Spaces:
Sleeping
Sleeping
File size: 14,188 Bytes
a55120a | 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 388 389 390 391 392 393 394 395 396 397 398 399 400 401 402 403 404 405 406 | """The System Metrics page for the Trackio UI (GPU metrics, etc.)."""
import gradio as gr
import pandas as pd
import trackio.utils as utils
from trackio.sqlite_storage import SQLiteStorage
from trackio.ui import fns
from trackio.ui.components.colored_checkbox import ColoredCheckboxGroup
from trackio.ui.helpers.run_selection import RunSelection
def get_runs(project) -> list[str]:
if not project:
return []
return SQLiteStorage.get_runs(project)
def refresh_runs(
project: str | None,
filter_text: str | None,
selection: RunSelection,
):
if project is None:
runs: list[str] = []
else:
runs = get_runs(project)
if filter_text:
runs = [r for r in runs if filter_text in r]
did_change = selection.update_choices(runs)
return (
fns.run_checkbox_update(selection) if did_change else gr.skip(),
gr.Textbox(label=f"Runs ({len(runs)})"),
selection,
)
def load_system_data(
project: str | None,
run: str | None,
) -> pd.DataFrame | None:
if not project or not run:
return None
logs = SQLiteStorage.get_system_logs(project, run)
if not logs:
return None
df = pd.DataFrame(logs)
if "timestamp" in df.columns:
df["timestamp"] = pd.to_datetime(df["timestamp"])
first_timestamp = df["timestamp"].min()
df["time"] = (df["timestamp"] - first_timestamp).dt.total_seconds()
df["run"] = run
return df
with gr.Blocks() as system_page:
with gr.Sidebar() as sidebar:
logo = fns.create_logo()
project_dd = fns.create_project_dropdown()
with gr.Group():
run_tb = gr.Textbox(label="Runs", placeholder="Type to filter...")
run_cb = ColoredCheckboxGroup(choices=[], colors=[], label="Runs")
gr.HTML("<hr>")
realtime_cb = gr.Checkbox(label="Refresh metrics realtime", value=True)
smoothing_slider = gr.Slider(
label="Smoothing Factor",
minimum=0,
maximum=20,
value=0,
step=1,
info="0 = no smoothing",
)
navbar = fns.create_navbar()
timer = gr.Timer(value=1)
run_selection_state = gr.State(RunSelection())
x_lim = gr.State(None)
last_system_update = gr.State({})
def toggle_timer(cb_value):
if cb_value:
return gr.Timer(active=True)
else:
return gr.Timer(active=False)
def update_x_lim(select_data: gr.SelectData):
return select_data.index
def check_system_metrics_update(project: str | None, runs: list[str]) -> dict:
if not project or not runs:
return {}
result = {}
for run in runs:
logs = SQLiteStorage.get_system_logs(project, run)
result[run] = len(logs) if logs else 0
return result
@gr.render(
triggers=[
system_page.load,
run_cb.change,
last_system_update.change,
smoothing_slider.change,
x_lim.change,
],
inputs=[
project_dd,
run_cb,
smoothing_slider,
x_lim,
run_selection_state,
],
show_progress="hidden",
queue=False,
)
def update_system_dashboard(
project,
runs,
smoothing_granularity,
x_lim_value,
selection,
):
dfs = []
original_runs = runs.copy() if runs else []
for run in runs:
df = load_system_data(project, run)
if df is not None:
dfs.append(df)
if not dfs:
if not SQLiteStorage.has_system_metrics(project) if project else True:
gr.Markdown(
"""
## No System Metrics Available
System metrics (GPU) will appear here once logged. To enable automatic GPU logging:
```python
import trackio
# GPU logging is auto-enabled when nvidia-ml-py is installed and a GPU is detected
run = trackio.init(project="my-project")
# Or explicitly enable it:
run = trackio.init(project="my-project", auto_log_gpu=True)
# You can also manually log GPU metrics:
trackio.log_gpu()
```
"""
)
else:
gr.Markdown("*Select runs to view system metrics*")
return
master_df = pd.concat(dfs, ignore_index=True)
if master_df.empty:
gr.Markdown("*No system metrics found for selected runs*")
return
x_column = "time"
numeric_cols = master_df.select_dtypes(include="number").columns
numeric_cols = [c for c in numeric_cols if c not in ["time", "timestamp"]]
if smoothing_granularity > 0:
window_size = max(3, min(smoothing_granularity, len(master_df)))
for col in numeric_cols:
master_df[col] = master_df.groupby("run")[col].transform(
lambda x: x.rolling(
window=window_size, center=True, min_periods=1
).mean()
)
ordered_groups, nested_metric_groups = utils.order_metrics_by_plot_preference(
list(numeric_cols)
)
all_runs = selection.choices if selection else original_runs
color_map = utils.get_color_mapping(all_runs, False)
metric_idx = 0
