85 lines
1.9 KiB
Python
Executable File
85 lines
1.9 KiB
Python
Executable File
#!/usr/bin/env python3
|
|
import sys
|
|
from pathlib import Path
|
|
import json
|
|
|
|
import numpy as np
|
|
|
|
# Necessary to load the local gguf package
|
|
sys.path.insert(0, str(Path(__file__).parent.parent))
|
|
|
|
from gguf import GGUFWriter # noqa: E402
|
|
|
|
# Example usage:
|
|
def writer_example() -> None:
|
|
# Example usage with a file
|
|
gguf_writer = GGUFWriter("cal_test.gguf", "cal_test")
|
|
|
|
tokens = [
|
|
"a",
|
|
"b",
|
|
"c",
|
|
"d",
|
|
"e",
|
|
"f",
|
|
"g",
|
|
"h",
|
|
"i",
|
|
"j",
|
|
"k",
|
|
"l",
|
|
"m",
|
|
"n",
|
|
"o",
|
|
"p",
|
|
"q",
|
|
"r",
|
|
"s",
|
|
"t",
|
|
"u",
|
|
"v",
|
|
"w",
|
|
"x",
|
|
"y",
|
|
"z",
|
|
"\u2581", #空格
|
|
"\n"
|
|
]
|
|
|
|
vocab_size = len(tokens)
|
|
embedding_len = 256
|
|
ctx_len = 2048
|
|
|
|
gguf_writer.add_block_count(1)
|
|
gguf_writer.add_context_length(ctx_len)
|
|
gguf_writer.add_embedding_length(embedding_len)
|
|
gguf_writer.add_vocab_size(vocab_size)
|
|
|
|
gguf_writer.add_head_count_kv(8)
|
|
gguf_writer.add_head_count(16)
|
|
gguf_writer.add_key_length(16)
|
|
gguf_writer.add_value_length(16)
|
|
gguf_writer.add_tokenizer_model("llama")
|
|
# 添加 tokenizer 列表到元数据
|
|
gguf_writer.add_token_list(tokens)
|
|
output_weight = np.ones((vocab_size, embedding_len), dtype=np.float32)
|
|
|
|
print(f"token 数量：{vocab_size}")
|
|
|
|
# token_embd_weight = np.ones((128000,embedding_len), dtype=np.float32)
|
|
# print(f"创建的张量形状: {output_weight.shape}")
|
|
# gguf_writer.add_tensor("output.weight", output_weight)
|
|
|
|
# attn_norm = np.ones((1))
|
|
# gguf_writer.add_tensor("blk.0.attn_norm.weight", attn_norm)
|
|
|
|
gguf_writer.write_header_to_file()
|
|
gguf_writer.write_kv_data_to_file()
|
|
gguf_writer.write_tensors_to_file()
|
|
|
|
gguf_writer.close()
|
|
|
|
|
|
if __name__ == '__main__':
|
|
writer_example()
|