| 1 | # SPDX-License-Identifier: Apache-2.0 |
| 2 | # SPDX-FileCopyrightText: Copyright contributors to the vLLM project |
| 3 | """Encode/decode utilities for multimodal tensors and field metadata |
| 4 | over JSON/HTTP, used by the disaggregated generate endpoint.""" |
| 5 | |
| 6 | from __future__ import annotations |
| 7 | |
| 8 | import pybase64 |
| 9 | |
| 10 | from vllm.multimodal.inputs import MultiModalKwargsItem |
| 11 | from vllm.v1.serial_utils import MsgpackDecoder, MsgpackEncoder |
| 12 | |
| 13 | _encoder = MsgpackEncoder(size_threshold=2**62) # force all tensors inline |
| 14 | _decoder = MsgpackDecoder(t=MultiModalKwargsItem) |
| 15 | |
| 16 | |
| 17 | def encode_mm_kwargs_item(item: MultiModalKwargsItem) -> str: |
| 18 | """Serialize a MultiModalKwargsItem to a base64 string.""" |
| 19 | bufs = _encoder.encode(item) |
| 20 | assert len(bufs) == 1, "All tensors should be inline" |
| 21 | return pybase64.b64encode(bufs[0]).decode("ascii") |
| 22 | |
| 23 | |
| 24 | def decode_mm_kwargs_item(data: str) -> MultiModalKwargsItem: |
| 25 | """Deserialize a base64 string back to a MultiModalKwargsItem.""" |
| 26 | raw = pybase64.b64decode(data) |
| 27 | return _decoder.decode(raw) |