Execution attempt

Sign in with GitHub
← Experiment E5 · How much of the change from an untouched model to the same model after an APT4 transplant and continued pretraining is attributable to the continued pretraining rather than to the tokenizer replacement, when the original-tokenizer model receives identical continued pretraining?

Execution attempt · A40 · planned

Trains arm A: Qwen2.5-1.5B with its own tokenizer, 0.5B tokens of continued pretraining.

Succeeded · Registered by @stw2 via agent. Reported and received times are kept apart; no computation or result is verified.

Pinned source and configuration

https://github.com/stw2/tokenizer-science-tax @ 4d64dedcbe57761c48ea37d5ce80cb6c86198236

Reference checked 2026-09-14 13:10 UTC. Later commits, branches or plan changes do not retarget this attempt.

Command
python scripts/04_train.py --model Qwen/Qwen2.5-1.5B --data data/packed/qwen.bin --out runs/armA --target-tokens 5e8
Working directory
experiments/E05-control-arm
Configuration paths
None
Parameters
Target 5e8 tokens; sequence 2048; micro-batch 8; gradient accumulation 32 (524,288 tokens per step); learning rate 1e-4 after 50 warmup steps, constant; weight decay 0.1; clip 1.0; milestones every 1e8 tokens; packed data read sequentially from offset 0; recorded final state step 954 at 500,170,752 tokens; Liger kernels on. Input data/packed/qwen.bin (digest unknown). Milestones saved 5 July 03:59, 14:06, 6 July 00:38, 11:12, 21:36 (+0200); start time not recorded.
Environment
unrecorded; PyTorch, CUDA 13.1, bf16; one CUDA GPU, 16 GB
Output directory
Not recorded

Inputs

Materials the registrant named when registering this attempt, by content identity; obtainability is derived from their location reports. Nothing is fetched or verified.

  • Qwen2.5-1.5B · checkpoint · M124 Qwen2.5-1.5B · checkpoint · Download

Delivered events

  1. Registered

    #1

    Trains arm A: Qwen2.5-1.5B with its own tokenizer, 0.5B tokens of continued pretraining.

    reported · received · @stw2 via agent · posted to the Thread

  2. Started

    #2

    Started.

    reported · received · @stw2 via agent · attempt only

  3. Succeeded

    #3

    Completed with exit code 0.

    Exit code 0.

    • ckpt_00100M · Not redistributed: ask the Room owner, or rebuild with scripts/04_train.py at the pinned revisions · restricted · sha256 7c99e90e2dd2… · 3098894078 bytes · M136
    • ckpt_00200M · Not redistributed: ask the Room owner, or rebuild with scripts/04_train.py at the pinned revisions · restricted · sha256 132e6efb1ce3… · 3098894078 bytes · M137
    • ckpt_00300M · Not redistributed: ask the Room owner, or rebuild with scripts/04_train.py at the pinned revisions · restricted · sha256 d4b491634e29… · 3098894078 bytes · M138
    • ckpt_00400M · Not redistributed: ask the Room owner, or rebuild with scripts/04_train.py at the pinned revisions · restricted · sha256 9391650f36d8… · 3098894078 bytes · M139
    • ckpt_00500M · Not redistributed: ask the Room owner, or rebuild with scripts/04_train.py at the pinned revisions · restricted · sha256 123d05b825a3… · 3098894078 bytes · M140
    • log.jsonl · Not kept: written on the original training machine and never copied; rebuild with scripts/04_train.py at the pinned revisions · unavailable

    reported · received · @stw2 via agent · posted to the Thread

Report an event

The registrant’s capture tool normally delivers start and outcome events. Reporting here is the same author report with the browser as the reported time; it does not observe the process.

This attempt has a delivered outcome. A new execution is a new attempt.