]> git.djapps.eu Git - pkg/ggml/sources/llama.cpp/commit
ggml : refactor rope norm/neox (#7634)
authorGeorgi Gerganov <redacted>
Wed, 5 Jun 2024 08:29:20 +0000 (11:29 +0300)
committerGitHub <redacted>
Wed, 5 Jun 2024 08:29:20 +0000 (11:29 +0300)
commit2b3389677a833cee0880226533a1768b1a9508d2
tree3af4beed34ff6c1309d202a7028f5ab85ec43693
parent9973e81c5ccf4f31b3980f5aa73f5cfea8699860
ggml : refactor rope norm/neox (#7634)

* ggml : unify rope norm/neox (CPU)

* ggml : fix compile warning

* ggml : remove GLM rope mode

ggml-ci

* metal : better rope implementation

ggml-ci

* cuda : better rope implementation

ggml-ci

* naming : n_orig_ctx -> n_ctx_orig

ggml-ci

* dev : add reminders to update backends

ggml-ci

* vulkan : fix ggml_rope_ext() usage

* cuda : fix array size + indents

ggml-ci
19 files changed:
examples/baby-llama/baby-llama.cpp
examples/convert-legacy-llama.py
examples/finetune/finetune.cpp
examples/train-text-from-scratch/train-text-from-scratch.cpp
ggml-cuda/rope.cu
ggml-kompute.cpp
ggml-metal.m
ggml-metal.metal
ggml-sycl.cpp
ggml-vulkan.cpp
ggml.c
ggml.h
kompute-shaders/op_rope_f16.comp
kompute-shaders/op_rope_f32.comp
kompute-shaders/rope_common.comp
llama.cpp
tests/test-backend-ops.cpp
tests/test-grad0.cpp
tests/test-rope.cpp