From 2fa5e9b04d9a444119bb83d992d7fce343a5db64 Mon Sep 17 00:00:00 2001
From: Manuel Candales <mcandales@meta.com>
Date: Fri, 28 Jun 2024 11:04:15 -0700
Subject: [PATCH] Update readme

Summary: Updates FACTO readme

Reviewed By: cbilgin

Differential Revision: D59165423

fbshipit-source-id: 3ff2ce4c3b780fd5d243288cc110db533484cc80
---
 README.md | 36 ++++++++++++++++++++++++++++++++++++
 1 file changed, 36 insertions(+)

diff --git a/README.md b/README.md
index f544da7..285521c 100644
--- a/README.md
+++ b/README.md
@@ -2,6 +2,42 @@
 
 Framework for Algorithmic Correctness Testing of Operators
 
+## InputGen
+
+InputGen is a Python library to generate inputs for torch operators, given certain specifications. These specifications can be provided by the user, or can be retrieved from a library of specifications such as SpecDB. They provide a complete description of the space of valid inputs to the operator.
+
+Here is an [overview](inputgen/overview.md) of InputGen
+
+## SpecDB
+
+SpecDB is a [database](specdb/db.py#L30) of specifications covering most of the Core ATen Ops. They have been developed using the ATen CPU kernels as a reference.
+
+## Example Usage
+
+The code below is a minimal example to test add.Tensor using FACTO.
+```python
+import torch
+from inputgen.argtuple.gen import ArgumentTupleGenerator
+from specdb.db import SpecDictDB
+
+# Retrieve the specification from SpecDB
+spec = SpecDictDB["add.Tensor"]
+
+# Initialize generator
+generator = ArgumentTupleGenerator(spec)
+
+op = torch.ops.aten.add.Tensor
+
+# Generate input tuples
+for posargs, inkwargs, outargs in generator.gen():
+    # Evaluate op with given inputs
+    op(*posargs, **inkwargs, **outargs)
+```
+
+## Calibrator
+
+Calibrator is under development. It is intended to be a tool for calibrating the specifications against the behavior of the reference op implementation.
+
 ## Reporting problems
 
 If you encounter a bug or some other problem with FACTO, please file an issue on