$ curl -fsSL https://ollama.com/install.sh | sh >>> Installing ollama to /usr/local [sudo] minimonk 암호: >>> Downloading ollama-linux-amd64.tar.zst ######################################################################## 100.0% >>> Creating ollama user... >>> Adding ollama user to render group... >>> Adding ollama user to video group... >>> Adding current user to ollama group... >>> Creating ollama systemd service... >>> Enabling and starting ollama service... Created symlink /etc/systemd/system/default.target.wants/ollama.service → /etc/systemd/system/ollama.service. >>> The Ollama API is now available at 127.0.0.1:11434. >>> Install complete. Run "ollama" from the command line. WARNING: No NVIDIA/AMD GPU detected. Ollama will run in CPU-only mode.
$ ollama Ollama 0.20.3
Chat with a model Start an interactive chat with a model
Launch Claude Code (not installed) Anthropic's coding tool with subagents
Recommended gemma4 Reasoning and code generation locally, ~16GB, (not downloaded) qwen3.5 Reasoning, coding, and visual understanding locally, ~11GB, (not downloaded) kimi-k2.5:cloud Multimodal reasoning with subagents qwen3.5:cloud Reasoning, coding, and agentic tool use with vision glm-5:cloud Reasoning and code generation minimax-m2.7:cloud Fast, efficient coding and real-world productivity
More ▸ llama3.2
↑/↓ navigate • enter select • ← back
$ ollama --help Large language model runner
Usage: ollama [flags] ollama [command]
Available Commands: serve Start Ollama create Create a model show Show information for a model run Run a model stop Stop a running model pull Pull a model from a registry push Push a model to a registry signin Sign in to ollama.com signout Sign out from ollama.com list List models ps List running models cp Copy a model rm Remove a model launch Launch the Ollama menu or an integration help Help about any command
Flags: -h, --help help for ollama --nowordwrap Don't wrap words to the next line automatically --verbose Show timings for response -v, --version Show version information
Use "ollama [command] --help" for more information about a command.
단발성으로 질문하고 속도를 보려고 하는데
음.. eval rate가 실제 체감되는 속도인가? 12 token/s
이정도 수치는 되어야 많이 지루하지 않게 읽는 속도로 말이 생성되는 듯.
$ ollama run llama3.2:latest "안녕" --verbose 안녕하세요! (Hello!) I'm here to help you. How can I assist you today?
1. **Analyze the Request:** The user asked "너에 대해서 소개해봐" (Tell me about yourself). 2. **Determine Identity & Core Facts (Self-Correction/Verification):** * Name: Gemma 4. * Developer: Google DeepMind. * Nature: Large Language Model (LLM). * Type: Open weights model. * Capabilities: Understands text/images, generates text. * Knowledge Cutoff: (Relevant for context, but usually stated implicitly) 3. **Structure the Introduction (Korean Context):** A good introduction should cover: * Greeting. * Name/Identity. * Developer/Origin. * Nature/Function (What I do). * Capabilities (What I can process). 4. **Draft the Response (Internal Monologue & Refinement):** * *Start warmly.* (안녕하세요!) * *State Identity.* (저는 Gemma 4입니다.) * *State Origin.* (Google DeepMind에서 개발했습니다.) * *State Nature.* (저는 대규모 언어 모델(LLM)입니다.) * *State Function/Role.* (사용자의 질문에 답하고, 정보를 제공하며, 다 양한 텍스트 기반 작업을 돕는 역할을 합니다.) * *State Modality/Input.* (텍스트와 이미지를 이해할 수 있습니다.) * *Concluding remark/Offer help.* (어떤 것을 도와드릴까요?) 5. **Final Output Generation (Review against constraints):** The generated response must adhere to the core identity (Gemma 4, LLM, Google DeepMind, open weights). (The draft looks good and adheres to the required identity.) ...done thinking.
안녕하세요! 저는 **Gemma 4**입니다.
저는 Google DeepMind에서 개발한 **대규모 언어 모델(LLM)**이며, 오픈 웨이트 (open weights) 모델로 공개되어 있습니다.
**저의 주요 특징은 다음과 같습니다:**
1. **정체성:** 저는 방대한 데이터를 학습하여 인간의 언어를 이해하고 생성하며 , 다양한 질문에 답하고 정보를 제공하는 능력을 가지고 있습니다. 2. **기능:** 텍스트 기반의 대화, 요약, 번역, 작문, 코드 생성 등 다양한 언어 관련 작업을 수행할 수 있습니다. 3. **입력 능력:** 저는 텍스트와 이미지를 이해하고 처리할 수 있습니다. (다만 , 출력은 텍스트로만 생성합니다.) 4. **목표:** 사용자분들이 필요한 정보를 얻고, 창의적인 작업을 수행하며, 복잡 한 문제를 해결하는 데 도움을 드릴 수 있도록 설계되었습니다.
Features Support three 18650 batteries (not included), 7800mAh large capacity, up to 12.6V output, stronger motor power. On-board S-8254AA + AO4407A Li-ion battery protection circuit, with anti-overcharge, anti-over-discharge, anti-over-current and short-circuit protection functions. Onboard APW7313 voltage regulator chip, which can provide stable 5V voltage to Jetson Nano. Onboard TB6612FNG dual H-bridge motor driver chip can drive the left and right two motor work. Onboard 0.91" 128×32 resolution OLED, real-time display of car IP address, memory, power, and other conditions. Onboard INA219 acquisition chip, convenient for real-time monitoring of battery voltage.
IMX219 (v2 camera) is supported by the legacy camera stack, as are OV5647 (v1) and IMX477 (HQ) cameras.
IMX708 (v3) and IMX296 (Global Shutter) are NOT supported, nor are ANY cameras on a Pi5. It's libcamera only for those options.
I'd boot Raspberry Pi OS properly and check "vcgencmd get_camera" to ensure that your camera is detected properly. With Bullseye you should be able to run the old raspistill app to confirm that it all works. Only once you've validated that would I move on to trying your bare metal approach.
