fix: address comments

Signed-off-by: Ilana Nguyen <inguyen@nvidia.com>

Author: ilana-n

README.md

@@ -63,7 +63,6 @@ Features
 | **[Template Endpoint](docs/tutorials/template-endpoint.md)** | Benchmark custom APIs with flexible Jinja2 request templates | Custom API formats, rapid prototyping, non-standard endpoints |
 | **[SGLang Image Generation](docs/tutorials/sglang-image-generation.md)** | Benchmark image generation APIs using SGLang with FLUX.1-dev model | Image generation testing, text-to-image benchmarking, extracting generated images |
 | **[Visualization & Plotting](docs/tutorials/plot.md)** | Generate PNG visualizations with automatic mode detection (single-run analysis or multi-run comparison) | Parameter sweep analysis, performance debugging, model comparison |
-|
 
 ### Working with Benchmark Data
 - **[Profile Exports](docs/tutorials/working-with-profile-exports.md)** - Parse and analyze `profile_export.jsonl` with Pydantic models, custom metrics, and async processing

docs/tutorials/plot.md

@@ -62,7 +62,7 @@ Compares metrics across multiple profiling runs to identify optimal configuratio
 - Multiple paths specified as arguments
 
 **Example:**
-```
+```text
 artifacts/sweep_qwen/
 ├── Qwen3-0.6B-concurrency1/
 ├── Qwen3-0.6B-concurrency2/
@@ -99,7 +99,7 @@ Analyzes performance over time for a single profiling run.
 - Directory contains `profile_export.jsonl` directly
 
