Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

[Hotfix][Core][VLM] Disable chunked prefill by default and prefix caching for multimodal models #8425

Merged
merged 3 commits into from
Sep 12, 2024
Merged

[Hotfix][Core][VLM] Disable chunked prefill by default and prefix caching for multimodal models #8425

Changes from 1 commit
Commits
Show all changes
3 commits
Select commit Hold shift + click to select a range
7f92511
patch
ywang96 Sep 12, 2024
ce7d366
address typo
ywang96 Sep 12, 2024
cb965db
preserve alphabetical order of list of models
ywang96 Sep 12, 2024
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
12 changes: 11 additions & 1 deletion vllm/engine/arg_utils.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -843,6 +843,13 @@ def create_engine_config(self) -> EngineConfig:
device_config = DeviceConfig(device=self.device)
model_config = self.create_model_config()

if model_config.is_multimodal_model:
if self.enable_prefix_caching:
logger.warning(
"--enable-prefix-caching is currently not "
"supported for multimodal models and has been disabled.")
self.enable_prefix_caching = False

cache_config = CacheConfig(
block_size=self.block_size if self.device != "neuron" else
self.max_model_len, # neuron needs block_size = max_model_len
Expand Down Expand Up @@ -874,7 +881,10 @@ def create_engine_config(self) -> EngineConfig:
# If not explicitly set, enable chunked prefill by default for
# long context (> 32K) models. This is to avoid OOM errors in the
# initial memory profiling phase.
if use_long_context:

# Chunked prefill is currently disabled for multimodal models by
# default.
if use_long_context and model_config.is_multimodal_model:
ywang96 marked this conversation as resolved.
Show resolved Hide resolved
is_gpu = device_config.device_type == "cuda"
use_sliding_window = (model_config.get_sliding_window()
is not None)
Expand Down
Loading