEED (#5855)

dranger003 committed 1 year ago

Verified 82f3e668

main : support special tokens as reverse/anti prompt (#5847)

dranger003 committed 1 year ago

Verified 5a51cc1b

cuda : fix data race in soft max (#5853)

slaren committed 1 year ago

Verified 67be2ce1

readme : add API changes section

ggerganov committed 1 year ago

Verified 231ae28f

llama : allow for user specified embedding pooling type (#5849)

iamlemec committed 1 year ago

Verified 475df1d6

gguf-dump : support i-quants (#5841)

Nindaleth committed 1 year ago

Verified 87c2e8b2

llama : fix llama_copy_state_data with fragmented KV cache (#5840)

compilade committed 1 year ago

Verified de9692a7

ci : schedule slow server tests only on Release or on demand (#5839)

phymbert committed 1 year ago

Verified e6029348

server : init http requests thread pool with --parallel if set (#5836)

phymbert committed 1 year ago

Verified 8ef969af

flake.lock: Update (#5842)

ggerganov committed 1 year ago

Verified fa974646

server: tests: passkey challenge / self-extend with context shift demo (#5832)

phymbert committed 1 year ago

Verified 97311342

llama : add abort_callback to interrupt computation (#5409)

Xarbirus committed 1 year ago

Verified 4a6e2d61

ggml : fix IQ3_S AVX implementation (#5834)

ggerganov committed 1 year ago

Verified 494c8703

convert : automatically fall back to HfVocab if tokenizer.model doesn't exist (#5821)

cebtenzzre committed 1 year ago

Verified 4d4d2366

convert-hf : make model class definitions self-contained (#5825)

cebtenzzre committed 1 year ago

Verified c7a0ad8e

ggml : IQ3_S improvements (#5829)

ikawrakow committed 1 year ago

Verified bbde6eb2

scripts : add pod-llama.sh

ggerganov committed 1 year ago

Verified ef2cd694

llama : refactor internal quantization functions (#5830)

ngxson committed 1 year ago

Verified 6c32d8c7

llama : fix segfault from unknown model arch name (#5820)

compilade committed 1 year ago

Verified 802da009

Support multiple GPUs (split mode) on SYCL backend (#5806)

NeoZhangJianyu committed 1 year ago

Verified 71564139

workflows : remove nocleanup arg for check-requirements.sh (#5826)

crasm committed 1 year ago

Verified 9bf297a0

build(nix): Introduce flake.formatter for `nix fmt` (#5687)

ditsuke committed 1 year ago

Verified cb5e8f7f

convert-hf-to-gguf : require einops for InternLM2ForCausalLM (#5792)

Nold360 committed 1 year ago

Verified da3b9ba2

Older