tinymind-native-8b-remote-handoff/native_axiom_regenesis_8b_remote_train_from_hf_bucket.ipynb

{
  "cells": [
    {
      "cell_type": "markdown",
      "metadata": {},
      "source": [
        "# TinyMind AxiomReGenesis 8B Target Remote Training\n",
        "This notebook trains the 8.22B-class native target. It does not claim quality until probes pass.\n"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {},
      "outputs": [],
      "source": [
        "!pip -q install -U 'huggingface_hub[hf_transfer]'\n",
        "!rm -rf /content/tinymind_remote && mkdir -p /content/tinymind_remote\n",
        "!hf sync hf://buckets/bbkdevops/unicosys-hypergraph-bucket/tinymind-native-8b-remote-handoff /content/tinymind_remote/handoff\n",
        "!unzip -q /content/tinymind_remote/handoff/tinymind_native_8b_remote_bundle.zip -d /content/tinymind_remote\n",
        "%cd /content/tinymind_remote/bundle\n"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {},
      "outputs": [],
      "source": [
        "!pip -q install -r requirements.txt || true\n",
        "import torch, sys\n",
        "print('python', sys.version)\n",
        "print('cuda available', torch.cuda.is_available())\n",
        "print('gpu', torch.cuda.get_device_name(0) if torch.cuda.is_available() else 'cpu')\n"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {},
      "outputs": [],
      "source": [
        "python -m train.cli native-axiom-regenesis-train --dataset /content/tinymind_remote/bundle/data/omni_round_curriculum.jsonl --out-dir /content/tinymind_axiom_regenesis_8b_target --max-steps 50000 --eval-records 256 --limit-records 10000 --dim 2816 --layers 48 --lanes 64 --seq-len 1024 --vocab-size 4096 --tokenizer-mode char_v1 --virtual-dim 1048576 --basis-rank 256 --facets 64 --learning-rate 1.2e-05 --train-batch-size 1 --device cuda\n"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {},
      "outputs": [],
      "source": [
        "python -m train.cli native-baseline-probe --out-dir /content/tinymind_axiom_regenesis_8b_probe --native-checkpoint /content/tinymind_axiom_regenesis_8b_target/checkpoint.pt --baseline-report /content/tinymind_remote/bundle/baseline/deep_core_probe_report.json --max-new-tokens 160 --device cuda\n"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {},
      "outputs": [],
      "source": [
        "python -m train.cli native-broad-probe --out-dir /content/tinymind_axiom_regenesis_8b_broad_probe --native-checkpoint /content/tinymind_axiom_regenesis_8b_target/checkpoint.pt --max-new-tokens 192 --device cuda\n"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {},
      "outputs": [],
      "source": [
        "!zip -qr /content/tinymind_axiom_regenesis_8b_results.zip /content/tinymind_axiom_regenesis_8b_target /content/tinymind_axiom_regenesis_8b_probe /content/tinymind_axiom_regenesis_8b_broad_probe\n",
        "files.download('/content/tinymind_axiom_regenesis_8b_results.zip')\n"
      ]
    }
  ],
  "metadata": {
    "kernelspec": {
      "display_name": "Python 3",
      "language": "python",
      "name": "python3"
    }
  },
  "nbformat": 4,
  "nbformat_minor": 5
}