 **Example:**
-```
+```text
 artifacts/single_run/
 └── profile_export.jsonl
 ```
@@ -231,7 +231,7 @@ experiment_classification:
 #### Example
 
 **Directory structure:**
-```
+```text
 artifacts/
 ├── baseline_moderate_io_isl100_osl200_streaming/           # Grey
 │   ├── concurrency_1/
@@ -330,8 +330,10 @@ The dashboard automatically detects visualization mode (multi-run comparison or
 - Token Throughput per GPU vs Interactivity
 
 **Single-run plots** (time series):
+```text
 - GPU Utilization Over Time
 - GPU Memory Usage Over Time
+```
 
 ![GPU Utilization and Throughput Over Time](../diagrams/plot_examples/single_run/time_series/gpu_utilization_and_throughput_over_time.png)
 
@@ -345,10 +347,12 @@ The dashboard automatically detects visualization mode (multi-run comparison or
 When timeslice data is available (via `--slice-duration` during profiling), plots show performance evolution across time windows.
 
 **Generated timeslice plots:**
+```text
 - TTFT Across Timeslices
 - ITL Across Timeslices
 - Throughput Across Timeslices
 - Latency Across Timeslices
+```
 
 **Timeslices enable easy outlier identification and bucketing analysis**. Each time window (bucket) shows avg/p50/p95 statistics, making it simple to spot which periods have outlier performance. Slice 0 often shows cold-start overhead, while later slices may reveal degradation. Flat bars across slices may indicate stable performance; increasing trends can suggest resource exhaustion. Potentially useful for quickly isolating performance issues to specific phases (warmup, steady-state, or degradation).
 

src/aiperf/common/enums/__init__.py

@@ -56,7 +56,6 @@
     FrequencyMetricUnit,
     FrequencyMetricUnitInfo,
     GenericMetricUnit,
-    MetricDictValueTypeT,
     MetricFlags,
     MetricOverTimeUnit,
     MetricOverTimeUnitInfo,
@@ -65,10 +64,8 @@
     MetricTimeUnit,
     MetricTimeUnitInfo,
     MetricType,
-    MetricUnitT,
     MetricValueType,
     MetricValueTypeInfo,
-    MetricValueTypeT,
     MetricValueTypeVarT,
     PlotMetricDirection,
     PowerMetricUnit,
@@ -145,7 +142,6 @@
     "LifecycleState",
     "MediaType",
     "MessageType",
-    "MetricDictValueTypeT",
     "MetricFlags",
     "MetricOverTimeUnit",
     "MetricOverTimeUnitInfo",
@@ -154,10 +150,8 @@
     "MetricTimeUnit",
     "MetricTimeUnitInfo",
     "MetricType",
-    "MetricUnitT",
     "MetricValueType",
     "MetricValueTypeInfo",
-    "MetricValueTypeT",
     "MetricValueTypeVarT",
     "ModelSelectionStrategy",
     "PlotMetricDirection",

src/aiperf/dataset/__init__.py

@@ -40,7 +40,6 @@
     BaseLoader,
     BasePublicDatasetLoader,
     CustomDatasetT,
-    Filename,
     MediaConversionMixin,
     MooncakeTrace,
     MooncakeTraceDatasetLoader,
@@ -71,7 +70,6 @@
     "CustomDatasetT",
     "DEFAULT_CORPUS_FILE",
     "DatasetManager",
-    "Filename",
     "ImageGenerator",
     "MP3_SUPPORTED_SAMPLE_RATES",
     "MediaConversionMixin",

src/aiperf/dataset/loader/__init__.py

@@ -33,7 +33,6 @@
     MultiTurnDatasetLoader,
 )
 from aiperf.dataset.loader.random_pool import (
-    Filename,
     RandomPoolDatasetLoader,
 )
 from aiperf.dataset.loader.sharegpt import (
@@ -49,7 +48,6 @@
     "BaseLoader",
     "BasePublicDatasetLoader",
     "CustomDatasetT",
-    "Filename",
     "MediaConversionMixin",
     "MooncakeTrace",
     "MooncakeTraceDatasetLoader",

src/aiperf/endpoints/__init__.py

@@ -35,7 +35,6 @@
     SolidoEndpoint,
 )
 from aiperf.endpoints.template_endpoint import (
-    NAMED_TEMPLATES,
     TemplateEndpoint,
 )
 
@@ -49,7 +48,6 @@
     "HFTeiRankingsEndpoint",
     "HuggingFaceGenerateEndpoint",
     "ImageGenerationEndpoint",
-    "NAMED_TEMPLATES",
     "NIMRankingsEndpoint",
     "SolidoEndpoint",
     "TemplateEndpoint",

src/aiperf/gpu_telemetry/__init__.py

@@ -8,7 +8,6 @@
 
 from aiperf.gpu_telemetry.constants import (
     DCGM_TO_FIELD_MAPPING,
-    GPU_TELEMETRY_METRICS_CONFIG,
     SCALING_FACTORS,
     get_gpu_telemetry_metrics_config,
 )
