ggerganov/llama.cpp

Pull Requests Commits

Remove custom pre attention scaling and use computed value instead.

abetlen committed 1 year ago

51f0bd50

Add custom kq scaling from Gemma2Attention

abetlen committed 1 year ago

a8942790

Merge branch 'add-gemma2-soft-capping' of github.com:ggerganov/llama.cpp into add-gemma2-soft-capping

abetlen committed 1 year ago

6f2464e3

Add default value for attention and final logit softcap value

abetlen committed 1 year ago

bb715992

Merge branch 'master' of github.com:ggerganov/llama.cpp into add-gemma2-soft-capping

abetlen committed 1 year ago

8edf73a7

fix code typo in llama-cli (#8198)

ngxson committed 1 year ago

Verified 72272b83

Update src/llama.cpp

abetlen committed 1 year ago

Verified 3a247181

Disable flash attention for Gemma2

abetlen committed 1 year ago

f4424c15

Add custom add_ functions

abetlen committed 1 year ago

d1137c20

abetlen committed 1 year ago

d3d3c4eb

Add attention and final logit softcapping.

abetlen committed 1 year ago

4d3f17b4

json: attempt to skip slow tests when running under emulator (#8189)

ochafik committed 1 year ago

Verified 8748d8ac

Add MiniCPM, Deepseek V2 chat template + clean up `llama_chat_apply_template_internal` (#8172)

ngxson committed 1 year ago

Verified 26a39bbd

Add SPM infill support (#8016)

CISC committed 1 year ago

Verified 38373cfb

cmake : allow user to override default options (#8178)

slaren committed 1 year ago

Verified b851b3fb

`json`: restore default additionalProperties to false, fix some pattern escapes (#8180)

ochafik committed 1 year ago

Verified 139cc621

llama: Add support for Gemma2ForCausalLM (#8156)

pculliton committed 1 year ago

Verified e57dc620

Add missing items in makefile (#8177)

ngxson committed 1 year ago

Verified a27aa50a

`json`: update grammars/README w/ examples & note about additionalProperties (#8132)

ochafik committed 1 year ago

Verified cb0b06a8

CI: fix release build (Ubuntu+Mac) (#8170)

loonerin committed 1 year ago

Verified 558f44bf

cmake : fix deprecated option names not working (#8171)

slaren committed 1 year ago

Verified 8172ee9d

Add chatml fallback for cpp `llama_chat_apply_template` (#8160)

ngxson committed 1 year ago

Verified 16791b8f

flake.lock: Update (#8071)

ggerganov committed 1 year ago

Verified ab367911

Control vector loading fixes (#8137)

jukofyork committed 1 year ago

Verified 97877eb1

Delete examples/llama.android/llama/CMakeLists.txt (#8165)

criminact committed 1 year ago

Verified 38795265

Add Qwen2MoE 57B-A14B model identifier (#8158)

CISC committed 1 year ago

Verified 6030c612

CUDA: fix MMQ stream-k for --split-mode row (#8167)

JohannesGaessler committed 1 year ago

Verified 85a267da

Added support for Viking pre-tokenizer (#8135)

kustaaya committed 1 year ago

Verified f675b20a

llama : fix CodeLlama FIM token checks (#8144)

CISC committed 1 year ago

Verified 911e35bb

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

criminact committed 1 year ago

Verified ac146628

Older