Unverified Commit 4d022cbc authored by Nicolò Lucchesi's avatar Nicolò Lucchesi Committed by GitHub
Browse files

[TPU][V1] Make `--disable_chunked_mm_input` mandatory for serving MM models (#16483)


Signed-off-by: default avatarNickLucche <nlucches@redhat.com>
parent 70de35a8
......@@ -120,6 +120,13 @@ class TpuPlatform(Platform):
assert not vllm_config.speculative_config, (
"Speculative decoding is not yet supported for TPU backend")
if scheduler_config.is_multimodal_model and not \
scheduler_config.disable_chunked_mm_input:
logger.warning("TPU does not support running Multimodal models"\
" without setting `--disable_chunked_mm_input`. " \
"Forcing --disable_chunked_mm_input.")
scheduler_config.disable_chunked_mm_input = True
@classmethod
def is_pin_memory_available(cls):
logger.warning("Pin memory is not supported on TPU.")
......
Markdown is supported
0% or .
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment