ggerganov/llama.cpp

Pull Requests Commits

Merge branch 'master' into llama-refactor

ggerganov committed 1 year ago

Verified afb39292

flake.nix: fix for rocm 5.7 (#3853)

Tungsten842 committed 1 year ago

Verified 07178c98

llama : add llm_build helper functions (#3848)

ggerganov committed 1 year ago

Verified 5baefef4

ggml : move FP16 <-> FP32 code to ggml-impl.h (#3861)

ggerganov committed 1 year ago

Verified 207b5190

Extend llama_kv_cache_seq_rm to allow matching any sequence (#3843)

KerfuffleV2 committed 1 year ago

Verified 6e08281e

make : remove unnecessary dependency on build-info.h (#3842)

cebtenzzre committed 1 year ago

Verified 2046eb43

llama : fix kv shift bug (#3835)

ggerganov committed 1 year ago

Verified 71a09da3

ggml : quantization refactoring (#3833)

ggerganov committed 1 year ago

Verified d69d777c

llama : remove obsolete map for layer counting

ggerganov committed 1 year ago

Verified 210e6e5d

llama : comment

ggerganov committed 1 year ago

Verified 79ad7344

llama : add functional header

ggerganov committed 1 year ago

Verified 76108793

llama : add layer index to all tensor names

ggerganov committed 1 year ago

Verified 8925cf9e

llama : refactor tensor offloading as callback

ggerganov committed 1 year ago

Verified 1e9c5443

llama : move refact in correct place + optimize graph input

ggerganov committed 1 year ago

Verified da936188

llama : try to fix build

ggerganov committed 1 year ago

Verified 739b85c9

llama : fix non-CUDA build

ggerganov committed 1 year ago

Verified 25cfbf67

llama : try to optimize offloading code

ggerganov committed 1 year ago

Verified b4ad03b3

llama : fix res_norm offloading

ggerganov committed 1 year ago

Verified 79617902

llama : do tensor offload only with CUDA

ggerganov committed 1 year ago

Verified e14aa461

llama : factor graph input into a function

ggerganov committed 1 year ago

Verified 0dc05b84

llama : support offloading result_norm + comments

ggerganov committed 1 year ago

Verified 4e98897e

llama : comments

ggerganov committed 1 year ago

Verified 51c4f9ee

llama : update offload log messages to print node index

ggerganov committed 1 year ago

Verified 3af87713

llama : offload rest of the models

ggerganov committed 1 year ago

Verified 83d2c437

llama : factor out tensor offloading outside the build call (wip)

ggerganov committed 1 year ago

Verified 38aca9e1

metal : disable kernel load log

ggerganov committed 1 year ago

Verified 5946d98f

llama : factor out ggml-alloc from graph graph build functions

ggerganov committed 1 year ago

Verified 8b2420d2

flake : update flake.lock for newer transformers version + provide extra dev shell (#3797)

Green-Sky committed 1 year ago

Verified ff3bad83

metal : try cwd for ggml-metal.metal if bundle lookup fails (#3793)

akx committed 1 year ago

Verified 82a6646e

issues : change label from bug to bug-unconfirmed (#3748)

ggerganov committed 1 year ago

Verified ba231e8a

Older