quantize : fix precedence of cli args (ggml-org#6541)

ggerganov · web-flow · commit b73e564b1608 · 2024-04-08T16:23:01.000+03:00
diff --git 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;
             }
 

Original file line number	Diff line number	Diff line change
`@@ -13562,10 +13562,10 @@ static void llama_model_quantize_internal(const std::string & fname_inp, const s`
`13562`	`13562`	`if (!params->pure && ggml_is_quantized(default_type)) {`
`13563`	`13563`	`new_type = llama_tensor_get_type(qs, new_type, tensor, ftype);`
`13564`	`13564`	`}`
`13565`		`- else if (params->token_embedding_type < GGML_TYPE_COUNT && strcmp(tensor->name, "token_embd.weight") == 0) {`
	`13565`	`+ if (params->token_embedding_type < GGML_TYPE_COUNT && strcmp(tensor->name, "token_embd.weight") == 0) {`
`13566`	`13566`	`new_type = params->token_embedding_type;`
`13567`	`13567`	`}`
`13568`		`- else if (params->output_tensor_type < GGML_TYPE_COUNT && strcmp(tensor->name, "output.weight") == 0) {`
	`13568`	`+ if (params->output_tensor_type < GGML_TYPE_COUNT && strcmp(tensor->name, "output.weight") == 0) {`
`13569`	`13569`	`new_type = params->output_tensor_type;`
`13570`	`13570`	`}`
`13571`	`13571`