Merge branch 'ggerganov:master' into fix-gptneox

Author: mmnga

convert-falcon-hf-to-gguf.py

@@ -11,11 +11,14 @@
 from pathlib import Path
 from typing import Any
 
-import gguf
 import numpy as np
 import torch
 from transformers import AutoTokenizer  # type: ignore[import]
 
+if 'NO_LOCAL_GGUF' not in os.environ:
+    sys.path.insert(1, str(Path(__file__).parent / 'gguf-py' / 'gguf'))
+import gguf
+
 
 def bytes_to_unicode():
     # ref: https://github.com/openai/gpt-2/blob/master/src/encoder.py

convert-gptneox-hf-to-gguf.py

@@ -11,11 +11,14 @@
 from pathlib import Path
 from typing import Any
 
-import gguf
 import numpy as np
 import torch
 from transformers import AutoTokenizer  # type: ignore[import]
 
+if 'NO_LOCAL_GGUF' not in os.environ:
+    sys.path.insert(1, str(Path(__file__).parent / 'gguf-py' / 'gguf'))
+import gguf
+
 # ref: https://github.com/openai/gpt-2/blob/master/src/encoder.py
 
 

convert-llama-7b-pth-to-gguf.py

@@ -7,9 +7,13 @@
 import sys
 from pathlib import Path
 
-import gguf
 import numpy as np
 
+import os
+if 'NO_LOCAL_GGUF' not in os.environ:
+    sys.path.insert(1, str(Path(__file__).parent / 'gguf-py' / 'gguf'))
+import gguf
+
 # Note: Does not support GGML_QKK_64
 QK_K = 256
 # Items here are (block size, type size)