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
96 lines
3.0 KiB
Python
96 lines
3.0 KiB
Python
#!/usr/bin/env python3.13
|
|
"""
|
|
Generate GGUF files from our trained Zen Eco models
|
|
"""
|
|
import os
|
|
import subprocess
|
|
from pathlib import Path
|
|
|
|
def convert_to_gguf(model_path, output_name):
|
|
"""Convert a model to GGUF format"""
|
|
print(f"🔄 Converting {output_name} to GGUF...")
|
|
|
|
# Ensure llama.cpp is available
|
|
llama_cpp = Path("/Users/z/work/zen/llama.cpp")
|
|
if not llama_cpp.exists():
|
|
print("⚠️ llama.cpp not found, cloning...")
|
|
subprocess.run([
|
|
"git", "clone",
|
|
"https://github.com/ggerganov/llama.cpp.git",
|
|
str(llama_cpp)
|
|
], check=True)
|
|
|
|
# Build llama.cpp
|
|
print("Building llama.cpp...")
|
|
subprocess.run(["cmake", "-B", "build"], cwd=llama_cpp, check=True)
|
|
subprocess.run(["cmake", "--build", "build", "--config", "Release"], cwd=llama_cpp, check=True)
|
|
|
|
# Create GGUF directory
|
|
gguf_dir = Path(f"/Users/z/work/zen/zen-eco/gguf")
|
|
gguf_dir.mkdir(exist_ok=True)
|
|
|
|
# Convert to GGUF F16
|
|
convert_script = llama_cpp / "convert_hf_to_gguf.py"
|
|
output_file = gguf_dir / f"{output_name}-f16.gguf"
|
|
|
|
print(f"Converting to F16 GGUF: {output_file}")
|
|
subprocess.run([
|
|
"python3.13", str(convert_script),
|
|
str(model_path),
|
|
"--outfile", str(output_file),
|
|
"--outtype", "f16"
|
|
], check=True)
|
|
|
|
# Quantize to Q4_K_M
|
|
quantize_exe = llama_cpp / "build" / "bin" / "llama-quantize"
|
|
q4_file = gguf_dir / f"{output_name}-Q4_K_M.gguf"
|
|
|
|
print(f"Creating Q4_K_M quantization...")
|
|
subprocess.run([
|
|
str(quantize_exe),
|
|
str(output_file),
|
|
str(q4_file),
|
|
"Q4_K_M"
|
|
], check=True)
|
|
|
|
# Quantize to Q5_K_M
|
|
q5_file = gguf_dir / f"{output_name}-Q5_K_M.gguf"
|
|
print(f"Creating Q5_K_M quantization...")
|
|
subprocess.run([
|
|
str(quantize_exe),
|
|
str(output_file),
|
|
str(q5_file),
|
|
"Q5_K_M"
|
|
], check=True)
|
|
|
|
print(f"✅ GGUF files created for {output_name}")
|
|
return True
|
|
|
|
def main():
|
|
print("🎯 Generating GGUF for all trained Zen Eco models")
|
|
print("=" * 50)
|
|
|
|
models = [
|
|
("/Users/z/work/zen/zen-eco/instruct/finetuned", "zen-eco-4b-instruct"),
|
|
("/Users/z/work/zen/zen-eco/thinking/finetuned", "zen-eco-4b-thinking"),
|
|
("/Users/z/work/zen/zen-eco/output/zen-eco-4b-agent", "zen-eco-4b-agent"),
|
|
]
|
|
|
|
for model_path, output_name in models:
|
|
if Path(model_path).exists():
|
|
try:
|
|
convert_to_gguf(model_path, output_name)
|
|
except Exception as e:
|
|
print(f"❌ Failed to convert {output_name}: {e}")
|
|
else:
|
|
print(f"⚠️ Model not found: {model_path}")
|
|
|
|
print("\n✅ GGUF generation complete!")
|
|
print("\n📦 Generated files:")
|
|
gguf_dir = Path("/Users/z/work/zen/zen-eco/gguf")
|
|
for gguf in gguf_dir.glob("*.gguf"):
|
|
size_gb = gguf.stat().st_size / (1024**3)
|
|
print(f" • {gguf.name}: {size_gb:.2f} GB")
|
|
|
|
if __name__ == "__main__":
|
|
main() |