ggml-org/llama.cpp

Pull Requests Commits

llama : add llm_build_inp_embd helper

ggerganov committed 1 year ago

Verified 7923b70c

llama : remove extra ; + deduplicate gate_b logic

ggerganov committed 1 year ago

Verified 2073347e

llama : enable warning about not offloaded tensors

ggerganov committed 1 year ago

Verified fc5a26aa

llama : normalize tensor names

ggerganov committed 1 year ago

Verified 0bfdcdd0

llama : update offload functions for KQ tensors

ggerganov committed 1 year ago

Verified 6669cd83

llama : fix input allocation logic

ggerganov committed 1 year ago

Verified 2926ef63

llama : add LLAMA_OFFLOAD_DEBUG + fix starcoder offloading

ggerganov committed 1 year ago

Verified a3f80013

ggerganov committed 1 year ago

Verified 792d1a1b

llama : add llm_build_kqv helper

ggerganov committed 1 year ago

Verified f39e6075

llama : remove obsolete comments in build graphs

ggerganov committed 1 year ago

Verified c9121fdd

llama : simplify falcon Q, K, V computation

ggerganov committed 1 year ago

Verified a104abea

llama : fix llm_build_k_shift to use n_head_kv instead of n_head

ggerganov committed 1 year ago

Verified 31a12f3d

llama : remove obsolete offload names

ggerganov committed 1 year ago

Verified 59908619

llama : add llm_build_kv_store helper

ggerganov committed 1 year ago

Verified 3e046259

llama : fix offloading after recent changes

ggerganov committed 1 year ago

Verified 909d6447

llama : add llm_build_k_shift helper

ggerganov committed 1 year ago

Verified 38728a0b

llama : add llm_build_ffn helper function (#3849)

ggerganov committed 1 year ago

Verified dbf836bb

llama : add llm_build_norm helper function

ggerganov committed 1 year ago

Verified 7db9c96d

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

Older