requirements.txt

# ── Web & API ──────────────────────────────────────────────────────────────
fastapi==0.115.12
uvicorn[standard]==0.34.2
pydantic==2.11.5
pydantic-settings==2.9.1
httpx==0.28.1
python-dotenv==1.0.1

# ── Data collection ─────────────────────────────────────────────────────────
feedparser==6.0.11
beautifulsoup4==4.12.3
lxml==5.3.0

# ── Data processing ─────────────────────────────────────────────────────────
pandas==2.2.3
numpy==2.2.6
scikit-learn==1.6.1
datasets==3.6.0

# ── ML core (install CUDA wheel separately for GPU training) ─────────────────
transformers==4.53.0
accelerate==1.7.0
peft==0.15.2
trl==0.9.6
bitsandbytes==0.45.5
# torch — install manually for your CUDA version:
# pip install torch --index-url https://download.pytorch.org/whl/cu124

# ── Distributed training (optional) ─────────────────────────────────────────
# deepspeed==0.16.7   # Linux/CUDA only — uncomment if using DeepSpeed

# ── Evaluation & export ──────────────────────────────────────────────────────
evaluate==0.4.3
sacrebleu==2.5.1
rouge-score==0.1.2
sentencepiece==0.2.0
onnx==1.17.0
onnxruntime==1.22.0

# ── HuggingFace publishing ───────────────────────────────────────────────────
huggingface_hub==0.30.2

# ── Demo & MLOps ─────────────────────────────────────────────────────────────
gradio==5.33.0
mlflow==2.22.0
wandb==0.19.11

# ── Dev & testing ────────────────────────────────────────────────────────────
pytest==8.3.5
pytest-asyncio==0.26.0
ruff==0.11.11