CLI Commands¶
Generated: December 12, 2025
Purpose: Reference for the aios CLI and subcommands
Status: Implemented
Overview¶
- Main entry point:
aios - File:
src/aios/cli/aios.py
Sub-commands overview: - hrm-hf – HuggingFace-based HRM training (see Core Training) - brains – Brain management - gui – Launch GUI (see GUI Features) - status – System status - datasets – Dataset management (see Datasets) - cache – Cache management - goals – Goals management - eval – Evaluation utilities (see Advanced Features → Evaluation) - artifacts – Artifacts management - cleanup – Cleanup utilities - crawl – Web crawling (see Tools & Integrations) - optimization – Optimization utilities (see Memory Optimization) - modelcard – Model card generation - agent – Agent commands - budgets – Budget management (see Advanced Features → Budgets) - core – Core commands - hf-cache – HuggingFace cache management - dml – DirectML utilities
HRM-HF Training¶
- Command:
aios hrm-hf - File:
src/aios/cli/hrm_hf_cli.py - Subcommand:
train-actv1– Train HRM models with ACT v1 - File:
src/aios/cli/hrm_hf/train_actv1.py - Deep dive: See Core Training and Memory Optimization docs
Key parameters (selection):
- Model: --model <name_or_path>
- Brain naming: --brain-name, --bundle-dir
- Training control: --steps, --batch-size, --lr, --max-seq-len, --iterate, --stop-file, --resume, --stop-after-epoch
- Architecture: --h-layers, --l-layers, --hidden-size, --expansion, --num-heads, --h-cycles, --l-cycles, --halt-max-steps, --window-size, --pos-encodings
- Memory optimization: --gradient-checkpointing|--no-gradient-checkpointing, --amp|--no-amp, --use-8bit-optimizer, --use-chunked-training, --chunk-size, --cpu-offload
- Dataset: --dataset-file, --ascii-only, --linear-dataset, --dataset-start-offset, --dataset-chunk-size
- Evaluation: --eval-file, --eval-batches, --log-file
- Multi-GPU: --ddp, --cuda-ids, --world-size, --parallel-independent, --strict
- DeepSpeed: --zero-stage <none|zero1|zero2|zero3>
- MoE: --use-moe, --num-experts, --num-experts-per-tok, --moe-capacity-factor, --auto-adjust-lr
- PEFT: --use-peft, --peft-method, --lora-r, --lora-alpha, --lora-dropout, --lora-target-modules
- Precision/Quant: --model-dtype fp32|fp16|bf16, --load-in-8bit, --load-in-4bit
- Inference hot‑reload: --inference-device, --hot-reload-steps
Brains Management¶
- Command:
aios brains - File:
src/aios/cli/brains.py - Subcommands: list, load, info, delete, export, import
- Related: Core Training → Brain Bundle System, GUI Features → Brains Panel
Datasets Management¶
- Command:
aios datasets - File:
src/aios/cli/datasets_cli.py - Features: list/download, scan, metadata, verification
- Related: Datasets doc
Goals Management¶
- Command:
aios goals - File:
src/aios/cli/goals_cli.py - Create/list/activate goals, link to experts, goal-driven training.
- Related: Dynamic Subbrains/MoE and Advanced Features → Orchestrator
Cache and HF Cache¶
aios cache→ Clear/show statsaios hf-cache→ Location, move, clear, size reporting
Evaluation¶
- Command:
aios eval - File:
src/aios/cli/eval_cli.py - Run evaluations, generate reports, compare models.
- Related: Advanced Features → Evaluation
Crawling¶
- Command:
aios crawl - File:
src/aios/cli/crawl_cli.py - Web crawling, dataset generation from web.
- Related: Tools & Integrations
Optimization¶
- Command:
aios optimization - File:
src/aios/cli/optimization_cli.py - Memory/VRAM estimation and parameter optimization.
- Related: Memory Optimization
Back to Feature Index: COMPLETE_FEATURE_INDEX.md • Back to Guide Index: ../INDEX.MD