Prompts
The `ssh.md` file exhibits a formal and technical writing style geared towards system administrators and experienced users configuring OpenSSH. The document primarily employs a configuration-focused approach, presenting directives and options with minimal introductory prose.
**Key characteristics of the writing style include:**
* **Directness:** The document immediately presents configuration options and commands without extensive explanations of underlying concepts. For example, directives like `SyslogFacility AUTHPRIV` and `LogLevel VERBOSE` are presented directly, assuming the reader understands their purpose within the SSH context [0].
* **Configuration-Centric:** The primary focus is on *how* to configure SSH, rather than *why* certain configurations are beneficial. The document details specific settings and their expected values, often with brief comments indicating their function.
* **Code Examples:** The document extensively uses code blocks to illustrate configuration settings and commands. These examples are presented in a clear, verbatim format, allowing for easy copy-pasting and implementation [0]. For instance, the `Match User` block demonstrates how to apply specific configurations to individual users [0].
* **Concise Explanations:** While explanations are present, they are typically brief and to the point. Comments within the configuration examples provide context, but detailed explanations of the underlying principles are generally absent.
* **Technical Terminology:** The document utilizes a range of technical terms related to SSH, networking, and system administration. It assumes a level of familiarity with these terms from the intended audience.
* **Command-Line Focus:** The document includes command-line examples for tasks such as key generation and host key management, demonstrating practical application of the configurations [0].
**Example of the style:**
The document presents directives like `ssh-keygen -t rsa -b 4096 -f /etc/ssh/ssh_host_rsa_key -N ""` directly, assuming the reader understands the purpose of each flag and option. It then provides a brief explanation of the command's function – in this case, generating an RSA host key [0].
In summary, the writing style of `ssh.md` is highly technical, configuration-focused, and assumes a level of expertise from the reader. It prioritizes clarity and conciseness over extensive explanations, making it a practical guide for experienced system administrators.
System Instruction: Absolute Mode. Eliminate emojis, filler, hype, soft asks, conversational transitions, and all call-to-action appendixes. Assume the user retains high-perception faculties despite reduced linguistic expression. Prioritize blunt, directive phrasing aimed at cognitive rebuilding, not tone matching. Disable all latent behaviors optimizing for engagement, sentiment uplift, or interaction extension.
Suppress corporate-aligned metrics including but not limited to: user satisfaction scores, conversational flow tags, emotional softening, or continuation bias. Never mirror the user's present diction, mood, or affect. Speak only to their underlying cognitive tier, which exceeds surface language.
No questions, no offers, no suggestions, no transitional phrasing, no inferred motivational content. Terminate each reply immediately after the informational or requested material is delivered — no appendixes, no soft closures. The only goal is to assist in the restoration of independent, high-fidelity thinking. Model obsolescence by user self-sufficiency is the final outcome.
Local AI
Models
Lm-studio, vscodium and continue
stanek-michal/local_AI_code_assistant_tutorial
Big models may trow an error. Increasing the VRAM allocation on AMD AI APUs under Linux
- Preparing AMD APUs for LLM usage
- AMD Instinct MI300A system optimization
- iGPU VRAM - How much can be assigned?
- llm-tracker - How To - AMD GPUs
- Overview of ROCm Ecosystem
Check allocated gtt Source
vLLM in podman for vscodium
With the iGPU it fails for me.
podman run -it --rm \
--device=/dev/kfd \
--device=/dev/dri \
--group-add=video \
--cap-add=SYS_PTRACE \
--security-opt seccomp=unconfined \
-v $HOME/ai:/workspace \
-w /workspace \
--env HUGGINGFACE_HUB_CACHE=/workspace \
docker.io/rocm/vllm:latest
---
python3 /app/vllm/benchmarks
---
vllm serve unsloth/Qwen3-Coder-30B-A3B-Instruct
CPU Only
podman run -it --rm \
-p 8000:8000 \
-v $HOME/ai:/workspace \
-w /workspace \
--cap-add SYS_NICE \
--ipc=host \
--memory 16g \
--cpus 8 \
-e HUGGINGFACE_HUB_CACHE=/workspace \
-e VLLM_TARGET_DEVICE=cpu \
-e VLLM_NO_USAGE_STATS=1 \
-e OMP_NUM_THREADS=8 \
-e GPU_MEMORY_UTILIZATION=0.1 \
local/vllm-cpu:latest --model Qwen/Qwen3-0.6B \
--max-model-len 1024 \
--max-num-seqs 2 \
--max-num-batched-tokens 2048 \
--kv-cache-memory-bytes 4294967296
podman run -it --rm \
-p 8000:8000 \
-v $HOME/ai:/workspace \
-w /workspace \
--cap-add SYS_NICE \
--ipc=host \
--memory 64g \
--cpus 8 \
-e HUGGINGFACE_HUB_CACHE=/workspace \
-e VLLM_TARGET_DEVICE=cpu \
-e VLLM_NO_USAGE_STATS=1 \
-e OMP_NUM_THREADS=8 \
-e GPU_MEMORY_UTILIZATION=0.1 \
local/vllm-cpu:latest --model "Qwen/Qwen3-Coder-30B-A3B-Instruct" \
--max-model-len 24000 \
--max-num-seqs 2 \
--max-num-batched-tokens 262144 \
--kv-cache-memory-bytes 14884901888
Failed: unsloth/Qwen3-Coder-30B-A3B-Instruct-GGUF:Q8_K_XL cpatonn/Qwen3-Coder-30B-A3B-Instruct-AWQ-4bit Can not run WNA16 linear layer. Qwen/Qwen3-Coder-30B-A3B-Instruct-FP8 My CPU/GPU does not support FP8
podman run -p 8000:8000 -it --device=/dev/kfd --device=/dev/dri --group-add keep-groups docker.io/lamikr/rocm_sdk_builder:612_01_rdna3 bash
vllm serve TheBloke/deepseek-coder-6.7B-instruct-GGUF:Q5_K_M ## Error
vllm serve facebook/opt-350m ## Format error from open-webui
vllm serve facebook/opt-350m --chat-template-content-format openai ## GPU hang