kklemon

These are compressed handoff notes for a group of agents trying to get a Factorio factory ready to launch a rocket. They mix production planning, experiments, inventory accounting, and instructions about who may control the game.

The central distinction is “tested in a disposable copy” versus “actually accomplished in the live game.” Much of the repetition exists to keep those separate. The immediate crisis develops toward the end: a destroyed coal belt shuts down most steam power, and the agents eventually discover that construction robots cannot reach it.

I’m interpreting the supplied notes, without inspecting the referenced files. I’ve joined wrapped lines and abbreviated long hashes in the quotations.

Some recurring shorthand:

Shorthand	Meaning

kklemon

# Multi-stage build for a ROCm + MIGraphX + ONNX Runtime image on Ubuntu 24.04.
#
# This Dockerfile intentionally uses:
# - A remote TheRock checkout as the ROCm source tree.
# - A separate AMDMIGraphX build, because this TheRock checkout does not
#   currently build/install MIGraphX itself.
# - ONNX Runtime's MIGraphX execution provider, not the deprecated ROCm EP.
#
# Build example:
#   docker build \

kklemon

{
  "version": "0.1.1",
  "notes": "",
  "platforms": {
    "windows-x86_64": {
      "signature": "",
      "url": ""
    },
    "darwin-x86_64": {
      "signature": "",

kklemon

from contextlib import asynccontextmanager
import os
from fastapi import FastAPI
from langchain_openai import ChatOpenAI
from sqlmodel import Field, Session, SQLModel, create_engine, select
from pydantic.json_schema import SkipJsonSchema


# class BaseJoke(SQLModel):
#     joke: str

kklemon

kklemon

import torch
import nibabel as nib
import numpy as np
from tensordidct import TensorDict, MemmapTensor

files_by_modality = {
    'flair': [...]
}

num_files = ...

kklemon

import torch.nn as nn
import torch.nn.functional as F


class FlashAttentionTransformerEncoder(nn.Module):
    def __init__(
        self,
        dim_model,
        num_layers,
        num_heads=None,

kklemon

import time
import torch
import torch.nn as nn
import torch.nn.functional as F


bz = 128
seq_len = 512
d_model = 64
n_heads = 8

kklemon

"""
Sample code for residue coloring in a dash-bio 3d molecule view

Requirements:
- dash
- dash-bio
- numpy
- biopython

pip install dash dash-bio numpy biopython

kklemon

import torch
import torch.distributed as dist
import torch.multiprocessing as mp

from torch.utils.data import IterableDataset, DataLoader


class DistributedIterableDataset(IterableDataset):
    """
    Example implementation of an IterableDataset that handles both multiprocessing (num_workers > 0)