Quickstart & Execution Recipes

Standard usage patterns and rapid prototyping code

6-Modality Rapid Execution Recipes

Select your target modality to view canonical CLI and SDK execution patterns:

# CLI Execution
termux-train diffusion-train --image-dir ./images --prompt "product photo" --output ./adapter.safetensors --epochs 5 --backend vulkan

# Python SDK
import termux_train as tt
tt.diffusion.train_diffusion_lora(image_dir="./images", output_path="./adapter.safetensors", epochs=5, backend="vulkan")

# CLI Execution
termux-train vision-train --data ./vlm.jsonl --output ./vlm_adapter.safetensors --epochs 3 --backend vulkan

# Node.js SDK
import { trainVision } from 'termux-train';
await trainVision({ data: './vlm.jsonl', output: './vlm_adapter.safetensors', epochs: 3, backend: 'vulkan' });

# CLI Execution
termux-train stt-train --data ./audio.jsonl --output ./stt_adapter.safetensors --epochs 4 --backend vulkan

# Python SDK
import termux_train as tt
tt.stt.train_stt_lora(data_source="./audio.jsonl", output_path="./stt_adapter.safetensors", epochs=4, backend="vulkan")

# CLI Execution
termux-train tts-train --data ./voice.jsonl --output ./tts_adapter.safetensors --epochs 5 --backend cpu

# Node.js SDK
import { trainTTS } from 'termux-train';
await trainTTS({ data: './voice.jsonl', output: './tts_adapter.safetensors', epochs: 5, backend: 'cpu' });

# CLI with GPU Slicing
termux-train train --model tiny-transformer --data ./corpus.txt --vocab-slice 4096 --chunk-layers 2 --stream-layers --backend vulkan

# Python SDK
from termux_train import TermuxTrainer
trainer = TermuxTrainer(model_type="lora", lora_rank=8, vocab_slice=4096, chunk_layers=2, backend="vulkan")
trainer.fit(data_path="./corpus.txt", epochs=3, output_path="./llm_adapter.safetensors")

# Fleet Worker Node (e.g. S21, S20)
termux-train cluster-worker --port 50052 --guard-band 300

# Coordinator Node (e.g. S25)
termux-train cluster-probe --fleet 192.168.1.101:50052,192.168.1.102:50052
termux-train train --data ./large.txt --virtual-ram-pool 192.168.1.101:50052,192.168.1.102:50052