Getting Started

# From PyPI (recommended)
pip install soup-cli

# Or from GitHub (latest dev)
pip install git+https://github.com/MakazhanAlpamys/Soup.git

pip install 'soup-cli[fast]'       # Unsloth 2-5x training speedup
pip install 'soup-cli[serve]'      # FastAPI inference server
pip install 'soup-cli[serve-fast]' # vLLM backend
pip install 'soup-cli[sglang]'     # SGLang backend
pip install 'soup-cli[eval]'       # lm-evaluation-harness
pip install 'soup-cli[data]'       # MinHash deduplication
pip install 'soup-cli[vision]'     # Pillow for vision fine-tuning
pip install 'soup-cli[audio]'      # librosa + soundfile
pip install 'soup-cli[qat]'        # Quantization-aware training
pip install 'soup-cli[liger]'      # Liger Kernel fused ops
pip install 'soup-cli[onnx]'       # ONNX export
pip install 'soup-cli[tensorrt]'   # TensorRT-LLM export
pip install 'soup-cli[ui]'         # Web UI dashboard
pip install 'soup-cli[wandb]'      # W&B logging
pip install 'soup-cli[deepspeed]'  # ZeRO distributed training
pip install 'soup-cli[ring-attn]'  # Ring FlashAttention

# Interactive wizard
soup init

# Or use a template
soup init --template chat

soup train --config soup.yaml

soup chat --model ./output

soup push --model ./output --repo your-username/my-model

# Export to GGUF for Ollama / llama.cpp
soup export --model ./output --format gguf --quant q4_k_m

# Start OpenAI-compatible server
soup serve --model ./output --port 8000

soup doctor

soup quickstart          # Creates sample data + config + trains TinyLlama
soup quickstart --dry-run # Just create files without training