diff options
Diffstat (limited to 'gguf-py/gguf/gguf_reader.py')
| -rw-r--r-- | gguf-py/gguf/gguf_reader.py | 12 |
1 files changed, 9 insertions, 3 deletions
diff --git a/gguf-py/gguf/gguf_reader.py b/gguf-py/gguf/gguf_reader.py index 1c10f575..33afac55 100644 --- a/gguf-py/gguf/gguf_reader.py +++ b/gguf-py/gguf/gguf_reader.py @@ -242,12 +242,15 @@ class GGUFReader: n_bytes = n_elems * type_size // block_size data_offs = int(start_offs + offset_tensor[0]) item_type: npt.DTypeLike - if ggml_type == GGMLQuantizationType.F32: + if ggml_type == GGMLQuantizationType.F16: + item_count = n_elems + item_type = np.float16 + elif ggml_type == GGMLQuantizationType.F32: item_count = n_elems item_type = np.float32 - elif ggml_type == GGMLQuantizationType.F16: + elif ggml_type == GGMLQuantizationType.F64: item_count = n_elems - item_type = np.float16 + item_type = np.float64 elif ggml_type == GGMLQuantizationType.I8: item_count = n_elems item_type = np.int8 @@ -257,6 +260,9 @@ class GGUFReader: elif ggml_type == GGMLQuantizationType.I32: item_count = n_elems item_type = np.int32 + elif ggml_type == GGMLQuantizationType.I64: + item_count = n_elems + item_type = np.int64 else: item_count = n_bytes item_type = np.uint8 |