Spaces:
Running
Running
File size: 6,602 Bytes
e7e5eda |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 |
#!/bin/bash
set -e
echo "π COMPREHENSIVE PHI-3.5-MINI CONVERSION & QUANTIZATION SYSTEM"
echo "================================================================"
# Configuration
MODEL_NAME="microsoft/Phi-3.5-mini-instruct"
OUTPUT_DIR="./phi3.5-mini-gguf-complete"
QUANT_TYPES=("Q2_K" "Q3_K_S" "Q3_K_M" "Q3_K_L" "Q4_0" "Q4_1" "Q4_K_S" "Q4_K_M" "Q5_0" "Q5_1" "Q5_K_S" "Q5_K_M" "Q6_K" "Q8_0" "F16" "F32")
# Create directories
mkdir -p $OUTPUT_DIR
mkdir -p $OUTPUT_DIR/artifacts
mkdir -p $OUTPUT_DIR/logs
# Install system dependencies
echo "π¦ INSTALLING SYSTEM DEPENDENCIES"
sudo apt-get update
sudo apt-get install -y \
build-essential \
cmake \
git \
python3 \
python3-pip \
python3-venv \
libcurl4-openssl-dev \
git-lfs \
wget \
curl \
pkg-config \
libopenblas-dev \
libatomic-dev
# Initialize Git LFS
git lfs install
# Create Python virtual environment
python3 -m venv phi3-convert
source phi3-convert/bin/activate
# Install Python dependencies
pip install --upgrade pip
pip install torch torchvision torchaudio --index-url https://download.pytorch.org/whl/cpu
pip install transformers>=4.43.0
pip install safetensors
pip install huggingface-hub
pip install accelerate
pip install protobuf
pip install sentencepiece
pip install ninja
pip install requests
pip install tqdm
pip install psutil
pip install GPUtil
# Clone and build llama.cpp with all optimizations
echo "π¨ BUILDING LLAMA.CPP WITH FULL OPTIMIZATIONS"
if [ ! -d "llama.cpp" ]; then
git clone https://github.com/ggerganov/llama.cpp
fi
cd llama.cpp
# Clean build directory
rm -rf build
mkdir -p build
cd build
# Configure with maximum optimizations
cmake .. \
-DCMAKE_BUILD_TYPE=Release \
-DLLAMA_NATIVE=ON \
-DLLAMA_CURL=ON \
-DLLAMA_OPENBLAS=ON \
-DLLAMA_BLAS=ON \
-DBLAS_LIBRARIES=/usr/lib/openblas-base/libopenblas.a \
-DLLAMA_CUDA=OFF \
-DLLAMA_METAL=OFF \
-DLLAMA_MPI=OFF \
-DLLAMA_CCACHE=ON \
-DLLAMA_BUILD_TESTS=ON \
-DLLAMA_BUILD_EXAMPLES=ON
# Build with maximum parallelism
make -j$(nproc)
# Return to main directory
cd ../..
# Download the model
echo "π₯ DOWNLOADING PHI-3.5-MINI-INSTRUCT MODEL"
if [ ! -d "Phi-3.5-mini-instruct" ]; then
git clone https://huggingface.co/microsoft/Phi-3.5-mini-instruct
fi
# Verify model files
echo "π VERIFYING MODEL FILES"
cd Phi-3.5-mini-instruct
ls -la
cd ..
# Create comprehensive conversion script
cat > complete_conversion.py << 'EOF'
import os
import sys
import subprocess
import json
import time
from datetime import datetime
class Phi3Converter:
def __init__(self, model_path, output_dir):
self.model_path = model_path
self.output_dir = output_dir
self.quant_types = ["Q2_K", "Q3_K_S", "Q3_K_M", "Q3_K_L", "Q4_0", "Q4_1",
"Q4_K_S", "Q4_K_M", "Q5_0", "Q5_1", "Q5_K_S", "Q5_K_M",
"Q6_K", "Q8_0", "F16", "F32"]
def run_command(self, cmd, description):
print(f"π {description}")
print(f" Command: {cmd}")
start_time = time.time()
result = subprocess.run(cmd, shell=True, capture_output=True, text=True)
end_time = time.time()
if result.returncode != 0:
print(f"β Error: {result.stderr}")
return False
print(f"β
Completed in {end_time - start_time:.2f}s")
return True
def convert_to_gguf(self, outtype="f16"):
output_file = f"{self.output_dir}/phi3.5-mini.{outtype}.gguf"
cmd = f"python3 llama.cpp/convert-hf-to-gguf.py {self.model_path} --outfile {output_file} --outtype {outtype}"
return self.run_command(cmd, f"Converting to GGUF ({outtype})")
def quantize_model(self, input_file, quant_type):
output_file = f"{self.output_dir}/phi3.5-mini.{quant_type}.gguf"
cmd = f"./llama.cpp/build/bin/quantize {input_file} {output_file} {quant_type}"
return self.run_command(cmd, f"Quantizing to {quant_type}")
def generate_model_card(self):
model_card = {
"model_name": "Phi-3.5-mini-instruct-GGUF-Comprehensive",
"base_model": "microsoft/Phi-3.5-mini-instruct",
"description": "Comprehensive GGUF quantization of Phi-3.5-mini-instruct for Saem's Tunes",
"quantizations": self.quant_types,
"conversion_date": datetime.now().isoformat(),
"file_sizes": {},
"recommended_use_cases": {
"Q2_K": "Mobile devices, extremely low memory",
"Q3_K_S": "Mobile devices, low memory",
"Q4_0": "Good balance for most use cases",
"Q4_K_M": "Recommended balance (quality/speed)",
"Q5_K_M": "High quality with reasonable size",
"Q8_0": "Maximum quality, larger size",
"F16": "Original quality, large size"
}
}
with open(f"{self.output_dir}/model_card.json", "w") as f:
json.dump(model_card, f, indent=2)
def run_comprehensive_conversion(self):
print("π STARTING COMPREHENSIVE CONVERSION PIPELINE")
# Convert to F16 first
if not self.convert_to_gguf("f16"):
return False
base_model = f"{self.output_dir}/phi3.5-mini.f16.gguf"
# Create all quantization variants
for quant_type in self.quant_types:
if quant_type not in ["F16", "F32"]:
self.quantize_model(base_model, quant_type)
# Convert to F32 separately
self.convert_to_gguf("f32")
# Generate model card
self.generate_model_card()
print("π COMPREHENSIVE CONVERSION COMPLETED!")
return True
if __name__ == "__main__":
converter = Phi3Converter("Phi-3.5-mini-instruct", "phi3.5-mini-gguf-complete")
converter.run_comprehensive_conversion()
EOF
# Run the comprehensive conversion
echo "π STARTING COMPREHENSIVE CONVERSION PROCESS"
python3 complete_conversion.py
# Generate file manifest
echo "π GENERATING FILE MANIFEST"
ls -la phi3.5-mini-gguf-complete/ > phi3.5-mini-gguf-complete/file_manifest.txt
# Create deployment package
echo "π¦ CREATING DEPLOYMENT PACKAGE"
tar -czvf phi3.5-mini-complete-gguf-$(date +%Y%m%d).tar.gz phi3.5-mini-gguf-complete/
echo "π COMPLETE CONVERSION SYSTEM FINISHED!"
echo "π Output directory: phi3.5-mini-gguf-complete/"
echo "π Total files generated: $(ls -1 phi3.5-mini-gguf-complete/*.gguf | wc -l)" |