@@ -24,7 +23,6 @@
 
 __all__ = [
     "DCGM_TO_FIELD_MAPPING",
-    "GPU_TELEMETRY_METRICS_CONFIG",
     "MetricsConfigLoader",
     "SCALING_FACTORS",
     "TelemetryDataCollector",

src/aiperf/plot/__init__.py

@@ -31,7 +31,6 @@
     DEFAULT_PNG_OUTPUT_DIR,
     DERIVED_METRIC_DIRECTIONS,
     LIGHT_THEME_COLORS,
-    METRIC_CATEGORY_RULES,
     NON_METRIC_KEYS,
     NVIDIA_BORDER_DARK,
     NVIDIA_BORDER_LIGHT,
@@ -56,8 +55,6 @@
 )
 from aiperf.plot.core import (
     DEFAULT_IGNORE_PARAMS,
-    DERIVED_METRICS_REGISTRY,
-    PLOT_TYPE_METADATA,
     DataLoader,
     DataSource,
     DerivedMetricCalculator,
@@ -243,7 +240,6 @@
     "DEFAULT_PLOT_HEIGHT",
     "DEFAULT_PLOT_WIDTH",
     "DEFAULT_PNG_OUTPUT_DIR",
-    "DERIVED_METRICS_REGISTRY",
     "DERIVED_METRIC_DIRECTIONS",
     "DashboardBuilder",
     "DashboardServer",
@@ -257,7 +253,6 @@
     "ExperimentClassificationConfig",
     "HistogramHandler",
     "LIGHT_THEME_COLORS",
-    "METRIC_CATEGORY_RULES",
     "MULTI_RUN_PLOT_TYPES",
     "MULTI_RUN_STAT_OPTIONS",
     "MetricSpec",
@@ -278,7 +273,6 @@
     "OUTLIER_RED",
     "PLOT_FONT_FAMILY",
     "PLOT_LOG_FILE",
-    "PLOT_TYPE_METADATA",
     "PROFILE_EXPORT_AIPERF_JSON",
     "PROFILE_EXPORT_GPU_TELEMETRY_JSONL",
     "PROFILE_EXPORT_JSONL",

src/aiperf/plot/core/__init__.py

@@ -6,7 +6,6 @@
 """
 
 from aiperf.plot.core.data_loader import (
-    DERIVED_METRICS_REGISTRY,
     DataLoader,
     DerivedMetricCalculator,
     RunData,
@@ -31,7 +30,6 @@
     get_nvidia_color_scheme,
 )
 from aiperf.plot.core.plot_specs import (
-    PLOT_TYPE_METADATA,
     DataSource,
     ExperimentClassificationConfig,
     MetricSpec,
@@ -55,14 +53,12 @@
 
 __all__ = [
     "DEFAULT_IGNORE_PARAMS",
-    "DERIVED_METRICS_REGISTRY",
     "DataLoader",
     "DataSource",
     "DerivedMetricCalculator",
     "ExperimentClassificationConfig",
     "MetricSpec",
     "ModeDetector",
-    "PLOT_TYPE_METADATA",
     "PlotGenerator",
     "PlotSpec",
     "PlotType",

src/aiperf/plot/core/data_loader.py

@@ -9,12 +9,12 @@
 formats suitable for visualization and analysis.
 """
 
-import json
 from fnmatch import fnmatch
 from pathlib import Path
 from typing import Any
 
 import numpy as np
+import orjson
 import pandas as pd
 from pydantic import Field
 
@@ -521,7 +521,7 @@ def _read_jsonl_with_error_handling(
                     try:
                         record = parse_func(line)
                         records.append(record)
-                    except (json.JSONDecodeError, Exception) as e:
+                    except (orjson.JSONDecodeError, Exception) as e:
                         corrupted_lines += 1
                         self.warning(
                             f"Skipping invalid line {line_num} in {jsonl_path}: {e}"
@@ -676,8 +676,8 @@ def _load_aggregated_json(self, json_path: Path) -> dict[str, Any]:
             raise DataLoadError("Required JSON file not found", path=str(json_path))
 
         try:
-            with open(json_path, encoding="utf-8") as f:
-                data = json.load(f)
+            with open(json_path, "rb") as f:
+                data = orjson.loads(f.read())
 
             if "metrics" in data and isinstance(data["metrics"], dict):
                 parsed_metrics = {}
@@ -691,7 +691,7 @@ def _load_aggregated_json(self, json_path: Path) -> dict[str, Any]:
 
             self.info(f"Loaded aggregated data from {json_path}")
             return data
-        except json.JSONDecodeError as e:
+        except orjson.JSONDecodeError as e:
             raise DataLoadError(
                 f"Failed to parse JSON file: {e}", path=str(json_path)
             ) from e
@@ -765,7 +765,7 @@ def _load_gpu_telemetry_jsonl(
             return None
 
         def parse_line(line: str) -> dict:
-            data = json.loads(line)
+            data = orjson.loads(line.encode("utf-8"))
 
             telemetry_data = data.pop("telemetry_data", {})
             flat_record = {**data, **telemetry_data}