Module molcrawl.preparation.download_guacamol

GuacaMol Dataset Download Script

Download the GuacaMol benchmark dataset from Figshare. https://figshare.com/projects/GuacaMol/56639

How to use: python src/preparation/download_guacamol.py

environmental variables: LEARNING_SOURCE_DIR: Base directory where dataset is saved (required)

Functions

def download_file(url, output_path, chunk_size=8192)

Expand source code

def download_file(url, output_path, chunk_size=8192):
    """
    Download and save file from URL.

    Falls back to copying from a sibling learning_source_* directory when the
    remote server blocks automated downloads (e.g. Figshare WAF challenge).

    Args:
        url: Download source URL
        output_path: save the first file
        chunk_size: Chunk size (bytes)
    """
    import shutil as _shutil

    output_path = Path(output_path)

    # Skip if already exists and has content (0-byte files are treated as missing)
    if output_path.exists() and output_path.stat().st_size > 0:
        print(f"✓ Already exists: {output_path.name}")
        return True
    if output_path.exists() and output_path.stat().st_size == 0:
        print(f"⚠ Found 0-byte file, re-downloading: {output_path.name}")
        output_path.unlink()

    print(f"Downloading {output_path.name}...")

    try:
        headers = {
            "User-Agent": (
                "Mozilla/5.0 (X11; Linux x86_64) AppleWebKit/537.36 "
                "(KHTML, like Gecko) Chrome/122.0.0.0 Safari/537.36"
            )
        }
        response = requests.get(url, stream=True, timeout=60, headers=headers)
        response.raise_for_status()

        # get file size
        total_size = int(response.headers.get("content-length", 0))

        # Download with progress bar
        with (
            open(output_path, "wb") as f,
            tqdm(
                desc=output_path.name,
                total=total_size,
                unit="B",
                unit_scale=True,
                unit_divisor=1024,
            ) as pbar,
        ):
            for chunk in response.iter_content(chunk_size=chunk_size):
                if chunk:
                    f.write(chunk)
                    pbar.update(len(chunk))

        # Verify we actually got content (WAF blocks return 0 bytes with 200/202)
        if output_path.stat().st_size == 0:
            output_path.unlink()
            raise requests.exceptions.RequestException(
                "Server returned 0 bytes — possible WAF/bot-protection block."
            )

        print(f"✓ Downloaded: {output_path.name}")
        return True

    except requests.exceptions.RequestException as e:
        print(f"✗ HTTP download failed: {e}", file=sys.stderr)
        # delete failed files
        if output_path.exists():
            output_path.unlink()

        # Fallback: copy from an existing learning_source_* directory
        existing = _find_existing_smiles(output_path.name)
        if existing:
            print(f"  ↪ Copying from existing dataset: {existing}", file=sys.stderr)
            _shutil.copy2(existing, output_path)
            print(f"✓ Copied: {output_path.name} ({output_path.stat().st_size:,} bytes)")
            return True

        print(
            f"  No local fallback found for {output_path.name}.\n"
            f"  Download manually from {url} and place in {output_path.parent}/",
            file=sys.stderr,
        )
        return False

Download and save file from URL.

Falls back to copying from a sibling learning_source_* directory when the remote server blocks automated downloads (e.g. Figshare WAF challenge).

Args

url

Download source URL

output_path

save the first file

chunk_size

Chunk size (bytes)

def download_guacamol(compounds_dir)

Expand source code

def download_guacamol(compounds_dir):
    """
    Download GuacaMol dataset

    Args:
        compounds_dir: compounds directorypath (example: learning_source_XXX/compounds）

    Raises:
        RuntimeError: If download fails
    """
    import logging

    logger = logging.getLogger(__name__)

    output_dir = Path(compounds_dir) / "benchmark" / "GuacaMol"
    output_dir.mkdir(parents=True, exist_ok=True)

    logger.info("Downloading GuacaMol benchmark from https://figshare.com/projects/GuacaMol/56639")
    logger.info(f"Destination: {output_dir}")

    success_count = 0
    total_count = len(GUACAMOL_URLS)

    for split, url in GUACAMOL_URLS.items():
        filename = f"guacamol_v1_{split}.smiles"
        output_path = output_dir / filename

        if download_file(url, output_path):
            success_count += 1

    if success_count < total_count:
        logger.warning(
            f"Figshare download incomplete ({success_count}/{total_count}). "
            "Trying HuggingFace fallback..."
        )
        if not _download_from_huggingface(output_dir):
            raise RuntimeError(
                f"GuacaMol download incomplete: {success_count}/{total_count} files downloaded "
                "and HuggingFace fallback also failed."
            )
        logger.info("✓ GuacaMol: All files obtained via HuggingFace fallback.")
    else:
        logger.info(f"✓ GuacaMol: All {total_count} files downloaded successfully")

Download GuacaMol dataset

Args

compounds_dir

compounds directorypath (example: learning_source_XXX/compounds）

Raises

RuntimeError

If download fails

def main()

Expand source code

def main():
    """Download GuacaMol dataset (for standalone execution)"""

    # Check LEARNING_SOURCE_DIR
    learning_source_dir = os.environ.get("LEARNING_SOURCE_DIR")
    if not learning_source_dir:
        print(
            "ERROR: Environment variable 'LEARNING_SOURCE_DIR' is not set.",
            file=sys.stderr,
        )
        print(
            "Please set LEARNING_SOURCE_DIR before running this script:",
            file=sys.stderr,
        )
        print("  export LEARNING_SOURCE_DIR='learning_20251104'", file=sys.stderr)
        sys.exit(1)

    compounds_dir = Path(learning_source_dir) / "compounds"

    try:
        download_guacamol(str(compounds_dir))
        print("\nNext steps:")
        print("  1. Run the GPT-2 preparation script:")
        print(
            f"     LEARNING_SOURCE_DIR={learning_source_dir} python src/compounds/dataset/prepare_gpt2.py assets/configs/compounds.yaml"
        )
        return 0
    except RuntimeError as e:
        print(f"\n✗ {e}", file=sys.stderr)
        return 1

Download GuacaMol dataset (for standalone execution)