ngxson/llama.cpp

Pull Requests Commits

vulkan: workaround for AMD Windows driver 16 bit unpack8 bug (#12472)

netrunnereve committed 1 year ago

Verified 30c42ef5

add batch.clear() and batch.n_tokens()

ngxson committed 1 year ago

3802ff2a

move various places to batch.add_text

ngxson committed 1 year ago

2cec1cff

model : do not repack if a GPU device is present (#12498)

ggerganov committed 1 year ago

Verified af04481e

add cpp batch.add_text wrapper

ngxson committed 1 year ago

2134cabf

Merge branch 'master' into xsn/private_batch_api

ngxson committed 1 year ago

c5a01763

chore : cleanup llama_model_loader::TENSOR_ usage (#12492)

CISC committed 1 year ago

Verified 960e7260

llama-tts : avoid crashes related to bad model file paths (#12482)

marcoStocchi committed 1 year ago

Verified ea1518e8

[SYCL] Fix build on Windows when ccache enabled (#9954) (#9976)

MakeDecisionWorth committed 1 year ago

Verified 1aa87ee5

sycl: cleanup oneDNN related code (#12097)

sgeor255 committed 1 year ago

Verified 9ffcc9e3

remove C API llama_batch_ext_init_from_text

ngxson committed 1 year ago

30f1db99

Merge pull request #16 from ggml-org/xsn/private_batch_api_pooling_none

ngxson committed 1 year ago

Verified bd51d63b

server : remove old commented code [no ci]

ggerganov committed 1 year ago

Verified b8b17327

webui : Prevent rerendering on textarea input (#12299)

woof-dog committed 1 year ago

Verified e0464306

server : avoid common_batch

ggerganov committed 1 year ago

Verified 8a23b4a5

llama : make Qwen2MoE QKV bias optional (#12477)

CISC committed 1 year ago

Verified dbb3a473

ggml : block interleaving support for Q4_K quantization for x86 AVX2 architecture (#12332)

Srihari-mcw committed 1 year ago

Verified 3d82dbcb

perplexity : avoid common_batch

ggerganov committed 1 year ago

Verified 76fd7d6f

convert : avoid calls to tokenizer.added_tokens_decoder (#12473)

bartowski1182 committed 1 year ago

Verified 732b5fbf

context : clear sets containing encoder output sequence ids before storing new values (#12470)

fairydreaming committed 1 year ago

Verified 568013d0

CUDA: Improve flash decoding kernel GPU occupancy for BS=1 case (#12183)

gaugarg-nv committed 1 year ago

Verified 517b5ddb

vulkan: optimize iq1 coopmat2 dequant functions (#12427)

jeffbolznv committed 1 year ago

Verified a9b59288

embedding : avoid common_batch

ggerganov committed 1 year ago

Verified 8b80d683

retrieval : avoid common_batch

ggerganov committed 1 year ago

Verified 6f54ee66

Fix visionOS build and add CI (#12415)

guusw committed 1 year ago

Verified 0fd8487b

android : fix permission

ngxson committed 1 year ago

32c2c41d

swift : adapt to new API

ggerganov committed 1 year ago

Verified 96ca6e8d

android : adapt to new API

ggerganov committed 1 year ago

Verified b0db7fc2

Merge pull request #15 from ggml-org/xsn/private_batch_api

ngxson committed 1 year ago

Verified 23d74073

llama : add support for GPT2, Bloom and CodeShell tied word embeddings (#12456)

CISC committed 1 year ago

Verified 108e53c2

Newer Older