From b73e564b16086845a8b4fffd26e22685d3e0c3db Mon Sep 17 00:00:00 2001 From: Georgi Gerganov Date: Mon, 8 Apr 2024 16:23:01 +0300 Subject: [PATCH] quantize : fix precedence of cli args (#6541) --- llama.cpp | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) diff --git a/llama.cpp b/llama.cpp index 9dde3efd0..89ea3fe1a 100644 --- a/llama.cpp +++ b/llama.cpp @@ -13562,10 +13562,10 @@ static void llama_model_quantize_internal(const std::string & fname_inp, const s if (!params->pure && ggml_is_quantized(default_type)) { new_type = llama_tensor_get_type(qs, new_type, tensor, ftype); } - else if (params->token_embedding_type < GGML_TYPE_COUNT && strcmp(tensor->name, "token_embd.weight") == 0) { + if (params->token_embedding_type < GGML_TYPE_COUNT && strcmp(tensor->name, "token_embd.weight") == 0) { new_type = params->token_embedding_type; } - else if (params->output_tensor_type < GGML_TYPE_COUNT && strcmp(tensor->name, "output.weight") == 0) { + if (params->output_tensor_type < GGML_TYPE_COUNT && strcmp(tensor->name, "output.weight") == 0) { new_type = params->output_tensor_type; }