CS348Project/llama.cpp
5
0
Fork 0
mirror of https://github.com/ggml-org/llama.cpp.git synced 2025-10-31 08:51:55 +00:00
Code Issues Packages Projects Releases Wiki Activity

gguf : add Python script to convert GGMLv3 LLaMA models to GGUF (#2682)

Browse Source 
* First pass at converting GGMLv3 LLaMA models to GGUF

* Cleanups, better output during conversion

* Fix vocab space conversion logic

* More vocab conversion fixes

* Add description to converted GGUF files

* Improve help text, expand warning

* Allow specifying name and description for output GGUF

* Allow overriding vocab and hyperparams from original model metadata

* Use correct params override var name

* Fix wrong type size for Q8_K

Better handling of original style metadata

* Set default value for gguf add_tensor raw_shape KW arg
This commit is contained in:
Kerfuffle
2023-08-21 08:45:52 -06:00
committed by GitHub
parent 6490ff7198
commit e06cbcee73
2 changed files with 374 additions and 12 deletions
Download Patch File Download Diff File Expand all files Collapse all files

52
gguf.py
View File

@@ -5,7 +5,7 @@ import tempfile
import numpy as np
from enum import IntEnum, auto
from typing import Any, IO, List
from typing import Any, IO, List, Optional
#
# constants
@@ -325,8 +325,20 @@ def get_tensor_name_map(arch: MODEL_ARCH, n_blocks: int) -> dict:
class GGMLQuantizationType(IntEnum):
F32 = 0
F16 = 1
F32 = 0
F16 = 1
Q4_0 = 2
Q4_1 = 3
Q5_0 = 6
Q5_1 = 7
Q8_0 = 8
Q8_1 = 9
Q2_K = 10
Q3_K = 11
Q4_K = 12
Q5_K = 13
Q6_K = 14
Q8_K = 15
class GGUFValueType(IntEnum):
@@ -359,7 +371,7 @@ class GGUFValueType(IntEnum):
class GGUFWriter:
def __init__(self, path: str, arch: str):
def __init__(self, path: str, arch: str, use_temp_file = True):
self.fout = open(path, "wb")
self.arch = arch
self.offset_tensor = 0
@@ -369,6 +381,8 @@ class GGUFWriter:
self.ti_data = b""
self.ti_data_count = 0
self.add_architecture()
self.use_temp_file = use_temp_file
self.tensors = []
def write_header_to_file(self):
self.fout.write(struct.pack("<I", GGUF_MAGIC))
@@ -476,8 +490,8 @@ class GGUFWriter:
def ggml_pad(x: int, n: int) -> int:
return ((x + n - 1) // n) * n
def add_tensor_info(self, name: str, tensor_shape: np.ndarray, tensor_dtype: np.dtype, tensor_nbytes: int):
assert tensor_dtype in (np.float32, np.float16), "Only F32 and F16 tensors are supported for now"
def add_tensor_info(self, name: str, tensor_shape: np.ndarray, tensor_dtype: np.dtype, tensor_nbytes: int, raw_dtype: Optional[GGMLQuantizationType] = None):
assert raw_dtype is not None or tensor_dtype in (np.float32, np.float16), "Only F32 and F16 tensors are supported for now"
encoded_name = name.encode("utf8")
self.ti_data += struct.pack("<I", len(encoded_name))
@@ -486,23 +500,30 @@ class GGUFWriter:
self.ti_data += struct.pack("<I", n_dims)
for i in range(n_dims):
self.ti_data += struct.pack("<I", tensor_shape[n_dims - 1 - i])
dtype = GGMLQuantizationType.F32 if tensor_dtype == np.float32 else GGMLQuantizationType.F16
if raw_dtype is None:
dtype = GGMLQuantizationType.F32 if tensor_dtype == np.float32 else GGMLQuantizationType.F16
else:
dtype = raw_dtype
self.ti_data += struct.pack("<I", dtype)
self.ti_data += struct.pack("<Q", self.offset_tensor)
self.offset_tensor += GGUFWriter.ggml_pad(tensor_nbytes, self.data_alignment)
self.ti_data_count += 1
def add_tensor(self, name: str, tensor: np.ndarray):
if not hasattr(self, "temp_file"):
def add_tensor(self, name: str, tensor: np.ndarray, raw_shape: Optional[np.ndarray] = None, raw_dtype: Optional[GGMLQuantizationType] = None):
if self.use_temp_file and not hasattr(self, "temp_file"):
self.temp_file = tempfile.SpooledTemporaryFile(mode="w+b", max_size=256*1024*1024)
self.temp_file.seek(0)
self.add_tensor_info(name, tensor.shape, tensor.dtype, tensor.nbytes)
self.add_tensor_info(name, raw_shape if raw_shape is not None else tensor.shape, tensor.dtype, tensor.nbytes, raw_dtype = raw_dtype)
pad = GGUFWriter.ggml_pad(tensor.nbytes, self.data_alignment) - tensor.nbytes
if not self.use_temp_file:
self.tensors.append((tensor, pad))
return
tensor.tofile(self.temp_file)
pad = GGUFWriter.ggml_pad(tensor.nbytes, self.data_alignment) - tensor.nbytes
if pad != 0:
self.temp_file.write(bytes([0] * pad))
@@ -524,6 +545,13 @@ class GGUFWriter:
if pad != 0:
self.fout.write(bytes([0] * pad))
if not self.use_temp_file:
for (currtensor, currpad) in self.tensors:
currtensor.tofile(self.fout)
if currpad != 0:
self.fout.write(bytes([0] * currpad))
return
self.temp_file.seek(0)
shutil.copyfileobj(self.temp_file, self.fout)