gguf-py : decouple adding metadata from writing in GGUFWriter (#7827)

Main changes of this PR is to consolidate GGUFWriter.add_key and GGUFWriter.add_val into GGUFWriter.add_key_value. 

In addition use_temp_file is now opt-in instead of opt-out defaulting to False.

Also GGUFWriter now does not require output file name until when actually writing to it.

And GGUFWriter doesn't really need to eagerly prepare the data layout of the metadata

Author: compilade

convert-hf-to-gguf.py

@@ -47,7 +47,7 @@ class Model:
     _model_classes: dict[str, type[Model]] = {}
 
     dir_model: Path
-    ftype: int
+    ftype: gguf.LlamaFileType
     is_big_endian: bool
     endianess: gguf.GGUFEndian
     use_temp_file: bool
@@ -94,7 +94,7 @@ def __init__(self, dir_model: Path, ftype: gguf.LlamaFileType, fname_out: Path,
         ftype_lw: str = ftype_up.lower()
         # allow templating the file name with the output ftype, useful with the "auto" ftype
         self.fname_out = fname_out.parent / fname_out.name.format(ftype_lw, outtype=ftype_lw, ftype=ftype_lw, OUTTYPE=ftype_up, FTYPE=ftype_up)
-        self.gguf_writer = gguf.GGUFWriter(self.fname_out, gguf.MODEL_ARCH_NAMES[self.model_arch], endianess=self.endianess, use_temp_file=self.use_temp_file)
+        self.gguf_writer = gguf.GGUFWriter(path=None, arch=gguf.MODEL_ARCH_NAMES[self.model_arch], endianess=self.endianess, use_temp_file=self.use_temp_file)
 
     @classmethod
     def __init_subclass__(cls):
@@ -324,13 +324,13 @@ def write_tensors(self):
 
     def write(self):
         self.write_tensors()
-        self.gguf_writer.write_header_to_file()
+        self.gguf_writer.write_header_to_file(self.fname_out)
         self.gguf_writer.write_kv_data_to_file()
         self.gguf_writer.write_tensors_to_file(progress=True)
         self.gguf_writer.close()
 
     def write_vocab(self):
-        self.gguf_writer.write_header_to_file()
+        self.gguf_writer.write_header_to_file(self.fname_out)
         self.gguf_writer.write_kv_data_to_file()
         self.gguf_writer.close()