ollama-bench 1.0.0

package/LICENSE

@@ -0,0 +1,21 @@
+MIT License
+
+Copyright (c) 2024 dalist.
+
+Permission is hereby granted, free of charge, to any person obtaining a copy
+of this software and associated documentation files (the "Software"), to deal
+in the Software without restriction, including without limitation the rights
+to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
+copies of the Software, and to permit persons to whom the Software is
+furnished to do so, subject to the following conditions:
+
+The above copyright notice and this permission notice shall be included in all
+copies or substantial portions of the Software.
+
+THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
+AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
+OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+SOFTWARE.

package/README.md

@@ -0,0 +1,44 @@
+# ollama-benchmark
+
+A command-line tool to benchmark Ollama models performance.
+
+## Prerequisites
+
+- Node.js >= 14.0.0
+- Ollama installed and running on your system
+
+## Installation
+
+You can install globally:
+```bash
+npm install -g ollama-bench
+```
+
+Or run directly with npx:
+```bash
+npx ollama-bench <model1> [model2] [model3] ...
+```
+
+## Usage
+
+```bash
+ollama-bench mistral llama2 codellama
+```
+
+This will:
+1. Pull the specified models if not already present
+2. Run a benchmark test on each model
+3. Compare and display the results
+4. Show the best performing model
+
+## Output
+
+The tool provides real-time feedback with:
+- Loading animations during operations
+- Color-coded status messages
+- Detailed benchmark results
+- Comparison of model performance
+
+## License
+
+MIT

package/dist/index.js

@@ -0,0 +1,143 @@
+#!/usr/bin/env node
+import ollama from 'ollama';
+/**
+ * Object containing ANSI color codes for text coloring.
+ */
+const colors = {
+    reset: '\x1b[0m',
+    green: '\x1b[32m',
+    yellow: '\x1b[33m',
+    red: '\x1b[31m',
+    cyan: '\x1b[36m',
+    magenta: '\x1b[35m',
+    blue: '\x1b[34m',
+};
+/**
+ * Object containing emoji characters for various status indicators.
+ */
+const emojis = {
+    rocket: '🚀',
+    check: '✅',
+    error: '❌',
+    hourglass: '⏳',
+    star: '⭐',
+    trophy: '🏆',
+    gear: '⚙️',
+};
+/**
+ * Applies color to the given text.
+ * @param text - The text to colorize.
+ * @param color - The color to apply.
+ * @returns The colorized text.
+ */
+function colorize(text, color) {
+    return `${colors[color]}${text}${colors.reset}`;
+}
+/**
+ * Creates a loading animation for the console.
+ * @param operation - The operation being performed.
+ * @param model - The model being processed.
+ * @returns An interval ID for the animation.
+ */
+function createLoadingAnimation(operation, model) {
+    const frames = ['⠋', '⠙', '⠹', '⠸', '⠼', '⠴', '⠦', '⠧', '⠇', '⠏'];
+    let i = 0;
+    let dots = 0;
+    return setInterval(() => {
+        const frame = frames[i];
+        const dotString = '.'.repeat(dots);
+        const operationText = colorize(`${operation} ${model}${dotString}`, 'blue');
+        process.stdout.write(`\r${frame} ${emojis.gear} ${operationText}`.padEnd(50));
+        i = (i + 1) % frames.length;
+        dots = (dots + 1) % 4;
+    }, 100);
+}
+/**
+ * Pulls a model from Ollama.
+ * @param model - The name of the model to pull.
+ */
+async function pullModel(model) {
+    console.log(colorize(`${emojis.rocket} Initiating pull for ${model}...`, 'yellow'));
+    const loadingAnimation = createLoadingAnimation('Pulling', model);
+    try {
+        const start = performance.now();
+        const response = await ollama.pull({ model, stream: true });
+        for await (const part of response) {
+            if (part.status === 'success') {
+                clearInterval(loadingAnimation);
+                const end = performance.now();
+                const duration = (end - start) / 1000;
+                console.log(`\r${colorize(`${emojis.check} Successfully pulled ${model} in ${duration.toFixed(2)} seconds`, 'green')}     `);
+                return;
+            }
+        }
+    }
+    catch (error) {
+        clearInterval(loadingAnimation);
+        console.log(`\r${colorize(`${emojis.error} Error pulling ${model}: ${error.message}`, 'red')}     `);
+    }
+}
+/**
+ * Benchmarks a model's performance.
+ * @param model - The name of the model to benchmark.
+ * @returns A promise that resolves to the benchmark result.
+ */
+async function benchmarkModel(model) {
+    const prompt = "Explain the theory of relativity in simple terms.";
+    console.log(colorize(`${emojis.hourglass} Initiating benchmark for ${model}...`, 'cyan'));
+    const loadingAnimation = createLoadingAnimation('Benchmarking', model);
+    try {
+        const response = await ollama.generate({
+            model,
+            prompt,
+            stream: false,
+        });
+        clearInterval(loadingAnimation);
+        const totalDuration = response.total_duration / 1e9; // Convert nanoseconds to seconds
+        const tokensPerSecond = response.eval_count / (response.eval_duration / 1e9);
+        console.log(`\r${colorize(`${emojis.star} Benchmark results for ${model}:`, 'cyan')}     `);
+        console.log(colorize(`  Total time: ${totalDuration.toFixed(2)} seconds`, 'yellow'));
+        console.log(colorize(`  Tokens generated: ${response.eval_count}`, 'yellow'));
+        console.log(colorize(`  Tokens per second: ${tokensPerSecond.toFixed(2)}`, 'yellow'));
+        console.log();
+        return { model, tokensPerSecond };
+    }
+    catch (error) {
+        clearInterval(loadingAnimation);
+        console.log(`\r${colorize(`${emojis.error} Error benchmarking ${model}: ${error.message}`, 'red')}     `);
+        return { model, tokensPerSecond: 0 };
+    }
+}
+/**
+ * The main function that orchestrates the model pulling and benchmarking process.
+ */
+export async function main() {
+    const models = process.argv.slice(2);
+    if (models.length === 0) {
+        console.log(colorize(`${emojis.error} Error: No models provided. Please specify at least one model.`, 'red'));
+        process.exit(1);
+    }
+    console.log(colorize(`${emojis.rocket} Ollama Benchmark Script`, 'cyan'));
+    console.log(colorize("=======================", 'cyan'));
+    // Pull models
+    for (const model of models) {
+        await pullModel(model);
+    }
+    console.log();
+    // Benchmark models
+    const results = [];
+    for (const model of models) {
+        const result = await benchmarkModel(model);
+        results.push(result);
+    }
+    // Find the best performing model
+    const bestModel = results.reduce((best, current) => current.tokensPerSecond > best.tokensPerSecond ? current : best);
+    console.log(colorize(`${emojis.trophy} Best performing model:`, 'magenta'));
+    console.log(colorize(`  ${bestModel.model} with ${bestModel.tokensPerSecond.toFixed(2)} tokens/second`, 'magenta'));
+}
+if (import.meta.url === import.meta.resolve(process.argv[1])) {
+    main().catch(error => {
+        console.error('Error:', error);
+        process.exit(1);
+    });
+}

