From 4760e7cc0b68570d58f55e8dda469805d1759d0d Mon Sep 17 00:00:00 2001 From: Georgi Gerganov Date: Mon, 13 Nov 2023 14:16:23 +0200 Subject: sync : ggml (backend v2) (#3912) * sync : ggml (backend v2) (wip) * sync : migrate examples and llama.cpp to dynamic graphs (wip) * sync : update tests + fix max op params to 64 ggml-ci * sync : ggml-cuda ggml-ci * llama : fix save/load state context size ggml-ci * sync : try to fix build on tvOS * sync : pass custom graph sizes in training examples * sync : update graph copies to new ggml API * sync : update sync-ggml.sh with new files * scripts : fix header in sync script * train : fix context size calculations * llama : increase inference graph size up to 4096 nodes * train : allocate grads for backward graphs * train : allocate grads for gb_tmp --- examples/export-lora/export-lora.cpp | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) (limited to 'examples/export-lora/export-lora.cpp') diff --git a/examples/export-lora/export-lora.cpp b/examples/export-lora/export-lora.cpp index d803cfd5..c8754ce7 100644 --- a/examples/export-lora/export-lora.cpp +++ b/examples/export-lora/export-lora.cpp @@ -240,7 +240,7 @@ static struct lora_data * load_lora(struct lora_info * info) { } struct ggml_init_params params_ggml; - params_ggml.mem_size = ggml_tensor_overhead() * GGML_MAX_NODES; + params_ggml.mem_size = ggml_tensor_overhead() * GGML_DEFAULT_GRAPH_SIZE; params_ggml.mem_buffer = NULL; params_ggml.no_alloc = true; result->ctx = ggml_init(params_ggml); @@ -334,7 +334,7 @@ static bool apply_lora(struct ggml_tensor * tensor, struct lora_data * lora, int float scaling = lora->info.scale * (float)lora->lora_alpha / (float)lora->lora_r; struct ggml_init_params params; - params.mem_size = GGML_OBJECT_SIZE + GGML_GRAPH_SIZE + ggml_tensor_overhead()*4 + GGML_MEM_ALIGN*5; + params.mem_size = GGML_OBJECT_SIZE + ggml_graph_overhead() + ggml_tensor_overhead()*4 + GGML_MEM_ALIGN*5; params.mem_buffer = NULL; params.no_alloc = true; struct ggml_context * ctx = NULL; -- cgit v1.2.3