# Reproducibility vLLM does not guarantee the reproducibility of the results by default, for the sake of performance. You need to do the following to achieve reproducible results: - For V1: Turn off multiprocessing to make the scheduling deterministic by setting `VLLM_ENABLE_V1_MULTIPROCESSING=0`. - For V0: Set the global seed (see below). Example: !!! warning Applying the above settings [changes the random state in user code](#locality-of-random-state). !!! note Even with the above settings, vLLM only provides reproducibility when it runs on the same hardware and the same vLLM version. Also, the online serving API (`vllm serve`) does not support reproducibility because it is almost impossible to make the scheduling deterministic in the online setting. ## Setting the global seed The `seed` parameter in vLLM is used to control the random states for various random number generators. If a specific seed value is provided, the random states for `random`, `np.random`, and `torch.manual_seed` will be set accordingly. However, in some cases, setting the seed will also [change the random state in user code](#locality-of-random-state). ### Default Behavior In V0, the `seed` parameter defaults to `None`. When the `seed` parameter is `None`, the random states for `random`, `np.random`, and `torch.manual_seed` are not set. This means that each run of vLLM will produce different results if `temperature > 0`, as expected. In V1, the `seed` parameter defaults to `0` which sets the random state for each worker, so the results will remain consistent for each vLLM run even if `temperature > 0`. !!! note It is impossible to un-specify a seed for V1 because different workers need to sample the same outputs for workflows such as speculative decoding. For more information, see: ### Locality of random state The random state in user code (i.e. the code that constructs [LLM][vllm.LLM] class) is updated by vLLM under the following conditions: - For V0: The seed is specified. - For V1: The workers are run in the same process as user code, i.e.: `VLLM_ENABLE_V1_MULTIPROCESSING=0`. By default, these conditions are not active so you can use vLLM without having to worry about accidentally making deterministic subsequent operations that rely on random state.