From 2b1d618287a9d850b73b6ada27493a22dfd52fa6 Mon Sep 17 00:00:00 2001 From: Marut Pandya Date: Tue, 18 Feb 2025 10:45:47 -0800 Subject: [PATCH] Revert "Enabling model caching." --- src/engine.py | 3 --- 1 file changed, 3 deletions(-) diff --git a/src/engine.py b/src/engine.py index ce82016..88a0101 100644 --- a/src/engine.py +++ b/src/engine.py @@ -24,10 +24,7 @@ class vLLMEngine: def __init__(self, engine = None): load_dotenv() # For local development self.engine_args = get_engine_args() - if os.getenv("MODEL_CACHE_ENABLE"): - self.engine_args.model = f"/runpod/cache/model/{os.getenv('MODEL_NAME')}/main" logging.info(f"Engine args: {self.engine_args}") - self.tokenizer = TokenizerWrapper(self.engine_args.tokenizer or self.engine_args.model, self.engine_args.tokenizer_revision, self.engine_args.trust_remote_code)