More verbose logging when downloading large datasets.

Issue facebookresearch#445.

compiler_gym/datasets/tar_dataset.py

@@ -92,9 +92,12 @@ def install(self) -> None:
             # Remove any partially-completed prior extraction.
             shutil.rmtree(self.site_data_path / "contents", ignore_errors=True)
 
-            logger.info("Downloading %s dataset", self.name)
+            logger.warning(
+                "Installing the %s dataset. This may take a few moments ...", self.name
+            )
+
             tar_data = io.BytesIO(download(self.tar_urls, self.tar_sha256))
-            logger.info("Unpacking %s dataset", self.name)
+            logger.info("Unpacking %s dataset to %s", self.name, self.site_data_path)
             with tarfile.open(
                 fileobj=tar_data, mode=f"r:{self.tar_compression}"
             ) as arc:

compiler_gym/third_party/llvm/__init__.py

@@ -4,6 +4,7 @@
 # LICENSE file in the root directory of this source tree.
 """Module for resolving paths to LLVM binaries and libraries."""
 import io
+import logging
 import shutil
 import sys
 import tarfile
@@ -16,6 +17,8 @@
 from compiler_gym.util.download import download
 from compiler_gym.util.runfiles_path import cache_path, site_data_path
 
+logger = logging.getLogger(__name__)
+
 # The data archive containing LLVM binaries and libraries.
 _LLVM_URL, _LLVM_SHA256 = {
     "darwin": (
@@ -37,6 +40,10 @@
 
 def _download_llvm_files(destination: Path) -> Path:
     """Download and unpack the LLVM data pack."""
+    logger.warning(
+        "Installing the CompilerGym LLVM environment runtime. This may take a few moments ..."
+    )
+
     # Tidy up an incomplete unpack.
     shutil.rmtree(destination, ignore_errors=True)