# ========================================
# Training Frameworks
# ========================================
omegaconf>=2.3.0  # Fix metadata issue with older versions
ms-swift[all]>=3.0.0
deepspeed==0.18.2
transformers>=4.51.0
accelerate>=0.30.0
peft>=0.10.0  # For LoRA/QLoRA if needed in future

# ========================================
# Acceleration & Optimization
# ========================================
vllm==0.11.2
# flash-attn>=2.0.0  # Install separately after environment setup: MAX_JOBS=4 pip install flash-attn --no-build-isolation
flashinfer-python==0.5.2

# ========================================
# Monitoring & Logging
# ========================================
tensorboard>=2.14.0
wandb>=0.15.0  # Optional: Weights & Biases for experiment tracking
tensorboardX>=2.6.0  # Extended TensorBoard features

# ========================================
# Data Processing & Utilities
# ========================================
datasets>=2.18.0
tokenizers>=0.15.0
numpy>=1.24.0,<2.0  # Pin to 1.x for vLLM/transformers compatibility
pandas>=2.0.0
tqdm>=4.65.0
sentencepiece>=0.2.0  # For some tokenizers
protobuf>=3.20.0
faiss-cpu==1.13.0
jsonlines
loguru
sentence_transformers
unstructured[local-inference]>=0.10.0
beautifulsoup4>=4.12.0
nltk>=3.8.0  # Required by unstructured

# ========================================
# Configuration & Development
# ========================================
pyyaml>=6.0
hydra-core>=1.3.0  # Optional: Advanced config management
python-dotenv>=1.0.0  # Environment variable management
rich>=13.0.0
typer>=0.9.0

# Core dependencies
instructor>=1.0.0
openai>=1.0.0
anthropic>=0.18.0
pydantic>=2.0.0,<2.12  # Constrained for gradio compatibility
duckdb
pyarrow

# NLP and text processing
spacy>=3.7.0
# Note: Install spaCy model separately after installation:
# python -m spacy download en_core_web_sm

# Temporal processing
dateparser>=1.2.0
pendulum>=3.0.0

# Spatial processing
geopy>=2.4.0
folium>=0.20.0  # Required for map visualization
geopandas>=0.14.0

# ========================================
# Distributed Training Utilities
# ========================================
mpi4py>=3.1.0  # For multi-node training
ninja>=1.11.0  # Fast C++ builds (for custom kernels)

# ========================================
# Model Serving & Inference (Future Use)
# ========================================
fastapi>=0.100.0  # For API serving
uvicorn>=0.23.0  # ASGI server



# MCP (Model Context Protocol)
mcp>=1.2.0

# Visualization
matplotlib>=3.7.0
seaborn>=0.12.0
plotly>=5.17.0

# Development tools
pytest>=7.4.0
pytest-cov>=4.1.0
black>=23.0.0
isort>=5.12.0
flake8>=6.0.0
mypy>=1.5.0
gdown