]> git.djapps.eu Git - pkg/ggml/sources/ggml/commit
kompute : improve backend to pass test_backend_ops (llama/10542)
authorSergio López <redacted>
Thu, 28 Nov 2024 11:51:38 +0000 (12:51 +0100)
committerGeorgi Gerganov <redacted>
Tue, 3 Dec 2024 19:05:37 +0000 (21:05 +0200)
commitd9dafe019177445cbcb388786115d24921608920
tree75628a3e7bee851f07fdbd889cafed8a5ecb4b8a
parentd22c42e74dc2d9af4816f0263fe396e742a9df18
kompute : improve backend to pass test_backend_ops (llama/10542)

* kompute: op_unary: reject unsupported parameters

Signed-off-by: Sergio Lopez <redacted>
* kompute: softmax: implement ALiBi support

Signed-off-by: Sergio Lopez <redacted>
* kompute: rope: implement neox and phi3 support

Signed-off-by: Sergio Lopez <redacted>
* kompute: op_mul_mat_q4_k permutted support

Signed-off-by: Sergio Lopez <redacted>
* kompute: op_mul_mat_[q4_0|q4_1|q8_0] permutted support

Signed-off-by: Sergio Lopez <redacted>
* kompute: op_mul_mat_f16 permutted support

Signed-off-by: Sergio Lopez <redacted>
* kompute: op_mul_mat_q6_k permutted support

Signed-off-by: Sergio Lopez <redacted>
---------

Signed-off-by: Sergio Lopez <redacted>
16 files changed:
src/ggml-kompute/CMakeLists.txt
src/ggml-kompute/ggml-kompute.cpp
src/ggml-kompute/kompute-shaders/common.comp
src/ggml-kompute/kompute-shaders/op_mul_mat_f16.comp
src/ggml-kompute/kompute-shaders/op_mul_mat_q4_k.comp
src/ggml-kompute/kompute-shaders/op_mul_mat_q6_k.comp
src/ggml-kompute/kompute-shaders/op_mul_mv_q_n.comp
src/ggml-kompute/kompute-shaders/op_mul_mv_q_n_pre.comp
src/ggml-kompute/kompute-shaders/op_rope_f16.comp [deleted file]
src/ggml-kompute/kompute-shaders/op_rope_f32.comp [deleted file]
src/ggml-kompute/kompute-shaders/op_rope_neox_f16.comp [new file with mode: 0644]
src/ggml-kompute/kompute-shaders/op_rope_neox_f32.comp [new file with mode: 0644]
src/ggml-kompute/kompute-shaders/op_rope_norm_f16.comp [new file with mode: 0644]
src/ggml-kompute/kompute-shaders/op_rope_norm_f32.comp [new file with mode: 0644]
src/ggml-kompute/kompute-shaders/op_softmax.comp
src/ggml-kompute/kompute-shaders/rope_common.comp