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

Fix viz #12

Merged
merged 2 commits into from
Dec 11, 2024
Merged

Fix viz #12

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
0e95e67
fix: viz mean
dPys Dec 11, 2024
a3731a9
fix: add missing vanilla graphblas import
dPys Dec 11, 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
13 changes: 9 additions & 4 deletions nxbench/benchmarks/benchmark.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -199,11 +199,15 @@ def prepare_benchmark(

if "graphblas" in backend and is_graphblas_available():
try:
gb = import_module("graphblas")
ga = import_module("graphblas_algorithms")
except ImportError:
logger.exception("graphblas_algorithms backend not available")
return None
try:
logger.info(
f"GraphBlas Algorithms nthreads={gb.ss.config['nthreads']} "
)
return ga.Graph.from_networkx(original_graph)
except Exception:
logger.exception("Error converting graph to graphblas format")
Expand All @@ -230,10 +234,11 @@ def do_benchmark(

try:
algo_func = get_algorithm_function(algo_config, backend)
if isinstance(algo_func, partial):
alg_func_name = algo_func.func.__name__
else:
alg_func_name = algo_func.__name__
alg_func_name = (
algo_func.func.__name__
if isinstance(algo_func, partial)
else algo_func.__name__
)
logger.debug(f"Got algorithm function: {alg_func_name}")
except (ImportError, AttributeError):
logger.exception(f"Function not available for backend {backend}")
Expand Down
2 changes: 2 additions & 0 deletions nxbench/configs/example.yaml
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -241,6 +241,7 @@ validation:
matrix:
backend:
- "networkx"
- "parallel"
- "graphblas"
num_threads:
- "1"
Expand All @@ -250,6 +251,7 @@ matrix:
env_config:
req:
- "networkx==3.4.2"
- "nx-parallel-0.3rc0.dev0"
- "graphblas_algorithms==2023.10.0"
pythons:
- "3.10"
Expand Down
127 changes: 88 additions & 39 deletions nxbench/viz/app.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -14,7 +14,6 @@ def run_server(port=8050, debug=False):
pd.DataFrame.iteritems = pd.DataFrame.items

essential_columns = ["algorithm", "execution_time", "memory_used"]

df = df.dropna(subset=essential_columns)

df["execution_time"] = pd.to_numeric(df["execution_time"], errors="coerce")
Expand All @@ -40,7 +39,6 @@ def run_server(port=8050, debug=False):
for col in string_columns:
df[col] = df[col].astype(str).str.strip().str.lower()

aggregation_columns = ["execution_time", "memory_used"]
group_columns = [
"algorithm",
"dataset",
Expand All @@ -56,8 +54,14 @@ def run_server(port=8050, debug=False):
"os",
"num_thread",
]
df = df.groupby(group_columns, as_index=False)[aggregation_columns].mean()
df.set_index(group_columns, inplace=True)

# compute both mean and count
df_agg = df.groupby(group_columns, as_index=False).agg(
mean_execution_time=("execution_time", "mean"),
mean_memory_used=("memory_used", "mean"),
sample_count=("execution_time", "size"),
)
df_agg.set_index(group_columns, inplace=True)

available_parcats_columns = [col for col in group_columns if col != "algorithm"]

Expand All @@ -72,12 +76,12 @@ def run_server(port=8050, debug=False):
options=[
{"label": alg.title(), "value": alg}
for alg in sorted(
df.index.get_level_values("algorithm").unique()
df_agg.index.get_level_values("algorithm").unique()
)
],
value=sorted(df.index.get_level_values("algorithm").unique())[
0
],
value=sorted(
df_agg.index.get_level_values("algorithm").unique()
)[0],
clearable=False,
style={"width": "100%"},
),
Expand Down Expand Up @@ -117,7 +121,7 @@ def run_server(port=8050, debug=False):
{"label": c.replace("_", " ").title(), "value": c}
for c in available_parcats_columns
],
value=available_parcats_columns, # default select all
value=available_parcats_columns,
multi=True,
style={"width": "100%"},
),
Expand All @@ -144,11 +148,12 @@ def run_server(port=8050, debug=False):
active_tab="parcats-tab",
style={"marginTop": "20px"},
),
dcc.Store(id="mean-values-store"),
]
)