for group_name in ordered_groups:
group_data = nested_metric_groups[group_name]
total_plot_count = sum(
1
for m in group_data["direct_metrics"]
if not master_df.dropna(subset=[m]).empty
) + sum(
sum(1 for m in metrics if not master_df.dropna(subset=[m]).empty)
for metrics in group_data["subgroups"].values()
)
group_label = (
f"{group_name} ({total_plot_count})"
if total_plot_count > 0
else group_name
)
with gr.Accordion(
label=group_label,
open=True,
key=f"sys-accordion-{group_name}",
preserved_by_key=["value", "open"],
):
if group_data["direct_metrics"]:
with gr.Draggable(
key=f"sys-row-{group_name}-direct", orientation="row"
):
for metric_name in group_data["direct_metrics"]:
metric_df = master_df.dropna(subset=[metric_name])
color = "run" if "run" in metric_df.columns else None
downsampled_df, updated_x_lim = utils.downsample(
metric_df,
x_column,
metric_name,
color,
x_lim_value,
)
if not metric_df.empty:
plot = gr.LinePlot(
downsampled_df,
x=x_column,
y=metric_name,
x_title="Time (seconds)",
y_title=metric_name.split("/")[-1],
color=color,
color_map=color_map,
colors_in_legend=original_runs,
title=metric_name,
key=f"sys-plot-{metric_idx}",
preserved_by_key=None,
buttons=["fullscreen", "export"],
x_lim=updated_x_lim,
min_width=400,
)
plot.select(
update_x_lim,
outputs=x_lim,
key=f"sys-select-{metric_idx}",
)
plot.double_click(
lambda: None,
outputs=x_lim,
key=f"sys-double-{metric_idx}",
)
metric_idx += 1
if group_data["subgroups"]:
for subgroup_name in sorted(group_data["subgroups"].keys()):
subgroup_metrics = group_data["subgroups"][subgroup_name]
subgroup_plot_count = sum(
1
for m in subgroup_metrics
if not master_df.dropna(subset=[m]).empty
)
subgroup_label = (
f"{subgroup_name} ({subgroup_plot_count})"
if subgroup_plot_count > 0
else subgroup_name
)
with gr.Accordion(
label=subgroup_label,
open=True,
key=f"sys-accordion-{group_name}-{subgroup_name}",
preserved_by_key=["value", "open"],
):
with gr.Draggable(
key=f"sys-row-{group_name}-{subgroup_name}",
orientation="row",
):
for metric_name in subgroup_metrics:
metric_df = master_df.dropna(subset=[metric_name])
color = (
"run" if "run" in metric_df.columns else None
)
downsampled_df, updated_x_lim = utils.downsample(
metric_df,
x_column,
metric_name,
color,
x_lim_value,
)
if not metric_df.empty:
plot = gr.LinePlot(
downsampled_df,
x=x_column,
y=metric_name,
x_title="Time (seconds)",
y_title=metric_name.split("/")[-1],
color=color,
color_map=color_map,
colors_in_legend=original_runs,
title=metric_name,
key=f"sys-plot-{metric_idx}",
preserved_by_key=None,
buttons=["fullscreen", "export"],
x_lim=updated_x_lim,
min_width=400,
)
plot.select(
update_x_lim,
outputs=x_lim,
key=f"sys-select-{metric_idx}",
)
plot.double_click(
lambda: None,
outputs=x_lim,
key=f"sys-double-{metric_idx}",
)
metric_idx += 1
gr.on(
[timer.tick],
fn=lambda: gr.Dropdown(info=fns.get_project_info()),
outputs=[project_dd],
show_progress="hidden",
api_visibility="private",
)
gr.on(
[timer.tick],
fn=refresh_runs,
inputs=[project_dd, run_tb, run_selection_state],
outputs=[run_cb, run_tb, run_selection_state],
show_progress="hidden",
api_visibility="private",
)
gr.on(
[timer.tick],
fn=check_system_metrics_update,
inputs=[project_dd, run_cb],
outputs=last_system_update,
show_progress="hidden",
api_visibility="private",
)
gr.on(
[system_page.load],
fn=fns.get_projects,
outputs=project_dd,
show_progress="hidden",
queue=False,
api_visibility="private",
).then(
fns.update_navbar_value,
inputs=[project_dd],
outputs=[navbar],
show_progress="hidden",
api_visibility="private",
queue=False,
)
gr.on(
[system_page.load, project_dd.change],
fn=refresh_runs,
inputs=[project_dd, run_tb, run_selection_state],
outputs=[run_cb, run_tb, run_selection_state],
show_progress="hidden",
queue=False,
api_visibility="private",
).then(
fns.update_navbar_value,
inputs=[project_dd],
outputs=[navbar],
show_progress="hidden",
api_visibility="private",
queue=False,
)
realtime_cb.change(
fn=toggle_timer,
inputs=realtime_cb,
outputs=timer,
api_visibility="private",
queue=False,
)
run_cb.input(
fn=fns.handle_run_checkbox_change,
inputs=[run_cb, run_selection_state],
outputs=run_selection_state,
api_visibility="private",
queue=False,
)
run_tb.input(
fn=refresh_runs,
inputs=[project_dd, run_tb, run_selection_state],
outputs=[run_cb, run_tb, run_selection_state],
api_visibility="private",
queue=False,
show_progress="hidden",
)
|