agentleFS
Sign inStart free
Seeed-ProjectsSeeed-Projects/Seeed-Jetson-DevelopTool/skills/claude/deepseek-quick-deploy/CLAUDE.md

Seeed-Jetson-DevelopTool / deepseek-quick-deploy

CLAUDE.md57 starsChanged 4 months agoMITView on GitHub

Quickly deploy DeepSeek-R1 on Jetson using Ollama. Two-command setup. Requires Jetson with >8GB RAM and JetPack 5.1.1+.

  • Pipes a download into a shell

The file

We split this CLAUDE.md into the rules it gives an agent

4 rules. About 0 are common, 1 is distinctive, and 3 are very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈1

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 3

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Distinctive:

    Run one phase at a time. After each phase:

    line 14
  2. Very specific:

    If output contains [STOP] → stop immediately, consult the failure decision tree below.

    line 16
  3. Very specific:

    If output ends with [OK] → tell the user "Phase N complete" and proceed to the next phase.

    line 17
  4. Very specific:

    Test with a simple prompt to confirm the model responds.

    line 63

≈ “Common” is an estimate: at least 1 in 5 of the 26,980 CLAUDE.md files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
CLAUDE.md

DeepSeek-R1 Quick Deploy on Jetson

Hardware required: Jetson with >8GB RAM, JetPack 5.1.1+. This is the fastest path to running DeepSeek-R1 locally. For optimized MLC-based deployment with better throughput, see the deploy-deepseek-mlc skill instead.

Execution model

Run one phase at a time. After each phase:

  • Relay all command output to the user.
  • If output contains [STOP] → stop immediately, consult the failure decision tree below.
  • If output ends with [OK] → tell the user "Phase N complete" and proceed to the next phase.

Prerequisites

RequirementDetail
RAM>8 GB (unified memory)
JetPack5.1.1 or later
InternetRequired for install script and model download
Disk space≥10 GB free for model weights

Phase 1 — Install Ollama

curl -fsSL https://ollama.com/install.sh | sh

Verify:

ollama --version
# Expected: ollama version x.x.x

[OK] when ollama --version prints a version string. [STOP] if the install script fails — see failure decision tree.

Phase 2 — Run DeepSeek-R1

ollama run deepseek-r1

The first run downloads the model weights (~4–8 GB depending on quantization). Subsequent runs start immediately from cache.

Expected: after download, an interactive prompt appears:

>>> Send a message (/? for help)

Test with a simple prompt to confirm the model responds. Type /bye to exit.

[OK] when the model responds to a prompt. [STOP] if the process is killed or hangs — see failure decision tree.

Failure decision tree

SymptomAction
curl install script fails with network errorCheck internet connectivity. If behind a proxy, set https_proxy env var before running curl. Try downloading the script manually and inspecting it.
curl install script fails with permission errorRun with sudo or ensure /usr/local/bin is writable.
Model download stalls or fails mid-wayRetry ollama run deepseek-r1 — Ollama resumes partial downloads. Check available disk space: df -h.
Process killed during model pull (OOM)Not enough RAM. Free memory by stopping other processes, or use a smaller quantization: ollama run deepseek-r1:1.5b.
Model loads but inference is very slowExpected on smaller Jetson modules. For better performance use the deploy-deepseek-mlc skill which uses MLC-optimized kernels.
ollama: command not found after installAdd Ollama to PATH: export PATH=$PATH:/usr/local/bin, then reload shell or open a new terminal.

Reference files

  • references/source.body.md — full source tutorial with additional context (reference only)

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

No reviews yet

People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.

  • It worked
  • Partly
  • It didn't
What did you use it for? e.g. Reviewing pull requests in a Rails app

Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.

Your trail

Files you open will collect here, so you can find them again.

Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.

Now do this for your own.

Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.

Check your setup, free