Text Generation
PEFT
Safetensors
MLX
English
lora
mixture-of-experts
hobbylm
agent
macos
tool-use
computer-use
on-device
Instructions to use Hobbylm/hobbylm-1B-computer-use with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- PEFT
How to use Hobbylm/hobbylm-1B-computer-use with PEFT:
from peft import PeftModel from transformers import AutoModelForCausalLM base_model = AutoModelForCausalLM.from_pretrained("harims95/hobbylm-1B-instruct") model = PeftModel.from_pretrained(base_model, "Hobbylm/hobbylm-1B-computer-use") - MLX
How to use Hobbylm/hobbylm-1B-computer-use with MLX:
# Make sure mlx-lm is installed # pip install --upgrade mlx-lm # if on a CUDA device, also pip install mlx[cuda] # Generate text with mlx-lm from mlx_lm import load, generate model, tokenizer = load("Hobbylm/hobbylm-1B-computer-use") prompt = "Once upon a time in" text = generate(model, tokenizer, prompt=prompt, verbose=True) - Notebooks
- Google Colab
- Kaggle
- Local Apps Settings
- LM Studio
- MLX LM
How to use Hobbylm/hobbylm-1B-computer-use with MLX LM:
Generate or start a chat session
# Install MLX LM uv tool install mlx-lm # Generate some text mlx_lm.generate --model "Hobbylm/hobbylm-1B-computer-use" --prompt "Once upon a time"
- Atomic Chat
Download scripts/setup.sh from Hobbylm/hobbylm-1B-computer-use: direct link, hf CLI and curl.
- Browser
- Download file 1.99 kB
-
https://huggingface.co/Hobbylm/hobbylm-1B-computer-use/resolve/main/scripts/setup.sh
- Command line
-
hf download hf://Hobbylm/hobbylm-1B-computer-use/scripts/setup.sh
-
curl -L -o setup.sh https://huggingface.co/Hobbylm/hobbylm-1B-computer-use/resolve/main/scripts/setup.sh
1.99 kB
| # Hobby setup for Apple Silicon Macs. Downloads pinned model files, clones/builds the pinned macos-use server, builds Swift tools. | |
| set -euo pipefail | |
| cd "$(dirname "$0")/.." | |
| mkdir -p data models vendor | |
| python3 -m venv .venv | |
| .venv/bin/pip install -q "mlx==0.32.3" "mlx-lm==0.31.3" "transformers==5.17.0" "tokenizers==0.23.2" "huggingface-hub==1.33.0" \ | |
| "numpy==2.5.3" "safetensors==0.8.0" | |
| # FP32 Instruct weights (sha256 d0899e0a...); skipped if already present | |
| [ -f models/hobbylm-1B-instruct/model.safetensors ] || .venv/bin/hf download harims95/hobbylm-1B-instruct \ | |
| --revision 8a90f6eb6a3d7c7eb952fd01b512a406821366ab --local-dir models/hobbylm-1B-instruct | |
| # MLX runtime (executed code) + tokenizer, pinned revision; the 3.7 GB MLX weights are skipped (Hobby runs FP32 + adapter) | |
| [ -d models/hobbylm-1B-mlx/runtime ] || .venv/bin/hf download harims95/hobbylm-1B-mlx \ | |
| --revision a2192174572dca075f8378db70cb3e8e70bfe521 --local-dir models/hobbylm-1B-mlx \ | |
| --include 'runtime/*' --include '*.json' --include '*.txt' # one --include per pattern; extra bare names would be read as files | |
| [ -d models/hobbylm-1B-mlx/runtime/mlx_hobbylm ] || { echo "MLX runtime download failed (models/hobbylm-1B-mlx/runtime)" >&2; exit 1; } | |
| # macos-use (Business Source License 1.1, see its LICENSE), pinned commit | |
| [ -d vendor/mcp-server-macos-use ] || git clone https://github.com/mediar-ai/mcp-server-macos-use vendor/mcp-server-macos-use | |
| git -C vendor/mcp-server-macos-use checkout -q 205c8880e8c06fa96a3e3a859e6edca728eb6c47 | |
| (cd vendor/mcp-server-macos-use && swift build -c release) | |
| # small Swift tools: checkers and system controls | |
| for t in calc_checker brightness media_key checker; do swiftc -O hobby/$t.swift -o data/$t; done | |
| swiftc -O hobby/HobbyBar.swift -o data/HobbyBar # Hobby bar: Spotlight-style panel on Shift-Command-H (⇧⌘H) | |
| echo "Done. Grant Accessibility to your terminal app, then: data/HobbyBar (⇧⌘H), or .venv/bin/python hobby/hobby_demo.py r4" | |