Support encoder-only models without KV-Cache (#21270)

Signed-off-by: Max de Bayser <maxdebayser@gmail.com>
Signed-off-by: Max de Bayser <mbayser@br.ibm.com>
Co-authored-by: Russell Bryant <rbryant@redhat.com>
This commit is contained in:
Maximilien de Bayser
2025-07-26 21:09:52 +08:00
committed by GitHub
co-authored by Russell Bryant
parent f27fdfc3ed
commit 1cd6eaba54
17 changed files with 352 additions and 99 deletions
+1
View File
@@ -93,6 +93,7 @@ def create_common_attn_metadata(
max_query_len=max_query_len,
block_table_tensor=block_table_tensor,
slot_mapping=slot_mapping,
causal=True,
)
-1
View File
@@ -13,7 +13,6 @@ UNSUPPORTED_MODELS_V1 = [
"openai/whisper-large-v3", # transcription
"facebook/bart-large-cnn", # encoder decoder
"state-spaces/mamba-130m-hf", # mamba1
"BAAI/bge-m3", # embedding
]
MODEL = "meta-llama/Llama-3.2-1B-Instruct"
+1 -2
View File
@@ -1,9 +1,8 @@
# SPDX-License-Identifier: Apache-2.0
# SPDX-FileCopyrightText: Copyright contributors to the vLLM project
import re
import pytest
import regex as re
import requests
import torch