SID Visualization Example
To run the demo you will need the following:
Docker
docker-compose
(Tested with version 1.29)
To run this demo, ensure all submodules are checked out:
git submodule update --init --recursive
Build Morpheus Dev Container
Before launching the demo, we need the dev container for Morpheus to be created:
export DOCKER_IMAGE_TAG="sid-viz"
./docker/build_container_dev.sh
Launch User Interface
We will use docker-compose to build and run the entire demo. To launch everything, run the following from the repo root:
Save the Morpheus repo directory:
export MORPHEUS_ROOT=$(git rev-parse --show-toplevel)
Ensure SID model is downloaded for deployment to Triton:
./scripts/fetch_data.py fetch models
Change to the example directory:
cd ${MORPHEUS_ROOT}/examples/sid_visualization
Launch the containers:
DOCKER_BUILDKIT=1 docker-compose up --build -d
The following GUI should be displayed when all containers have completed launching:
Build Morpheus
Once docker-compose command has completed and GUI is displayed, exec into the container to build and run Morpheus:
Exec into the morpheus container:
docker-compose exec morpheus bash
Inside the container, compile morpheus:
BUILD_DIR=build-docker ./scripts/compile.sh
Install morpheus with an extra dependency:
pip install -e .
Verify Morpheus is installed:
morpheus --version
Ensure the data has been downloaded:
./scripts/fetch_data.py fetch examples
Keep this shell in the Morpheus Dev container running. It will be used later to start Morpheus.
Running Morpheus
After the GUI has been launched, Morpheus now needs to be started. In the same shell used to build Morpheus (the one running the Morpheus Dev container), run the following:
python examples/sid_visualization/run.py \
--debug --use_cpp=False --num_threads=1 \
--triton_server_url=triton:8001 \
--input_file=./examples/data/sid_visualization/group1-benign-2nodes.jsonlines \
--input_file=./examples/data/sid_visualization/group2-benign-50nodes.jsonlines \
--input_file=./examples/data/sid_visualization/group3-si-50nodes.jsonlines \
--input_file=./examples/data/sid_visualization/group4-benign-49nodes.jsonlines
Note: The first run of this script will take a few minutes to allow Triton to convert the deployed ONNX model to TensorRT. Subsequent runs will not include this conversion step so will be much quicker.
This launch will use all of the available datasets. Each dataset will show up as one batch in the visualization. Here is a description of each dataset:
examples/data/sid_visualization/group1-benign-2nodes.jsonlines
Small scale with 2 nodes, no SID
examples/data/sid_visualization/group2-benign-50nodes.jsonlines
Scale up to 50 nodes, no SID
examples/data/sid_visualization/group3-si-50nodes.jsonlines
50 nodes, with SID from a single node
examples/data/sid_visualization/group4-benign-49nodes.jsonlines
Isolate bad node leaving 49 nodes, no SID
The following is a screenshot after all four batches have been processed:
Use the slider or the following buttons to step through the inferences batches in the visualization:
Step to previous inference batch |
|
Step to next inference batch |
|
Step through all inference batches from beginning |
|
Pause animation |
The visualization on the right shows nodes in the current inference batch represented as
green spheres. White (benign) and red (SI) packets are shown flowing between the node connections.
While the animation is running, you can click the pause button or toggle off Simulating
. Once paused,
you will be able to hover over an individual packet to view its contents.
Changing the dataset does not require relaunching the GUI. Simply re-run Morpheus with the new dataset and the GUI will be updated.
It’s also possible to launch the demo using the Morpheus CLI using the following:
DEMO_DATASET="examples/data/sid_visualization/group1-benign-2nodes.jsonlines"
morpheus --log_level=DEBUG \
run --num_threads=1 --pipeline_batch_size=1024 --model_max_batch_size=32 --edge_buffer_size=4 --use_cpp=False \
pipeline-nlp --model_seq_length=256 \
from-file --filename=${DEMO_DATASET} \
deserialize \
preprocess --vocab_hash_file=morpheus/data/bert-base-uncased-hash.txt --truncation=True --do_lower_case=True --add_special_tokens=False \
inf-triton --model_name=sid-minibert-onnx --server_url=triton:8001 --force_convert_inputs=True \
monitor --description Inference\Rate --unit=inf \
add-class \
gen-viz
Note, this launch method is more useful for showing performance than showing capability.