ggerganov/llama.cpp

Pull Requests Commits

convert : use non-fast T5 tokenizer

ggerganov committed 1 year ago

Verified 703764a3

llama : UGM tokenizer init with UNK tokens instead of PAD

ggerganov committed 1 year ago

Verified 17bb0eae

convert : add t5 tokenizer tests

ggerganov committed 1 year ago

Verified 9eb5d561

llama : quantization-related fixes for T5

sszymczy committed 1 year ago

6dc9eb40

llama : whitespace formatting

sszymczy committed 1 year ago

7d7fff46

Merge remote-tracking branch 'upstream/master' into t5-clean-3

sszymczy committed 1 year ago

7293243d

Fix llama-android.cpp for error - "common/common.h not found" (#8145)

criminact committed 1 year ago

Verified ac146628

clip : suppress unused variable warnings (#8105)

danbev committed 1 year ago

Verified 9b31a40c

scripts : fix filename sync

ggerganov committed 1 year ago

Verified c70d117c

ci : publish new docker images only when the files change (#8142)

slaren committed 1 year ago

Verified ae5d0f4b

ggml : add GGML_CUDA_USE_GRAPHS option, restore GGML_CUDA_FORCE_CUBLAS (cmake) (#8140)

slaren committed 1 year ago

Verified 31ec3993

make : fix missing -O3 (#8143)

slaren committed 1 year ago

Verified c7ab7b61

ggerganov committed 1 year ago

Verified f2d48fff

authors : regen

ggerganov committed 1 year ago

Verified 4713bf30

devops : remove clblast + LLAMA_CUDA -> GGML_CUDA (#8139)

ggerganov committed 1 year ago

Verified 0e814dfc

readme : update API notes

ggerganov committed 1 year ago

Verified a95631ee

llama : make pos_bias contiguous for CUDA

sszymczy committed 1 year ago

c4ded1a8

llama : updated llm_build_ffn() calls to new API in build_t5()

sszymczy committed 1 year ago

bad0cafe

llama : reorganize source code + improve CMake (#8006)

ggerganov committed 1 year ago

Verified f3f65429

Merge branch 'ggerganov:master' into t5-clean-3

fairydreaming committed 1 year ago

Verified 1c8d37a2

llama : add inference support and model types for T5 and FLAN-T5 model families

sszymczy committed 1 year ago

45681a57

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

Older