From 254a7a7a5ff4c874ff8488f1f5cbdd7e9c89d682 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Johannes=20G=C3=A4=C3=9Fler?= Date: Wed, 14 Jun 2023 19:47:19 +0200 Subject: CUDA full GPU acceleration, KV cache in VRAM (#1827) * Fixed CUDA RoPE * ggml_cuda_mul_mat_vec_p021 * ggml_cuda_scale * ggml_cuda_diag_mask_inf * ggml_is_permuted * ggml_cuda_cpy * flatten rows for ggml_cuda_op * Added a --low-vram option * Fixed Windows performance * Fixed LLAMA_CUDA_DMMV_Y > 1 for WizardLM --- ggml.h | 1 + 1 file changed, 1 insertion(+) (limited to 'ggml.h') diff --git a/ggml.h b/ggml.h index f2a9176..9b0c846 100644 --- a/ggml.h +++ b/ggml.h @@ -485,6 +485,7 @@ extern "C" { GGML_API bool ggml_is_transposed(const struct ggml_tensor * tensor); GGML_API bool ggml_is_contiguous(const struct ggml_tensor * tensor); + GGML_API bool ggml_is_permuted (const struct ggml_tensor * tensor); // use this to compute the memory overhead of a tensor GGML_API size_t ggml_tensor_overhead(void); -- cgit v1.2.3