Only attempt to use mmap for the lora base model if it is supported

slaren · slaren · commit 63da54e0164a · 2023-04-16T18:30:27.000+02:00
diff --git a/llama.cpp b/llama.cpp
@@ -1840,7 +1840,9 @@ int llama_apply_lora_from_file_internal(struct llama_context * ctx, const char *
         model_loader->ggml_ctx = base_ctx;
 
         // maybe this should in llama_model_loader
-        model_loader->mapping.reset(new llama_mmap(&model_loader->file_loaders.at(0)->file, false));
+        if (model_loader->use_mmap) {
+            model_loader->mapping.reset(new llama_mmap(&model_loader->file_loaders.at(0)->file, /* prefetch */ false));
+        }
     }
 
     // read tensors and apply

Original file line number	Diff line number	Diff line change
`@@ -1840,7 +1840,9 @@ int llama_apply_lora_from_file_internal(struct llama_context * ctx, const char *`
`1840`	`1840`	`model_loader->ggml_ctx = base_ctx;`
`1841`	`1841`
`1842`	`1842`	`// maybe this should in llama_model_loader`
`1843`		`- model_loader->mapping.reset(new llama_mmap(&model_loader->file_loaders.at(0)->file, false));`
	`1843`	`+ if (model_loader->use_mmap) {`
	`1844`	`+ model_loader->mapping.reset(new llama_mmap(&model_loader->file_loaders.at(0)->file, /* prefetch */ false));`
	`1845`	`+ }`
`1844`	`1846`	`}`
`1845`	`1847`
`1846`	`1848`	`// read tensors and apply`