sammcj

#!/usr/bin/env python3
# Repack a split GGUF so one tensor (default: the qwen4exp engram table
# per_layer_token_embd.weight) sits ALONE in the final shard, with no other tensors
# sharing that file. Motivation: on Metal, llama.cpp wires the mmap'd regions that
# back GPU tensors; a tensor interleaved with them in the same file gets wired along
# for the ride (measured +24 GiB on Qwen3.8-Flash-Next). A file containing only
# CPU-side tensors keeps its own mapping and stays pageable.
#
# Tensor bytes are copied verbatim (no requantization). Output layout for an N-file
# input: shard 1 = byte-for-byte copy of input shard 1 (metadata-only; split.count and

sammcj

/**
 * Local Models Extension
 *
 * Manage self-hosted LLM endpoints entirely from the TUI.
 * Commands:
 *   /local-models  - Open the local models manager
 *
 * Config is persisted across sessions. Models show up in /model selector
 * automatically when their endpoint is reachable.
 *

sammcj

#!/usr/bin/env -S uv run --script
# /// script
# requires-python = ">=3.11"
# dependencies = [
#     "gguf>=0.10",
#     "jinja2>=3.1",
#     "numpy",
#     "tqdm",
# ]
# ///

sammcj

# /// script
# requires-python = ">=3.11"
# dependencies = [
#   "mlx>=0.31.1",
#   "mlx-lm @ git+https://github.com/AirRunner/mlx-lm.git@feat/mtp-native",
# ]
# ///
"""MTP vs non-MTP benchmark: measure prefill and generation separately.

Usage:

sammcj

default	21:23:56.820679+1100	handy	[com.apple.controlcenter:698AD2B4-774D-4104-93CC-06FD218815D9-Aux[1]-NSStatusItemView] Sending action(s) in update: NSSceneFenceAction
default	21:23:56.821468+1100	handy	[com.apple.controlcenter:698AD2B4-774D-4104-93CC-06FD218815D9-Aux[1]-NSStatusItemView] Sending action(s) in update: NSSceneFenceAction
default	21:23:56.822402+1100	handy	[com.apple.controlcenter:698AD2B4-774D-4104-93CC-06FD218815D9] Sending action(s) in update: NSSceneFenceAction
error	21:23:56.827498+1100	handy	Warning: -[NSWindow makeKeyWindow] called on <RecordingOverlayPanel: 0xbca502800> windowNumber=f550 which returned NO from -[NSWindow canBecomeKeyWindow].
default	21:23:56.827718+1100	handy	0x11f0ed290 - [PID=47698] ProcessThrottler::setThrottleState: Updating process assertion type to 3 (foregroundActivities=1, backgroundActivities=1)
default	21:23:56.827758+1100	handy	0x11f0ed200 - [PID=47698] WebProcessProxy::didChangeThrottleState: type=2
default	21:23:56.827763+1100	handy	0x11f0ed200 - [PID=4769

sammcj

#!/usr/bin/env bash
# ${HOME}/.openclaw/workspace/tools/skill-audit.sh

set -euo pipefail

# skill-audit.sh - Audit OpenClaw skills for suspicious patterns before installation
# Usage: skill-audit.sh <skill-path-or-url>
# Exit codes: 0 = clean, 1 = suspicious patterns found, 2 = error

readonly SCRIPT_NAME="${0##*/}"

sammcj

services:
  &name vllm:
    <<: [*ai-common, *gpu]
    container_name: *name
    hostname: *name
    profiles:
      - *name
    # image: vllm/vllm-openai:cu130-nightly
    build:

sammcj

{
    "triton_version": "3.5.1",
    "1": {
        "BLOCK_SIZE_M": 64,
        "BLOCK_SIZE_N": 128,
        "BLOCK_SIZE_K": 128,
        "GROUP_SIZE_M": 32,
        "num_warps": 8,
        "num_stages": 3
    },

sammcj

AM2R PortMaster Setup Guide

This guide explains how to install AM2R v1.5.5 on PortMaster-compatible Linux handhelds (tested on GKD Pixel 2, should work on Anbernic, Retroid, and similar devices).

Required Files

You need three zip files. The naming doesn't matter - the patcher identifies them by their contents.

File	Size	Description	Source

sammcj

ZWNobyAiYzNWa2J5QnliU0F0Y21ZZ0x3PT0iIHwgYmFzZTY0IC1kIHwgc2gK