@app.callback(
Output("benchmark-graph", "figure"),
[Output("benchmark-graph", "figure"), Output("mean-values-store", "data")],
[
Input("algorithm-dropdown", "value"),
Input("color-toggle", "value"),
Expand All @@ -159,7 +164,7 @@ def update_graph(selected_algorithm, color_by, selected_dimensions):
selected_algorithm = selected_algorithm.lower()

try:
filtered_df = df.xs(selected_algorithm, level="algorithm")
filtered_df = df_agg.xs(selected_algorithm, level="algorithm")
except KeyError:
fig = go.Figure()
fig.update_layout(
Expand All @@ -174,7 +179,7 @@ def update_graph(selected_algorithm, color_by, selected_dimensions):
}
],
)
return fig
return fig, []

if filtered_df.empty:
fig = go.Figure()
Expand All @@ -190,15 +195,18 @@ def update_graph(selected_algorithm, color_by, selected_dimensions):
}
],
)
return fig
return fig, []

if color_by == "execution_time":
color_values = filtered_df["execution_time"]
mean_values = filtered_df["mean_execution_time"]
colorbar_title = "Execution Time (s)"
else:
color_values = filtered_df["memory_used"]
mean_values = filtered_df["mean_memory_used"]
colorbar_title = "Memory Used (GB)"

counts = filtered_df["sample_count"].values
color_values = mean_values.values

dims = [
{
"label": dim_col.replace("_", " ").title(),
Expand All @@ -207,24 +215,27 @@ def update_graph(selected_algorithm, color_by, selected_dimensions):
for dim_col in selected_dimensions
]

parcats = go.Parcats(
dimensions=dims,
line={
"color": color_values,
"colorscale": "Tealrose",
"showscale": True,
"colorbar": {"title": colorbar_title},
},
hoverinfo="count",
fig = go.Figure()
fig.add_trace(
go.Parcats(
dimensions=dims,
line={
"color": color_values,
"colorscale": "Tealrose",
"showscale": True,
"colorbar": {"title": colorbar_title},
},
counts=counts,
hoverinfo="count",
hovertemplate="Count: %{count}\nMean: REPLACE_ME<extra></extra>",
)
)

fig = go.Figure(data=parcats)
fig.update_layout(
title=f"Benchmark Results for {selected_algorithm.title()}",
template="plotly_white",
)

return fig
return fig, color_values.tolist()

@app.callback(
Output("violin-graph", "figure"),
Expand All @@ -237,7 +248,7 @@ def update_graph(selected_algorithm, color_by, selected_dimensions):
def update_violin(selected_algorithm, color_by, selected_dimensions):
selected_algorithm = selected_algorithm.lower()
try:
filtered_df = df.xs(selected_algorithm, level="algorithm").reset_index()
filtered_df = df_agg.xs(selected_algorithm, level="algorithm").reset_index()
except KeyError:
fig = go.Figure()
fig.update_layout(
Expand Down Expand Up @@ -270,7 +281,11 @@ def update_violin(selected_algorithm, color_by, selected_dimensions):
)
return fig

y_metric = "execution_time" if color_by == "execution_time" else "memory_used"
y_metric = (
"mean_execution_time"
if color_by == "execution_time"
else "mean_memory_used"
)
y_label = "Execution Time" if color_by == "execution_time" else "Memory Used"

violin_dimension = selected_dimensions[0] if selected_dimensions else "backend"
Expand All @@ -296,6 +311,7 @@ def update_violin(selected_algorithm, color_by, selected_dimensions):
"cpu",
"os",
"num_thread",
"sample_count",
],
title=f"{y_label} Distribution for {selected_algorithm.title()}",
)
Expand All @@ -304,22 +320,55 @@ def update_violin(selected_algorithm, color_by, selected_dimensions):

app.clientside_callback(
"""
function(hoverData) {
setTimeout(() => {
const tooltips = document.querySelectorAll('.hoverlayer .hovertext');
tooltips.forEach(tooltip => {
const textNode = tooltip.querySelector('text');
if (textNode && textNode.textContent.includes('Count:')) {
textNode.textContent = textNode.textContent.replace('Count:',
'Mean:');
function(hoverData, meanValues) {
if (!hoverData || !hoverData.points || hoverData.points.length === 0) {
return null;
}

if (!meanValues) {
// No mean values available yet
return null;
}

var point = hoverData.points[0];
var pointIndex = point.pointNumber;
var meanValue = meanValues[pointIndex];

const tooltips = document.querySelectorAll('.hoverlayer .hovertext');
// Create a MutationObserver that waits for the tooltip to appear
const observer = new MutationObserver(mutations => {
let replaced = false;
mutations.forEach(mutation => {
if (mutation.type === 'childList') {
const tooltips = document.querySelectorAll('.hoverlayer .
hovertext text');
tooltips.forEach(tNode => {
if (tNode.textContent.includes('REPLACE_ME')) {
tNode.textContent = tNode.textContent.replace(
'REPLACE_ME',
meanValue.toFixed(3)
);
replaced = true;
}
});
}
});
}, 100);
// Once replaced, disconnect the observer to stop unnecessary monitoring
if (replaced) {
observer.disconnect();
}
});

const hoverlayer = document.querySelector('.hoverlayer');
if (hoverlayer) {
observer.observe(hoverlayer, { childList: true, subtree: true });
}

return null;
}
""",
Output("hover-text-hack", "children"),
[Input("benchmark-graph", "hoverData")],
[Input("benchmark-graph", "hoverData"), Input("mean-values-store", "data")],
)

app.run_server(port=port, debug=debug)
Expand Down
Loading