Merge branch 'gguf-read' into agent-example

Author: ochafik

gguf-py/examples/read_template.py

@@ -0,0 +1,15 @@
+#!/usr/bin/env python3
+import sys
+from pathlib import Path
+
+sys.path.insert(0, str(Path(__file__).parent.parent))
+from gguf.gguf_reader import GGUFReader
+
+if __name__ == '__main__':
+    if len(sys.argv) < 2:
+        print("Usage: read_template.py <path_to_gguf_file>")
+        sys.exit(1)
+    gguf_file_path = sys.argv[1]
+
+    reader = GGUFReader(gguf_file_path)
+    print(reader.read_field(reader.fields['tokenizer.chat_template']))

gguf-py/gguf/gguf_reader.py

@@ -121,6 +121,21 @@ def __init__(self, path: os.PathLike[str] | str, mode: Literal['r' | 'r+' | 'c']
     # Fetch a key/value metadata field by key.
     def get_field(self, key: str) -> Union[ReaderField, None]:
         return self.fields.get(key, None)
+    
+    def read_field(self, field):
+        if not field.types:
+            return None
+        if field.types[:1] == [GGUFValueType.ARRAY]:
+            itype = field.types[-1]
+            if itype == GGUFValueType.STRING:
+                return [str(bytes(field.parts[idx]), encoding="utf-8") for idx in field.data]
+            else:
+                return [pv for idx in field.data for pv in field.parts[idx].tolist()]
+        elif field.types[0] == GGUFValueType.STRING:
+            return str(bytes(field.parts[-1]), encoding="utf-8")
+        else:
+            assert(field.types[0] in self.gguf_scalar_to_np)
+            return field.parts[-1].tolist()[0]
 
     # Fetch a tensor from the list by index.
     def get_tensor(self, idx: int) -> ReaderTensor:

gguf-py/scripts/gguf-dump.py

@@ -41,11 +41,7 @@ def dump_metadata(reader: GGUFReader, args: argparse.Namespace) -> None:
             pretty_type = str(field.types[-1].name)
         print(f'  {n:5}: {pretty_type:10} | {len(field.data):8} | {field.name}', end = '')
         if len(field.types) == 1:
-            curr_type = field.types[0]
-            if curr_type == GGUFValueType.STRING:
-                print(' = {0}'.format(repr(str(bytes(field.parts[-1]), encoding='utf8')[:60])), end = '')
-            elif field.types[0] in reader.gguf_scalar_to_np:
-                print(' = {0}'.format(field.parts[-1][0]), end = '')
+            print(' = {0}'.format(repr(reader.read_field(field))[:60]), end = '')
         print()
     if args.no_tensors:
         return
@@ -75,17 +71,7 @@ def dump_metadata_json(reader: GGUFReader, args: argparse.Namespace) -> None:
         metadata[field.name] = curr
         if field.types[:1] == [GGUFValueType.ARRAY]:
             curr["array_types"] = [t.name for t in field.types][1:]
-            if not args.json_array:
-                continue
-            itype = field.types[-1]
-            if itype == GGUFValueType.STRING:
-                curr["value"] = [str(bytes(field.parts[idx]), encoding="utf-8") for idx in field.data]
-            else:
-                curr["value"] = [pv for idx in field.data for pv in field.parts[idx].tolist()]
-        elif field.types[0] == GGUFValueType.STRING:
-            curr["value"] = str(bytes(field.parts[-1]), encoding="utf-8")
-        else:
-            curr["value"] = field.parts[-1].tolist()[0]
+        curr["value"] = reader.read_field(field)
     if not args.no_tensors:
         for idx, tensor in enumerate(reader.tensors):
             tensors[tensor.name] = {