package/package.json

@@ -0,0 +1,31 @@
+{
+    "name": "ollama-bench",
+    "version": "1.0.0",
+    "description": "A CLI tool to benchmark Ollama models performance",
+    "main": "dist/index.js",
+    "type": "module",
+    "bin": {
+        "ollama-benchmark": "./dist/index.js"
+    },
+    "scripts": {
+        "build": "tsc",
+        "start": "node dist/index.js",
+        "dev": "tsc && node dist/index.js"
+    },
+    "keywords": ["ollama", "benchmark", "ai", "models", "cli"],
+    "author": "dalist1",
+    "license": "MIT",
+    "dependencies": {
+        "ollama": "latest"
+    },
+    "devDependencies": {
+        "@types/node": "^20.17.5",
+        "typescript": "^5.6.3"
+    },
+    "engines": {
+        "node": ">=14.0.0"
+    },
+    "publishConfig": {
+        "access": "public"
+    }
+}

package/src/index.ts

@@ -0,0 +1,181 @@
+#!/usr/bin/env node
+
+import ollama from 'ollama';
+
+/**
+ * Represents the available color codes for text coloring.
+ */
+type Color = 'reset' | 'green' | 'yellow' | 'red' | 'cyan' | 'magenta' | 'blue';
+
+/**
+ * Represents the available emoji keys.
+ */
+type Emoji = 'rocket' | 'check' | 'error' | 'hourglass' | 'star' | 'trophy' | 'gear';
+
+/**
+ * Object containing ANSI color codes for text coloring.
+ */
+const colors: Record<Color, string> = {
+  reset: '\x1b[0m',
+  green: '\x1b[32m',
+  yellow: '\x1b[33m',
+  red: '\x1b[31m',
+  cyan: '\x1b[36m',
+  magenta: '\x1b[35m',
+  blue: '\x1b[34m',
+};
+
+/**
+ * Object containing emoji characters for various status indicators.
+ */
+const emojis: Record<Emoji, string> = {
+  rocket: '🚀',
+  check: '✅',
+  error: '❌',
+  hourglass: '⏳',
+  star: '⭐',
+  trophy: '🏆',
+  gear: '⚙️',
+};
+
+/**
+ * Applies color to the given text.
+ * @param text - The text to colorize.
+ * @param color - The color to apply.
+ * @returns The colorized text.
+ */
+function colorize(text: string, color: Color): string {
+  return `${colors[color]}${text}${colors.reset}`;
+}
+
+/**
+ * Creates a loading animation for the console.
+ * @param operation - The operation being performed.
+ * @param model - The model being processed.
+ * @returns An interval ID for the animation.
+ */
+function createLoadingAnimation(operation: string, model: string): NodeJS.Timeout {
+  const frames: string[] = ['⠋', '⠙', '⠹', '⠸', '⠼', '⠴', '⠦', '⠧', '⠇', '⠏'];
+  let i = 0;
+  let dots = 0;
+  return setInterval(() => {
+    const frame = frames[i];
+    const dotString = '.'.repeat(dots);
+    const operationText = colorize(`${operation} ${model}${dotString}`, 'blue');
+    process.stdout.write(`\r${frame} ${emojis.gear} ${operationText}`.padEnd(50));
+    i = (i + 1) % frames.length;
+    dots = (dots + 1) % 4;
+  }, 100);
+}
+
+/**
+ * Pulls a model from Ollama.
+ * @param model - The name of the model to pull.
+ */
+async function pullModel(model: string): Promise<void> {
+  console.log(colorize(`${emojis.rocket} Initiating pull for ${model}...`, 'yellow'));
+  const loadingAnimation = createLoadingAnimation('Pulling', model);
+  try {
+    const start = performance.now();
+    const response = await ollama.pull({ model, stream: true });
+    for await (const part of response) {
+      if (part.status === 'success') {
+        clearInterval(loadingAnimation);
+        const end = performance.now();
+        const duration = (end - start) / 1000;
+        console.log(`\r${colorize(`${emojis.check} Successfully pulled ${model} in ${duration.toFixed(2)} seconds`, 'green')}     `);
+        return;
+      }
+    }
