GGUF-Splitter / src /processing.py
Felladrin's picture
Initial commit
2de2584
"""Main processing logic for GGUF splitting"""
import logging
import pathlib
import re
import shutil
import subprocess
import tempfile
import time
from typing import Optional
import gradio as gr
from huggingface_hub import HfApi
from .config import DOWNLOAD_TIMEOUT, MAX_DOWNLOAD_RETRIES, RUN_LOCALLY
from .gguf_utils import calculate_optimal_split_size, split_gguf_file
from .hf_utils import (
check_repo_exists,
extract_quantization,
extract_username,
validate_repo_id,
)
from .logging_config import setup_logging
try:
from huggingface_hub.errors import HFValidationError
except ImportError:
from huggingface_hub.utils import HFValidationError
logger = setup_logging()
def update_status(status_markdown: str, text: str) -> str:
"""Update status display"""
return f"{status_markdown}\n\n{text}"
def _validate_download(gguf_path: pathlib.Path) -> None:
"""Validate that download completed successfully"""
logger.info(f"Validating download: {gguf_path}")
if not gguf_path.exists():
logger.error("Download validation failed - file not found")
raise Exception("Download failed - file not found")
if gguf_path.stat().st_size == 0:
logger.error("Download validation failed - file is empty")
raise Exception("Download failed - file is empty")
logger.info(f"Download validation successful: {gguf_path.stat().st_size} bytes")
def save_locally(
split_files: list[pathlib.Path],
source_repo_id: str,
gguf_filename: str,
max_size_mb: int,
output_dir: pathlib.Path,
status_callback=None,
) -> str:
"""Save split files locally with proper organization"""
if status_callback:
status_callback(f"💾 Saving {len(split_files)} split files locally...")
model_name = source_repo_id.split("/")[-1]
model_name = re.sub(r"-?GGUF$", "", model_name, flags=re.IGNORECASE)
model_name = model_name.rstrip("-")
sharded_dir_name = f"{model_name}-sharded"
sharded_path = output_dir / sharded_dir_name
sharded_path.mkdir(exist_ok=True)
for split_file in split_files:
destination = sharded_path / split_file.name
shutil.move(str(split_file), str(destination))
readme_content = _generate_readme_content(
source_repo_id, len(split_files), max_size_mb
)
readme_path = sharded_path / "README.md"
readme_path.write_text(readme_content)
if status_callback:
status_callback(f"✅ Files saved to {sharded_path}")
return str(sharded_path)
def _cleanup_temp_files(
gguf_path: pathlib.Path,
split_files: list[pathlib.Path],
sharded_path: Optional[pathlib.Path] = None,
) -> None:
"""Clean up temporary files"""
logger.info("Cleaning up temporary files")
if gguf_path.exists():
try:
gguf_path.unlink()
logger.info(f"Removed original GGUF file: {gguf_path}")
except (OSError, PermissionError) as e:
logger.warning(f"Failed to remove GGUF file {gguf_path}: {e}")
for split_file in split_files:
if split_file.exists():
try:
split_file.unlink()
logger.info(f"Removed split file: {split_file}")
except (OSError, PermissionError) as e:
logger.warning(f"Failed to remove split file {split_file}: {e}")
if sharded_path and sharded_path.exists():
try:
shutil.rmtree(sharded_path)
logger.info(f"Removed sharded directory: {sharded_path}")
except (OSError, PermissionError) as e:
logger.warning(f"Failed to remove sharded directory {sharded_path}: {e}")
logger.info("Temporary file cleanup completed")
def upload_to_hf(
split_files: list[pathlib.Path],
repo_name: str,
source_repo_id: str,
oauth_token: gr.OAuthToken,
max_size_mb: int,
make_public: bool,
status_callback=None,
) -> str:
"""Upload split files to Hugging Face Hub"""
if status_callback:
status_callback(
f"☁️ Uploading {len(split_files)} split files to Hugging Face..."
)
api = HfApi(token=oauth_token.token)
try:
repo_url = api.create_repo(repo_name, private=not make_public, exist_ok=True)
if status_callback:
status_callback(f"✅ Repository created: {repo_url}")
except Exception as e:
if status_callback:
status_callback(f"❌ Failed to create repository: {str(e)}")
return ""
try:
for i, split_file in enumerate(split_files):
if status_callback:
status_callback(
f"📤 Uploading file {i + 1}/{len(split_files)}: {split_file.name}"
)
api.upload_file(
path_or_fileobj=str(split_file),
path_in_repo=split_file.name,
repo_id=repo_name,
)
readme_content = _generate_readme_content(
source_repo_id, len(split_files), max_size_mb
)
if status_callback:
status_callback("📄 Uploading README.md...")
