From: Matthew Tejo <redacted>
Date: Tue, 7 Nov 2023 07:43:59 +0000 (-0800)
Subject: Use params when loading models in llava-cli (#3976)
X-Git-Tag: upstream/0.0.4488~2992
X-Git-Url: https://git.djapps.eu/?a=commitdiff_plain;h=54b4df8886103b436a4bb3b60f4d84824f9e8868;p=pkg%2Fggml%2Fsources%2Fllama.cpp

Use params when loading models in llava-cli (#3976)

llava-cli was loading models with default params and ignoring settings
from the cli. This switches to a generic function to load the params
from the cli options.
---

diff --git a/examples/llava/llava-cli.cpp b/examples/llava/llava-cli.cpp
index 19374c67..633afd1d 100644
--- a/examples/llava/llava-cli.cpp
+++ b/examples/llava/llava-cli.cpp
@@ -242,18 +242,16 @@ static struct llava_context * llava_init(gpt_params * params) {
 
     llama_backend_init(params->numa);
 
-    llama_model_params model_params = llama_model_default_params();
+    llama_model_params model_params = llama_model_params_from_gpt_params(*params);
+
     llama_model * model = llama_load_model_from_file(params->model.c_str(), model_params);
     if (model == NULL) {
         fprintf(stderr , "%s: error: unable to load model\n" , __func__);
         return NULL;
     }
 
-    llama_context_params ctx_params = llama_context_default_params();
-
+    llama_context_params ctx_params = llama_context_params_from_gpt_params(*params);
     ctx_params.n_ctx           = params->n_ctx < 2048 ? 2048 : params->n_ctx; // we need a longer context size to process image embeddings
-    ctx_params.n_threads       = params->n_threads;
-    ctx_params.n_threads_batch = params->n_threads_batch == -1 ? params->n_threads : params->n_threads_batch;
 
     llama_context * ctx_llama = llama_new_context_with_model(model, ctx_params);