adaptai / platform /dataops /dto /integrations /huggingface_config.yaml
ADAPT-Chase's picture
Add files using upload-large-folder tool
fd357f4 verified
# Hugging Face Hub Integration Configuration
# Data Transfer Operations Framework
# Uses HF Hub with Xet backend for artifact storage
# Repository Configuration
repositories:
artifacts:
id: "adaptai/dto-artifacts"
description: "Primary artifact storage for models, checkpoints, and experiments"
type: "model"
private: true
datasets:
id: "adaptai/dto-datasets"
description: "Dataset storage for training and evaluation data"
type: "dataset"
private: true
models:
id: "adaptai/dto-models"
description: "Model storage for trained models and checkpoints"
type: "model"
private: true
# Artifact Management Policies
artifacts:
# Retention policies
retention_days: 90
auto_cleanup: false # HF handles storage management
# Compression settings (handled by HF+Xet)
auto_compress: true
# Integrity verification
checksum_verification: true
checksum_algorithm: "sha256"
# Synchronization Settings
sync:
# Automatic sync intervals (in seconds)
interval: 3600 # 1 hour
# Conflict resolution
conflict_policy: "timestamp" # newest wins
# Bandwidth throttling (handled by HF)
max_bandwidth_mbps: 1000
# Retry settings
max_retries: 3
retry_delay: 30
# Security Configuration
security:
# Authentication
auth_method: "token"
token_env_var: "HF_TOKEN"
# Encryption (handled by HF infrastructure)
encryption_enabled: true
# Access control
read_only: false
# Monitoring and Metrics
monitoring:
# Prometheus metrics
enable_metrics: true
metrics_port: 9092
# Health checks
health_check_interval: 300
# Logging
log_level: "INFO"
log_file: "/var/log/dto-huggingface.log"
# Integration Hooks
hooks:
# Pre-upload hooks
pre_upload: "/opt/dto/hooks/pre-hf-upload.sh"
# Post-upload hooks
post_upload: "/opt/dto/hooks/post-hf-upload.sh"
# Pre-download hooks
pre_download: "/opt/dto/hooks/pre-hf-download.sh"
# Post-download hooks
post_download: "/opt/dto/hooks/post-hf-download.sh"
# Performance Optimization
performance:
# Parallel transfers (handled by HF)
max_parallel_transfers: 8
# Chunk size for large files (handled by HF+Xet)
chunk_size_mb: 64
# Cache settings
cache_enabled: true
cache_size_gb: 50
# Connection pooling
max_connections: 12
connection_timeout: 30
# Xet Backend Configuration (Transparently handled by Hugging Face)
xet_backend:
enabled: true
compression: "zstd"
deduplication: true
versioning: true
# Hugging Face Specific Settings
huggingface:
# API endpoint
endpoint: "https://huggingface.co"
# Timeout settings
timeout: 30
# Rate limiting
requests_per_hour: 5000
# Community features
enable_comments: false
enable_discussions: false