+  } catch (error) {
+    clearInterval(loadingAnimation);
+    console.log(`\r${colorize(`${emojis.error} Error pulling ${model}: ${(error as Error).message}`, 'red')}     `);
+  }
+}
+
+/**
+ * Represents the result of a model benchmark.
+ */
+interface BenchmarkResult {
+  model: string;
+  tokensPerSecond: number;
+}
+
+/**
+ * Benchmarks a model's performance.
+ * @param model - The name of the model to benchmark.
+ * @returns A promise that resolves to the benchmark result.
+ */
+async function benchmarkModel(model: string): Promise<BenchmarkResult> {
+  const prompt = "Explain the theory of relativity in simple terms.";
+  console.log(colorize(`${emojis.hourglass} Initiating benchmark for ${model}...`, 'cyan'));
+  const loadingAnimation = createLoadingAnimation('Benchmarking', model);
+  
+  try {
+    const response = await ollama.generate({
+      model,
+      prompt,
+      stream: false,
+    });
+
+    clearInterval(loadingAnimation);
+    const totalDuration = response.total_duration / 1e9; // Convert nanoseconds to seconds
+    const tokensPerSecond = response.eval_count / (response.eval_duration / 1e9);
+
+    console.log(`\r${colorize(`${emojis.star} Benchmark results for ${model}:`, 'cyan')}     `);
+    console.log(colorize(`  Total time: ${totalDuration.toFixed(2)} seconds`, 'yellow'));
+    console.log(colorize(`  Tokens generated: ${response.eval_count}`, 'yellow'));
+    console.log(colorize(`  Tokens per second: ${tokensPerSecond.toFixed(2)}`, 'yellow'));
+    console.log();
+
+    return { model, tokensPerSecond };
+  } catch (error) {
+    clearInterval(loadingAnimation);
+    console.log(`\r${colorize(`${emojis.error} Error benchmarking ${model}: ${(error as Error).message}`, 'red')}     `);
+    return { model, tokensPerSecond: 0 };
+  }
+}
+
+/**
+ * The main function that orchestrates the model pulling and benchmarking process.
+ */
+export async function main(): Promise<void> {
+  const models = process.argv.slice(2);
+
+  if (models.length === 0) {
+    console.log(colorize(`${emojis.error} Error: No models provided. Please specify at least one model.`, 'red'));
+    process.exit(1);
+  }
+
+  console.log(colorize(`${emojis.rocket} Ollama Benchmark Script`, 'cyan'));
+  console.log(colorize("=======================", 'cyan'));
+
+  // Pull models
+  for (const model of models) {
+    await pullModel(model);
+  }
+
+  console.log();
+
+  // Benchmark models
+  const results: BenchmarkResult[] = [];
+  for (const model of models) {
+    const result = await benchmarkModel(model);
+    results.push(result);
+  }
+
+  // Find the best performing model
+  const bestModel = results.reduce((best, current) => 
+    current.tokensPerSecond > best.tokensPerSecond ? current : best
+  );
+
+  console.log(colorize(`${emojis.trophy} Best performing model:`, 'magenta'));
+  console.log(colorize(`  ${bestModel.model} with ${bestModel.tokensPerSecond.toFixed(2)} tokens/second`, 'magenta'));
+}
+
+if (import.meta.url === import.meta.resolve(process.argv[1])) {
+  main().catch(error => {
+      console.error('Error:', error);
+      process.exit(1);
+  });
+}

package/tsconfig.json

@@ -0,0 +1,15 @@
+{
+  "compilerOptions": {
+    "target": "ES2020",
+    "module": "ES2020",
+    "moduleResolution": "node",
+    "outDir": "./dist",
+    "rootDir": "./src",
+    "strict": true,
+    "esModuleInterop": true,
+    "skipLibCheck": true,
+    "forceConsistentCasingInFileNames": true
+  },
+  "include": ["src/**/*"],
+  "exclude": ["node_modules", "dist"]
+}