rasptorch

Use the rasptorch CLI to create tensors, inspect Vulkan (GPU) availability, build models, and run training.

Security Scan

Instructions are coherent with a CLI-wrapper skill: it only directs an agent to run the rasptorch CLI (or module) and explicitly limits writes/long-running actions to user-approved cases.

Purpose & Capability

The name/description (manage tensors, inspect Vulkan, build/train models) align with the runtime instructions, which call the rasptorch CLI or python -m rasptorch. Required binaries (rasptorch, python/py) are appropriate for this purpose.

Instruction Scope

SKILL.md instructs only CLI invocations and prefers JSON output. It explicitly warns to ask user approval before writing files, launching Streamlit UI, starting long-running servers, or training. It does not instruct reading unrelated files, harvesting environment variables, or sending data to unexpected endpoints.

Install Mechanism

This is an instruction-only skill with no install spec and no bundled code. Nothing is downloaded or written by the skill itself.

Credentials

No environment variables, credentials, or config paths are requested. The declared runtime requirements (presence of rasptorch/python) are proportionate to a CLI integration.

Persistence & Privilege

always is false and the skill does not request elevated or persistent system privileges. It does not modify other skills or system-wide settings. Autonomous invocation is allowed (platform default) but that is expected for skills and not by itself a concern.

Assessment

This skill is a thin wrapper around the rasptorch CLI and appears internally consistent. Before installing or enabling it: 1) ensure you trust the rasptorch binary you will run (the skill executes that binary — if the binary is untrusted it could do anything your account can). 2) Confirm you want the agent to be able to run long-running tasks, start a Streamlit server, or save/load model files; the skill correctly asks for approval for these actions but you should enforce that in your agent policy. 3) If you need stronger isolation, run rasptorch and the agent in a sandboxed environment or inspect the rasptorch package source prior to use.

rasptorch skill

Use this skill when you need to operate rasptorch from an OpenClaw agent by calling its agent-native CLI and consuming JSON output.

Command rule

• Prefer JSON output: always pass --json when returning results to the user/agent.
• Prefer explicit device selection:
  • --device auto (default): use GPU when Vulkan is working, else CPU
  • --device gpu: force GPU (will error / fall back depending on Vulkan availability)
  • --device cpu: force CPU

How to run the CLI

Prefer the installed console script:

rasptorch --json info

If the console script is unavailable, use module invocation from this repo:

python -m rasptorch.CLI.cli --json info

Quick checks (read-only)

Environment + Vulkan status

rasptorch --json info

Key fields:

• vulkan_available
• vulkan_using_real_gpu
• vulkan_status

Tensor workflows

Create tensors

rasptorch --json tensor random --shape 2,3,4 --device cpu --dtype float32
rasptorch --json tensor zeros  --shape 3,4   --device auto
rasptorch --json tensor ones   --shape 5,10  --device gpu

Notes:

• Keep shapes small unless the user explicitly requests large workloads.
• Prefer --device auto unless the user asks to force CPU/GPU.

Model workflows (stateful)

Model commands keep a session-like registry of created models.

Create models

rasptorch --json model linear --input-size 10 --hidden-sizes 32,16 --output-size 2
rasptorch --json model mlp --layers 64,32,16,2
rasptorch --json model cnn --in-channels 3 --out-channels 16,32
rasptorch --json model gru --input-size 8 --hidden-size 32 --num-layers 1
rasptorch --json model transformer --vocab-size 32000 --d-model 256 --num-heads 8 --num-layers 4

List / remove

rasptorch --json model list
rasptorch --json model remove --model-id <model-id>

Save / load (writes files)

Only do this when the user requests it (or approves writing files):

rasptorch --json model save --model-id <model-id> --path ./model.pkl
rasptorch --json model load --path ./model.pkl

Train (writes compute + may take time)

Only run training after user approval.

rasptorch --json model train --model-id <model-id> --epochs 5 --lr 0.001 --batch-size 32 --device auto --optimizer Adam

UI / chat modes

• rasptorch ui starts a Streamlit server (long-running). Only launch with user approval.
• rasptorch chat starts an interactive REPL. Avoid in automated tool flows.