[Frontend] Add routed-experts prompt offset (#51906)

Signed-off-by: aoshen02 <aoshen@inferact.ai>
Co-authored-by: OpenAI Codex <codex@openai.com>
Co-authored-by: mergify[bot] <37929162+mergify[bot]@users.noreply.github.com>
This commit is contained in:
aoshen02
2026-08-13 08:09:00 -07:00
committed by GitHub
co-authored by OpenAI Codex mergify[bot] <37929162+mergify[bot]@users.noreply.github.com>
parent f3c1638927
commit 7553aac77b
13 changed files with 114 additions and 61 deletions
+16
View File
@@ -1,5 +1,9 @@
# SPDX-License-Identifier: Apache-2.0
# SPDX-FileCopyrightText: Copyright contributors to the vLLM project
import io
import numpy as np
import pybase64
import pytest
import torch
@@ -12,6 +16,7 @@ from vllm.utils.serial_utils import (
Endianness,
MmMetadataDType,
binary2tensor,
numpy2base64,
tensor2binary,
)
@@ -19,6 +24,17 @@ FLOAT_EMBED_DTYPES = tuple(EMBED_DTYPES.keys())
INTEGER_EMBED_DTYPES = tuple(MM_METADATA_DTYPES.keys())
def test_numpy2base64_round_trip():
array = np.arange(24, dtype=np.uint8).reshape(2, 3, 4)
decoded = np.load(
io.BytesIO(pybase64.b64decode(numpy2base64(array))),
allow_pickle=False,
)
np.testing.assert_array_equal(decoded, array)
def _build_integer_tensor(
embed_dtype: MmMetadataDType, shape: tuple[int, ...]
) -> torch.Tensor: