vbatts/llama.cpp
1
0
Fork 0
Code Issues Pull requests Projects Releases Packages Wiki Activity Actions

gguf : support custom alignment value

Browse source
This commit is contained in:
M. Yusuf Sarıgöz 2023-07-31 09:59:36 +03:00
parent 6b3a7b9f4f
commit 7aa0a0e7f7
3 changed files with 10 additions and 2 deletions
Download patch file Download diff file Expand all files Collapse all files

6
gguf.py
View file

@ -220,6 +220,9 @@ class GGUFWriter:
self.write_uint32(
constants.KEY_GENERAL_QUANTIZATION_VERSION, quantization_version)
def write_custom_alignment(self, alignment: int):
self.write_uint32(constants.KEY_GENERAL_ALIGNMENT, alignment)
def write_context_length(self, llm: str, length: int):
self.write_uint32(
constants.KEY_LLM_CONTEXT_LENGTH.format(llm=llm), length)
@ -292,11 +295,12 @@ class GGUFWriter:
if __name__ == "__main__":
# Example usage with a file
gguf_writer = GGUFWriter.open("example.gguf")
gguf_writer.write_header(2, 3)
gguf_writer.write_header(2, 4)
gguf_writer.write_architecture("llama")
gguf_writer.write_uint32("answer", 42) # Write a 32-bit integer
gguf_writer.write_float32("answer_in_float", 42.0) # Write a 32-bit float
gguf_writer.write_custom_alignment(64)
tensor1 = np.ones((32,), dtype=np.float32) * 100.0
tensor2 = np.ones((32,), dtype=np.float32) * 101.0
gguf_writer.write_tensor_info("tensor0", tensor1)