[datasets] Add a new Benchmark class.

A benchmark represents that particular program that is being compiled.

Issue facebookresearch#45.

compiler_gym/datasets/BUILD

@@ -6,10 +6,16 @@ load("@rules_python//python:defs.bzl", "py_library")
 
 py_library(
     name = "datasets",
-    srcs = ["__init__.py"],
+    srcs = [
+        "__init__.py",
+        "benchmark.py",
+        "dataset.py",
+    ],
     visibility = ["//visibility:public"],
     deps = [
-        ":dataset",
+        "//compiler_gym:validation_result",
+        "//compiler_gym/service/proto",
+        "//compiler_gym/util",
     ],
 )
 

compiler_gym/datasets/__init__.py

@@ -3,6 +3,11 @@
 # This source code is licensed under the MIT license found in the
 # LICENSE file in the root directory of this source tree.
 """Manage datasets of benchmarks."""
+from compiler_gym.datasets.benchmark import (
+    Benchmark,
+    BenchmarkInitError,
+    BenchmarkSource,
+)
 from compiler_gym.datasets.dataset import (
     LegacyDataset,
     activate,
@@ -11,4 +16,13 @@
     require,
 )
 
-__all__ = ["LegacyDataset", "require", "activate", "deactivate", "delete"]
+__all__ = [
+    "activate",
+    "Benchmark",
+    "BenchmarkInitError",
+    "BenchmarkSource",
+    "deactivate",
+    "delete",
+    "LegacyDataset",
+    "require",
+]