Upload examples/benchmark.py with huggingface_hub

examples/benchmark.py

@@ -0,0 +1,109 @@
+#!/usr/bin/env python3
+"""
+Sixpert K1 - Quick Benchmark Script
+====================================
+Runs basic performance benchmarks for Sixpert K1 including:
+- Token generation speed (tokens/second)
+- Context processing speed
+- Memory usage estimation
+
+Usage:
+    python benchmark.py --model SixpertK1.gguf
+"""
+
+import argparse
+import time
+import sys
+
+try:
+    from llama_cpp import Llama
+except ImportError:
+    print("Installing llama-cpp-python...")
+    import subprocess
+    subprocess.check_call([sys.executable, "-m", "pip", "install", "llama-cpp-python"])
+    from llama_cpp import Llama
+
+
+def benchmark_generation(model_path: str, tokens: int = 512):
+    """Benchmark token generation speed."""
+    print("\n=== Generation Benchmark ===")
+    print(f"Generating {tokens} tokens...\n")
+
+    llm = Llama(
+        model_path=model_path,
+        n_ctx=4096,
+        n_gpu_layers=-1,
+        verbose=False,
+    )
+
+    start = time.time()
+    output = llm(
+        "<|im_start|>user\nWrite a detailed essay about artificial intelligence.<|im_end|>\n<|im_start|>assistant\n",
+        max_tokens=tokens,
+        temperature=0.7,
+        stream=False,
+    )
+    elapsed = time.time() - start
+
+    gen_tokens = len(output["choices"][0]["text"].split())
+    tokens_per_sec = tokens / elapsed
+
+    print(f"Generated: {tokens} tokens")
+    print(f"Time: {elapsed:.2f}s")
+    print(f"Speed: {tokens_per_sec:.1f} tokens/sec")
+    print(f"Est. words: ~{gen_tokens}")
+
+
+def benchmark_context(model_path: str, context_length: int = 8192):
+    """Benchmark context processing speed."""
+    print(f"\n=== Context Processing Benchmark ===")
+    print(f"Processing {context_length} token context...\n")
+
+    llm = Llama(
+        model_path=model_path,
+        n_ctx=context_length + 512,
+        n_gpu_layers=-1,
+        verbose=False,
+    )
+
+    # Create a long context prompt
+    filler = "The quick brown fox jumps over the lazy dog. " * (context_length // 10)
+    prompt = f"<|im_start|>user\n{filler}\nSummarize the above text in one sentence.<|im_end|>\n<|im_start|>assistant\n"
+
+    start = time.time()
+    output = llm(prompt, max_tokens=100, stream=False)
+    elapsed = time.time() - start
+
+    prompt_tokens = output["usage"]["prompt_eval_count"]
+    eval_time = output["usage"].get("prompt_eval_time", 1000) / 1000
+
+    print(f"Context tokens: {prompt_tokens}")
+    print(f"Processing time: {eval_time:.2f}s")
+    print(f"Speed: {prompt_tokens / eval_time:.1f} tokens/sec")
+
+
+def main():
+    parser = argparse.ArgumentParser(description="Sixpert K1 Benchmark")
+    parser.add_argument("--model", type=str, default="SixpertK1.gguf", help="Path to GGUF model")
+    parser.add_argument("--gen-tokens", type=int, default=512, help="Generation benchmark tokens")
+    parser.add_argument("--ctx-length", type=int, default=8192, help="Context benchmark length")
+    parser.add_argument("--all", action="store_true", help="Run all benchmarks")
+
+    args = parser.parse_args()
+
+    print("=" * 60)
+    print("  Sixpert K1 Benchmark Suite")
+    print("  Precision Logic Engine")
+    print("=" * 60)
+
+    if args.all or True:
+        benchmark_generation(args.model, args.gen_tokens)
+        benchmark_context(args.model, args.ctx_length)
+
+    print("\n" + "=" * 60)
+    print("  Benchmark complete!")
+    print("=" * 60)
+
+
+if __name__ == "__main__":
+    main()