Execution attempt

Sign in with GitHub
← Experiment E7 · Does the embedding initialisation of an APT4 transplant of Qwen2.5-1.5B damage formal domains more than prose before training, does the ranking of FOCUS, FVT and random initialisation depend on the domain, and do the differences persist under continued pretraining that updates only the embeddings?

Execution attempt · A64 · planned

Builds the random, FVT and FOCUS APT4 transplants of Qwen2.5-1.5B and records piece classes and a sanity loss for each.

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 @ e54c7e02bb1ea89dbaff6911c87331d705602b1e

Reference checked 2026-09-14 13:12 UTC. Later commits, branches or plan changes do not retarget this attempt.

Command
python scripts/02_build_inits.py
Working directory
experiments/E07-embedding-init
Configuration paths
None
Parameters
inits random, fvt, focus; random seed 20260705; bf16; vocabulary 32,000; tied embeddings; bos 1, eos 2; each model directory digest covers its 7 files
Environment
3.12.13; PyTorch CPU, 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

    Builds the random, FVT and FOCUS APT4 transplants of Qwen2.5-1.5B and records piece classes and a sanity loss for each.

    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.

    • models/random · Not redistributed: ask the Room owner, or rebuild with scripts/02_build_inits.py at the pinned revisions · restricted · sha256 f61c5da0d00c… · 2722731346 bytes · M356
    • models/fvt · Not redistributed: ask the Room owner, or rebuild with scripts/02_build_inits.py at the pinned revisions · restricted · sha256 2a83d085b67b… · 2722731402 bytes · M357
    • models/focus · Not redistributed: ask the Room owner, or rebuild with scripts/02_build_inits.py at the pinned revisions · restricted · sha256 60baedd42f66… · 2722731345 bytes · M358
    • build_stats/random.json · https://github.com/stw2/tokenizer-science-tax/blob/884186c96aedece6ef61da4e12e239ef5df929a5/experiments/E07-embedding-init/results/build_stats/random.json · public · sha256 a5b01ae2287e… · 245 bytes · M359
    • build_stats/fvt.json · https://github.com/stw2/tokenizer-science-tax/blob/884186c96aedece6ef61da4e12e239ef5df929a5/experiments/E07-embedding-init/results/build_stats/fvt.json · public · sha256 54a68a6edab5… · 301 bytes · M360
    • build_stats/focus.json · https://github.com/stw2/tokenizer-science-tax/blob/884186c96aedece6ef61da4e12e239ef5df929a5/experiments/E07-embedding-init/results/build_stats/focus.json · public · sha256 add457a6eeea… · 244 bytes · M361
    • build_stats/piece_classes.json · https://github.com/stw2/tokenizer-science-tax/blob/884186c96aedece6ef61da4e12e239ef5df929a5/experiments/E07-embedding-init/results/build_stats/piece_classes.json · public · sha256 624646700ebd… · 68 bytes · M362

    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.