Merge pull request #2 from FasterAI-Labs/release/v0.1.0

Release v0.1.0

Author: nathanhubens

CHANGELOG.md

@@ -0,0 +1,29 @@
+# Changelog
+
+## v0.1.0 (2026-03-24)
+
+First stable release of fasterbench.
+
+### Features
+
+- **Unified `benchmark()` API** — Single entry point returning typed `BenchmarkResult` with 5 metric categories
+- **Size metrics** — Parameter count, disk size via `compute_size()`
+- **Speed metrics** — Latency, throughput, batch/thread/latency sweeps via `compute_speed_multi()`
+- **Compute metrics** — MACs/FLOPs via thop and torchprofile backends via `compute_compute()`
+- **Memory metrics** — CPU (psutil) and GPU (CUDA) memory tracking via `compute_memory_multi()`
+- **Energy metrics** — Power consumption and carbon footprint via codecarbon via `compute_energy_multi()`
+- **`LayerProfiler`** — Per-layer profiling with hook-based measurement (speed, memory, size, compute)
+- **Radar plots** — Multi-model comparison visualization via `create_radar_plot()`
+- **`BenchmarkResult`** — Supports both typed access (`result.size.size_mib`) and dict access (`result["size_mib"]`)
+- **Serialization** — `.as_dict()`, `.to_dataframe()`, `.to_json()` on all result types
+
+### Infrastructure
+
+- Migrated to nbdev3 with `pyproject.toml` (PEP 621)
+- CI via GitHub Actions (nbdev3-ci workflow)
+- Documentation via Quarto + GitHub Pages
+- Suppressed spurious logging from thop and codecarbon
+
+## v0.0.1
+
+Initial development release.

fasterbench/__init__.py

@@ -1,11 +1,13 @@
-__version__ = "0.0.1"
 """Comprehensive benchmarking toolkit for deep learning models"""
 
 # AUTOGENERATED! DO NOT EDIT! File to edit: ../nbs/index.ipynb.
 
 # %% auto #0
 __all__ = []
 
+# %% ../nbs/index.ipynb #version_cell
+__version__ = "0.1.0"
+
 # %% ../nbs/index.ipynb #8b6f8c52
 from .benchmark import benchmark, BenchmarkResult
 from .size import SizeMetrics, compute_size, get_model_size, get_num_parameters

nbs/index.ipynb

@@ -20,6 +20,17 @@
     "#| default_exp __init__"
    ]
   },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "id": "version_cell",
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "#| export\n",
+    "__version__ = \"0.1.0\""
+   ]
+  },
   {
    "cell_type": "code",
    "execution_count": null,

pyproject.toml

@@ -10,8 +10,8 @@ readme = "README.md"
 requires-python = ">=3.7"
 license = {text = "Apache-2.0"}
 authors = [{name = "nathanhubens", email = "nathan.hubens@gmail.com"}]
-keywords = ['nbdev', 'jupyter', 'notebook', 'python']
-classifiers = ["Natural Language :: English", "Intended Audience :: Developers", "Development Status :: 3 - Alpha", "Programming Language :: Python :: 3", "Programming Language :: Python :: 3 :: Only"]
+keywords = ['benchmark', 'deep-learning', 'pytorch', 'profiling', 'latency', 'energy']
+classifiers = ["Natural Language :: English", "Intended Audience :: Developers", "Development Status :: 4 - Beta", "Programming Language :: Python :: 3", "Programming Language :: Python :: 3 :: Only"]
 dependencies = ['torch', 'fastcore', 'prettytable', 'tqdm', 'codecarbon', 'torchprofile', 'thop', 'plotly']
 
 [project.urls]
@@ -28,5 +28,6 @@ version = {attr = "fasterbench.__version__"}
 include = ["fasterbench"]
 
 [tool.nbdev]
+version = "0.1.0"
 tst_flags = 'notest'
 jupyter_hooks = true