06 Split-GPU + Graphistry¶

Source: notebooks/06-split-gpu-graphistry-llamatelemetry-v0-1-1-e1.ipynb

Notebook focus¶

This page is a cell-by-cell walkthrough of the notebook, explaining the intent of each step and showing the exact code executed.

Cell-by-cell walkthrough¶

Cell 1 (Markdown)¶

06 Split-GPU + Graphistry¶

Run llama.cpp on GPU 0 and Graphistry/RAPIDS graph analytics on GPU 1.

What you will learn: - Use split_gpu_session to partition GPUs - Build knowledge graphs with GraphistryBuilders - Visualize graphs with GraphistrySession (optional credentials)

Requirements: Kaggle T4 x2. Graphistry credentials (optional) for visualization.

Cell 2 (Markdown)¶

1) Install¶

Cell 3 (Code)¶

Summary: Installs required dependencies and runtime tools.

!pip -q install git+https://github.com/llamatelemetry/llamatelemetry.git@v0.1.1

Cell 4 (Markdown)¶

2) Split GPU session¶

The context manager pins the LLM workload to one GPU and reserves the other for graph/analytics tasks.

Cell 5 (Code)¶

Summary: Imports core libraries: llamatelemetry.

from llamatelemetry.kaggle import split_gpu_session

with split_gpu_session(llm_gpu=0, graph_gpu=1) as ctx:
    print("LLM server kwargs:", ctx["llm_server_kwargs"])
    print("Graph GPU ID:", ctx.get("graph_gpu"))

Cell 6 (Markdown)¶

3) Build a knowledge graph¶

GraphistryBuilders.knowledge_graph() creates node/edge DataFrames from entity and relationship lists.

Cell 7 (Code)¶

Summary: Imports core libraries: llamatelemetry. Sets up Graphistry for graph visualization or analytics.

from llamatelemetry.graphistry import GraphistryBuilders

entities = [
    {"id": "llamatelemetry", "label": "llamatelemetry SDK"},
    {"id": "llama.cpp", "label": "llama.cpp"},
    {"id": "otel", "label": "OpenTelemetry"},
    {"id": "graphistry", "label": "Graphistry"},
]
relationships = [
    {"source": "llamatelemetry", "target": "llama.cpp", "type": "wraps"},
    {"source": "llamatelemetry", "target": "otel", "type": "exports_to"},
    {"source": "llamatelemetry", "target": "graphistry", "type": "visualizes_with"},
]

nodes_df, edges_df = GraphistryBuilders.knowledge_graph(entities, relationships)
print("Nodes:")
print(nodes_df)
print("\nEdges:")
print(edges_df)

Cell 8 (Markdown)¶

4) Visualize with Graphistry (optional)¶

Requires Graphistry credentials stored in Kaggle Secrets: - GRAPHISTRY_USERNAME - GRAPHISTRY_PASSWORD - GRAPHISTRY_SERVER (optional, defaults to hub.graphistry.com)

Cell 9 (Code)¶

Summary: Sets up Graphistry for graph visualization or analytics.

# from llamatelemetry.graphistry import GraphistrySession
#
# session = GraphistrySession.from_kaggle_secrets()
# g = session.connector.create_graph(edges_df, nodes_df=nodes_df)
# g.plot()