api.upload_file(
path_or_fileobj=readme_content.encode(),
path_in_repo="README.md",
repo_id=repo_name,
)
if status_callback:
status_callback(
f"✅ Upload completed! Repository: https://huggingface.co/{repo_name}"
)
return f"https://huggingface.co/{repo_name}"
except Exception as e:
if status_callback:
status_callback(f"❌ Upload failed: {str(e)}")
return ""
def _download_with_retry(
gguf_url: str,
gguf_path: pathlib.Path,
max_retries: int = MAX_DOWNLOAD_RETRIES,
status_callback=None,
) -> bool:
"""Download file with retry logic and exponential backoff"""
logger.info(f"Starting download attempt for {gguf_url}")
for attempt in range(max_retries):
try:
download_cmd = [
"curl",
"-L",
"--fail",
"--max-time",
str(DOWNLOAD_TIMEOUT),
"--progress-bar",
gguf_url,
"-o",
str(gguf_path),
]
if status_callback:
status_callback(
f"📥 Downloading file (attempt {attempt + 1}/{max_retries})..."
)
subprocess.run(download_cmd, check=True, capture_output=False)
if gguf_path.exists() and gguf_path.stat().st_size > 0:
logger.info(f"Download successful: {gguf_path.stat().st_size} bytes")
if status_callback:
status_callback(
f"✅ Download successful ({gguf_path.stat().st_size} bytes)"
)
return True
else:
raise Exception("Download validation failed - file empty or missing")
except subprocess.CalledProcessError as e:
logger.error(f"Download attempt {attempt + 1} failed: {e}")
if status_callback:
status_callback(f"⚠️ Download attempt {attempt + 1} failed, retrying...")
if attempt < max_retries - 1:
wait_time = 2**attempt
logger.info(f"Waiting {wait_time}s before retry")
if status_callback:
status_callback(f"⏳ Waiting {wait_time}s before retry...")
time.sleep(wait_time)
except Exception as e:
logger.error(f"Download validation failed: {e}")
if status_callback:
status_callback(f"⚠️ Download validation failed: {e}")
logger.error("All download attempts failed")
if status_callback:
status_callback("❌ All download attempts failed")
return False
def _generate_readme_content(
source_repo_id: str, split_count: int, max_size_mb: int
) -> str:
"""Generate README content for split GGUF files"""
return f"""---
base_model: {source_repo_id}
---
Sharded GGUF version of [{source_repo_id}](https://huggingface.co/{source_repo_id}).
"""
def process_split_request(
repo_name: str,
source_repo_id: str,
gguf_filename: str,
make_public: bool,
oauth_token: Optional[gr.OAuthToken],
status_display: str,
) -> str:
"""Main function to handle GGUF splitting and upload/save"""
gguf_path: Optional[pathlib.Path] = None
split_files: list[pathlib.Path] = []
sharded_path: Optional[pathlib.Path] = None
tmp_path: Optional[pathlib.Path] = None
api: Optional[HfApi] = None
username: Optional[str] = None
output_dir: Optional[pathlib.Path] = None
try:
if not RUN_LOCALLY and not oauth_token:
return update_status(status_display, "❌ Please sign in first.")
if not source_repo_id or not gguf_filename:
return update_status(
status_display, "❌ Please select a model and GGUF file."
)
if not RUN_LOCALLY and (not repo_name or not repo_name.strip()):
return update_status(status_display, "❌ Please provide a repository name.")
try:
validate_repo_id(source_repo_id)
except HFValidationError as e:
return update_status(
status_display, f"❌ Invalid source repository ID format: {str(e)}"
)
status_display = update_status(
status_display, "⏳ Validating repository exists..."
)
try:
if RUN_LOCALLY:
temp_api = HfApi()
elif oauth_token is not None:
temp_api = HfApi(token=oauth_token.token)
else:
return update_status(status_display, "Please sign in first.")
if not check_repo_exists(source_repo_id, temp_api):
return update_status(
status_display,
"Source repository does not exist or is not accessible.",
)
except Exception:
return update_status(
status_display, "Unable to verify repository existence."
)
if not RUN_LOCALLY:
if oauth_token is None:
return update_status(status_display, "Please sign in first.")
api = HfApi(token=oauth_token.token)
user_info = api.whoami()
username = extract_username(user_info)
if not username:
return update_status(
status_display, "❌ Unable to determine your Hugging Face username."
