[Core][Frontend] Support Passing Multimodal Processor Kwargs (#8657)

Signed-off-by: Alex-Brooks <Alex.Brooks@ibm.com>
This commit is contained in:
Alex Brooks
2024-09-23 07:44:48 +00:00
committed by GitHub
parent d23679eb99
commit 9b8c8ba119
16 changed files with 590 additions and 117 deletions
+2 -1
View File
@@ -235,7 +235,7 @@ class LLMEngine:
"decoding_config=%r, observability_config=%r, "
"seed=%d, served_model_name=%s, use_v2_block_manager=%s, "
"num_scheduler_steps=%d, enable_prefix_caching=%s, "
"use_async_output_proc=%s)",
"use_async_output_proc=%s, mm_processor_kwargs=%s)",
VLLM_VERSION,
model_config.model,
speculative_config,
@@ -268,6 +268,7 @@ class LLMEngine:
scheduler_config.num_scheduler_steps,
cache_config.enable_prefix_caching,
model_config.use_async_output_proc,
model_config.mm_processor_kwargs,
)
# TODO(woosuk): Print more configs in debug mode.
from vllm.plugins import load_general_plugins