mirror of
https://github.com/zenlm/zen-eco.git
synced 2026-10-30 14:58:11 +00:00
zeekay
bcc96f6f48
- Apache 2.0 License - Zen Research DAO under Zoo Labs Inc (501(c)(3)) - San Francisco-based non-profit for open AI research - Copyright 2025 Zen Research Authors
72 lines
2.1 KiB
Python
Executable File
72 lines
2.1 KiB
Python
Executable File
#!/usr/bin/env python3.13
|
|
"""
|
|
Convert models to GGUF format
|
|
Simple wrapper around llama.cpp conversion
|
|
"""
|
|
import argparse
|
|
import os
|
|
import subprocess
|
|
from pathlib import Path
|
|
|
|
def main():
|
|
parser = argparse.ArgumentParser()
|
|
parser.add_argument("--model_path", type=str, required=True)
|
|
parser.add_argument("--output_path", type=str, required=True)
|
|
parser.add_argument("--quantization", type=str, default="q4_k_m")
|
|
args = parser.parse_args()
|
|
|
|
print(f"📦 Converting to GGUF: {args.model_path}")
|
|
print(f"📊 Quantization: {args.quantization}")
|
|
|
|
# Check if llama.cpp is available
|
|
llama_cpp_path = "/Users/z/work/llama.cpp"
|
|
if not os.path.exists(llama_cpp_path):
|
|
print("⚠️ llama.cpp not found, downloading...")
|
|
subprocess.run([
|
|
"git", "clone",
|
|
"https://github.com/ggerganov/llama.cpp",
|
|
llama_cpp_path
|
|
], check=True)
|
|
|
|
# Build llama.cpp
|
|
print("Building llama.cpp...")
|
|
subprocess.run(["make", "-C", llama_cpp_path], check=True)
|
|
|
|
# Convert to GGUF
|
|
convert_script = f"{llama_cpp_path}/convert-hf-to-gguf.py"
|
|
|
|
# First convert to FP16 GGUF
|
|
temp_gguf = args.output_path.replace(".gguf", "_fp16.gguf")
|
|
|
|
print("Converting to GGUF FP16...")
|
|
subprocess.run([
|
|
"python3.13", convert_script,
|
|
args.model_path,
|
|
"--outfile", temp_gguf,
|
|
"--outtype", "f16"
|
|
], check=True)
|
|
|
|
# Then quantize
|
|
if args.quantization != "f16":
|
|
print(f"Quantizing to {args.quantization}...")
|
|
quantize_bin = f"{llama_cpp_path}/llama-quantize"
|
|
|
|
subprocess.run([
|
|
quantize_bin,
|
|
temp_gguf,
|
|
args.output_path,
|
|
args.quantization
|
|
], check=True)
|
|
|
|
# Remove temp file
|
|
os.remove(temp_gguf)
|
|
else:
|
|
os.rename(temp_gguf, args.output_path)
|
|
|
|
# Verify output
|
|
output_size = os.path.getsize(args.output_path) / (1024 * 1024 * 1024)
|
|
print(f"✅ GGUF created: {args.output_path}")
|
|
print(f"📏 Size: {output_size:.2f} GB")
|
|
|
|
if __name__ == "__main__":
|
|
main() |