)
status_display = update_status(status_display, "⏳ Downloading GGUF file...")
if RUN_LOCALLY:
output_dir = pathlib.Path("./output")
output_dir.mkdir(exist_ok=True)
tmp_path = output_dir / f"temp_{int(time.time())}"
tmp_path.mkdir(exist_ok=True)
else:
tmp_dir = tempfile.mkdtemp()
tmp_path = pathlib.Path(tmp_dir)
def status_callback(message: str) -> None:
nonlocal status_display
status_display = update_status(status_display, message)
logger.info(f"Status update: {message}")
gguf_url = (
f"https://huggingface.co/{source_repo_id}/resolve/main/{gguf_filename}"
)
gguf_path = tmp_path / gguf_filename
if not _download_with_retry(
gguf_url, gguf_path, status_callback=status_callback
):
return update_status(status_display, "❌ Failed to download GGUF file")
_validate_download(gguf_path)
status_callback("⏳ Calculating optimal split size...")
model_name = source_repo_id.split("/")[-1]
model_name = re.sub(r"-?GGUF$", "", model_name, flags=re.IGNORECASE)
model_name = model_name.rstrip("-")
quantization = extract_quantization(gguf_filename)
output_prefix = tmp_path / f"{model_name}-{quantization}"
max_size_mb = calculate_optimal_split_size(str(gguf_path), str(output_prefix))
status_callback(f"⏳ Splitting GGUF file with max size {max_size_mb}M...")
output_pattern = tmp_path / f"{model_name}-{quantization}"
if not split_gguf_file(
str(gguf_path),
str(output_pattern),
max_size_mb,
status_callback=status_callback,
):
return update_status(status_display, "❌ Failed to split GGUF file")
split_files = list(tmp_path.glob(f"{model_name}-{quantization}-*.gguf"))
if not split_files:
return update_status(status_display, "❌ No split files generated")
if RUN_LOCALLY:
if output_dir is None:
return update_status(status_display, "❌ Output directory error.")
status_display = update_status(
status_display, f"⏳ Saving {len(split_files)} split files locally..."
)
quantization = extract_quantization(gguf_filename)
final_output_dir = output_dir / f"{model_name}_{quantization}"
final_output_dir.mkdir(exist_ok=True)
for split_file in split_files:
target = final_output_dir / split_file.name
shutil.copy2(split_file, target)
readme_content = _generate_readme_content(
source_repo_id, len(split_files), max_size_mb
)
with open(final_output_dir / "README.md", "w") as f:
f.write(readme_content)
_cleanup_temp_files(gguf_path, split_files)
tmp_path.rmdir()
success_message = f"""✅ GGUF file split and saved locally!
📂 Output directory: {final_output_dir.absolute()}
🔍 Created {len(split_files)} split files with max size {max_size_mb}M each
💾 Files saved to local disk.
"""
return update_status(status_display, success_message)
else:
if api is None or username is None or oauth_token is None:
return update_status(status_display, "❌ Authentication error.")
status_display = update_status(
status_display, "⏳ Creating new repository..."
)
if "/" in repo_name:
new_repo_name = repo_name
else:
new_repo_name = f"{username}/{repo_name}"
api.create_repo(
repo_id=new_repo_name,
repo_type="model",
exist_ok=True,
private=not make_public,
)
status_display = update_status(
status_display, f"⏳ Uploading {len(split_files)} split files..."
)
sharded_path = tmp_path / "sharded"
sharded_path.mkdir(exist_ok=True)
for split_file in split_files:
target = sharded_path / split_file.name
target.symlink_to(split_file)
api.upload_file(
path_or_fileobj=str(split_file),
path_in_repo=split_file.name,
repo_id=new_repo_name,
token=oauth_token.token,
)
status_display = update_status(status_display, "⏳ Creating README...")
readme_content = _generate_readme_content(
source_repo_id, len(split_files), max_size_mb
)
api.upload_file(
path_or_fileobj=readme_content.encode(),
path_in_repo="README.md",
repo_id=new_repo_name,
token=oauth_token.token,
)
status_display = update_status(
status_display, "🧹 Cleaning up temporary files..."
)
_cleanup_temp_files(gguf_path, split_files, sharded_path)
success_message = f"""✅ GGUF file split and uploaded successfully!
📂 New repository: https://huggingface.co/{new_repo_name}
🔍 Created {len(split_files)} split files with max size {max_size_mb}M each
🧹 Temporary files cleaned up successfully!
"""
return update_status(status_display, success_message)
except Exception as e:
return update_status(status_display, f"❌ Error: {str(e)}")
finally:
if tmp_path and tmp_path.exists() and not RUN_LOCALLY:
try:
shutil.rmtree(tmp_path)
except Exception:
pass