Ollama: 模型管理

# Pull different sizes for different tasks
ollama pull llama3.2:3b    # Fast, for simple queries
ollama pull llama3.1:8b    # Balanced
ollama pull llama3.1:70b   # Powerful, for complex cases

# List all local models
ollama list
# NAME              ID              SIZE      MODIFIED
# llama3.2:latest   a80c4feeb02f    2.0 GB    2 hours ago
# mistral:latest    2b3e5c5f1e3d    4.1 GB    1 day ago

# Pull a specific model
ollama pull llama3.2
# pulling manifest... done
# success

# Show model details
ollama show llama3.2
# Model
#   architecture: llama
#   parameters:    3.2B
#   quantization:  Q4_K_M
#   context length: 131072

# Delete a model to free space
ollama rm llama3.2:3b
# deleted 'llama3.2:3b'

NAME                    ID              SIZE    
llama3.2:latest        a80...          2.0 GB  
mistral:latest         61...           4.1 GB

# Default: auto-select quantization (usually Q4_K_M)
ollama pull llama3.2

# Explicit: pull specific quantization
ollama pull llama3.2:q4_K_M     # Balanced (recommended)
ollama pull llama3.2:q8_0       # Higher quality, larger file

# Pull by parameter size
ollama pull llama3.1:8b         # 8 billion parameters
ollama pull llama3.1:70b        # 70 billion parameters (needs 40GB+ VRAM)

# List all available variants
ollama show llama3.2 --modelfile

pulling manifest... 
success

# Alice's SupportBot: multi-language customer service
ollama pull qwen2.5:7b          # Best Chinese + multilingual
ollama pull nomic-embed-text    # For RAG knowledge base

# Alice's GlobalShop: SQL generation
ollama pull codellama:7b        # Code-focused model

# Alice's SaaS: metrics analysis
ollama pull llama3.1:8b         # General reasoning

# Check total disk usage
ollama list | awk '{sum+=$3} END {print "Total: " sum/1024/1024/1024 " GB"}'

NAME                    ID              SIZE    
llama3.2:latest        a80...          2.0 GB  
mistral:latest         61...           4.1 GB

# Check disk usage per model
ollama list

# Remove unused models to free space
ollama rm llama3.1:70b    # Free ~40GB
ollama rm mistral:latest  # Free ~4GB

# Change model storage path (if disk full)
export OLLAMA_MODELS=/data/ollama/models
ollama serve

# Verify new path
ls $OLLAMA_MODELS/blobs/

NAME                    ID              SIZE    
llama3.2:latest        a80...          2.0 GB  
mistral:latest         61...           4.1 GB

# Download a GGUF file from HuggingFace
wget https://huggingface.co/.../model-q4_K_M.gguf

# Create a Modelfile pointing to the GGUF
cat > Modelfile <<EOF
FROM ./model-q4_K_M.gguf
EOF

# Build and register with Ollama
ollama create my-custom-model -f Modelfile

# Run it like any other model
ollama run my-custom-model "Hello"

I'm a helpful AI assistant running locally on your machine...

#!/bin/bash
# ============================================
# Comprehensive: Model management dashboard
# Pull, inspect, clean, and report models
# ============================================

MODEL_STORE="$HOME/.ollama/models"

# Function: Show model inventory report
show_inventory() {
    echo "=== Ollama Model Inventory ==="
    echo ""
    ollama list | while read -r name id size modified; do
        if [ "$name" != "NAME" ]; then
            echo "Model: $name"
            echo "  ID:       $id"
            echo "  Size:     $size"
            echo "  Modified: $modified"
            echo ""
        fi
    done
    echo "Total disk: $(du -sh $MODEL_STORE 2>/dev/null | cut -f1)"
}

# Function: Pull models for SupportBot stack
pull_support_stack() {
    echo "=== Pulling SupportBot Model Stack ==="
    local models=(
        "qwen2.5:7b"           # Main chat model
        "nomic-embed-text"     # Embedding for RAG
        "llama3.2:3b"          # Fast classifier
    )
    for model in "${models[@]}"; do
        echo "Pulling $model..."
        ollama pull "$model"
    done
    echo "=== Stack Ready ==="
}

# Function: Clean old/unused models
clean_unused() {
    echo "=== Models Available for Cleanup ==="
    ollama list
    echo ""
    read -p "Enter model name to remove (or 'cancel'): " model
    if [ "$model" != "cancel" ] && [ -n "$model" ]; then
        ollama rm "$model"
        echo "Removed: $model"
    fi
}

# Main menu
echo "1. Show inventory"
echo "2. Pull SupportBot stack"
echo "3. Clean unused models"
read -p "Choose: " choice
case $choice in
    1) show_inventory ;;
    2) pull_support_stack ;;
    3) clean_unused ;;
esac