From 3285d9118f0874a54fd1922c6e650db4937e52f0 Mon Sep 17 00:00:00 2001 From: Mikhail Yevchenko Date: Thu, 21 May 2026 19:33:41 +0000 Subject: [PATCH] Enhance completions benchmark generator to extract words from a fallback Perl copyright file --- workers/openai/worker.py | 16 +++++++++++----- 1 file changed, 11 insertions(+), 5 deletions(-) diff --git a/workers/openai/worker.py b/workers/openai/worker.py index 00aa336..22c39f9 100644 --- a/workers/openai/worker.py +++ b/workers/openai/worker.py @@ -1,5 +1,6 @@ import random import os +import re from vastai import Worker, WorkerConfig, HandlerConfig, LogActionConfig, BenchmarkConfig @@ -22,11 +23,16 @@ def completions_benchmark_generator() -> dict: WORD_LIST = [] - with open(__file__, 'r') as f: - # Use regex to extract words from the source code - import re - source_code = f.read() - WORD_LIST = re.findall(r'\b\w+\b', source_code) + # Try to load from perl copyright file first + try: + with open("/usr/share/doc/perl/copyright", 'r') as f: + source_code = f.read() + WORD_LIST = re.findall(r'\b\w+\b', source_code) + except (FileNotFoundError, IOError): + # Fallback to loading from python file + with open(__file__, 'r') as f: + source_code = f.read() + WORD_LIST = re.findall(r'\b\w+\b', source_code) prompt = " ".join(random.choices(WORD_LIST, k=int(250))) model = os.environ.get("MODEL_NAME")