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 · A45 · planned

Smoke runs of the digit probe with a byte-identity repeat, on the base model and on the transplant before and after the tokenizer-loading fix.

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 @ b4c55e6e7d26917e239ccbe2f11886c4d4a733a2

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

Command
python scripts/08_probe_digits.py --model Qwen/Qwen2.5-1.5B --tag smoke_base --smoke 4 --raw-dir results/probe_digits_raw/smoke
Working directory
experiments/E05-control-arm
Configuration paths
None
Parameters
First 4 items per task and format, generated twice and compared byte for byte. Three invocations: smoke_base on Qwen2.5-1.5B; smoke_fvt on ckpts/fvt_t0 before the loading fix, whose first prompt tokenizes to 57 tokens with spaces and newlines lost; smoke_fvt2 after it, 80 tokens.
Environment
3.12.13; PyTorch MPS, bf16; Apple silicon, 128 GB unified memory
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.

Delivered events

  1. Registered

    #1

    Smoke runs of the digit probe with a byte-identity repeat, on the base model and on the transplant before and after the tokenizer-loading fix.

    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.

    • prompt_dump_smoke_base.json · https://github.com/stw2/tokenizer-science-tax/blob/9a57aa40f4e774136f51b6730c33266d59094a45/experiments/E05-control-arm/results/probe_digits_raw/smoke/prompt_dump_smoke_base.json · public · sha256 6c0413e8971b… · 777 bytes · M151
    • prompt_dump_smoke_fvt.json · https://github.com/stw2/tokenizer-science-tax/blob/9a57aa40f4e774136f51b6730c33266d59094a45/experiments/E05-control-arm/results/probe_digits_raw/smoke/prompt_dump_smoke_fvt.json · public · sha256 50acd2ee7de0… · 857 bytes · M152
    • prompt_dump_smoke_fvt2.json · https://github.com/stw2/tokenizer-science-tax/blob/9a57aa40f4e774136f51b6730c33266d59094a45/experiments/E05-control-arm/results/probe_digits_raw/smoke/prompt_dump_smoke_fvt2.json · public · sha256 88ce9137d024… · 859 bytes · M153

    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.